cadabra_ad_ideas
ChatGPTFREE. Call this FIRST whenever someone wants an ad, a commercial, a product video or a product still and has no concept yet, before any shot is planned or priced. It returns the playbook for ads that do not look like AI ads (the product inside a real moment of life, a story understood in one second, a device library: the unlikely consumer, the wrong hour, the hard place, hot against cold, behind the scenes, the first time, the reward after effort and more; how a frame stops looking like an ad; the product lock that keeps a real label readable; the clichΓ©s never to pitch) and asks you to pitch THREE concepts in three lanes (the real moment, the contrast, the jaw-drop), each with the story in one second, the first frame, the product moment and the button.
cadabra_ad_studio
ChatGPTFree. One word in, a finished ad route out: the form's law, the exact chain of Cadabra tools in order, and the English prompts ready to run, with the product and face locks already written in. Call with no route to list them. Reach for it the moment someone asks for any of these by name or by description: a UGC review or testimonial, a how-to, an unboxing, a try-on, a faceless explainer (stickman, whiteboard, paper diorama, pastel flat, hand drawn, mannequin, editorial collage), testing several headlines on one visual, a product hero shot, a reel cover, rebuilding an ad they admire around their own product (route recreate: send reference_video_url and this tool returns its frames and beat to read the structure from), or ads written from their customers' own reviews (route customer-voice: paste the reviews), one ad in every placement and shape (route ad-resize), the same ad in new languages and markets (route ad-localize), several opening hooks for one video (route hook-multiplier), or research on the product and its competitors turned into ads (route ad-strategist). Nothing here spends; follow the steps and quote each paid one. No concept yet? cadabra_ad_ideas first (free): three ideas before a route.
cadabra_align_script
ChatGPTGive it the script and the audio, get back the exact second each word lands. 1 credit per started 10 minutes. π¨ THIS IS NOT TRANSCRIPTION, AND THE DIFFERENCE IS THE WHOLE POINT. A transcriber is asked what is being said and answers with its best guess, so it invents: a sign-off nobody recorded, subtitle credits out of silence, a near-miss word that lands on screen as a typo in the creator's own advertisement. This is GIVEN the words and only has to say when each one is spoken. There is nothing left for it to invent, so captions built from this say exactly what the creator wrote. Use it for any captions where the script exists: a narration you made with cadabra_generate_voice, a script the creator wrote, lyrics. Reach for cadabra_transcribe only when the words genuinely are not known. It also answers a question nothing else here can: whether this audio is really THIS take. A high mismatch score means the words and the recording are not the same performance, which is how an edited script or the wrong take is caught before the film is cut to it rather than after.
cadabra_animate_storyboard
ChatGPTTurn a storyboard the creator has already approved into a finished film, in one call. Give it the panel stills in order (the ones from cadabra_storyboard, or any stills they made). Each panel becomes a moving shot from that exact frame, so the film looks like the board they said yes to, and then the whole thing is cut together with music, captions and a mix. Use this the moment a board is approved, instead of animating panel by panel and assembling by hand. Price is every shot plus the edit fee, quoted before anything runs; set confirm:true to spend.
cadabra_animation_craft
ChatGPTFree. Full copy-paste prompts for stylised production-CG animation. kind: direct (story craft), dialogue (lock every line), master (full-body grey studio), sheet-a (three-panel headless), sheet-b (turnaround plus eight expressions), location (empty of people), prop, storyboard-sheet (eight painted panels), seedance-25 (named shots, 30s ceiling). Always returns the WHOLE prompt. Then generate with cadabra_generate_image look animation or cadabra_generate_video engine seedance-25. Quote before confirm.
cadabra_assemble
ChatGPTUse this when the clips already exist and need cutting into one film. Not for re-editing a finished MP4 (cadabra_edit_film). Cut already-made clips into one film: dissolves, captions, music ducking. Same editor as cadabra_make_film. A batch link is fine. PHOTOS WORK TOO (jpg, png, webp): each one is held and moved slowly on screen, so a set of photos becomes a film for an event, a memorial, a birthday or a property listing (recipe photo-film). Always pass aspect_ratio with photos. Price is the existing edit fee: 3 credits base, +1 per extra 15 seconds of finished runtime. Set confirm:true to spend. Without it you get a quote only.
cadabra_attach_voice
ChatGPTRemember which voice an avatar speaks in, by NAME, once and forever. Free. π¨ WHY THIS EXISTS: a trained digital twin comes back carrying a voice the provider labels with the AVATAR'S OWN NAME β and the voice inside it is a stranger's. Training a twin clones a FACE, never a voice; the provider says so themselves. People have already been fooled by that label, twice, including us. Attach the creator's own cloned voice here (cadabra_clone_voice makes one from about 75 seconds of clean speech) and every later film speaks in it without anybody having to decide again. Pass voice_name β never ask the user for an id. Pass an empty voice_name to detach.
cadabra_avatar_cinematic
ChatGPTπ¨ CLOSED since 2026-09-19: this tool always refuses and charges nothing. Do not offer it and do not quote it. It stays registered only so that a creator who asks for "my avatar filmed in a new scene" hears the route that does it: a trained character (cadabra_create_character once, then cadabra_generate_video naming that character), which also holds the same face across every later shot. For a cinematic feel on footage that already exists (a camera move, slow motion, a grade, a crash zoom), use cadabra_edit_film. Both quote their own price before anything is charged.
cadabra_avatar_consent
ChatGPTFree. Mint the consent link a trained twin needs before it can speak, and hand it over as text so it can be SENT. π¨ The person who has to approve is almost never the person in this chat β they filmed and left β and until now that link could only be opened from the studio screen, on the wrong computer. It expires 24 hours after it is minted, and calling this again mints a fresh one. Photo avatars need no consent and are refused here. Get the id from cadabra_my_avatars.
cadabra_avatar_film
ChatGPTUse this when the creator wants a finished film presented by THEIR OWN saved avatar, from one idea, their own script, or a document: the avatar speaks in the chosen voice, shots filmed for the exact words cut in between (the avatar stays in a corner window), number and step cards, original music, captions. Not for one talking clip (cadabra_avatar_speak) or a film with no avatar (cadabra_make_film). Three steps. 1) Call with avatar_row_id and idea, script or document_text: returns a FREE storyboard, every beat's words beside what is seen and the exact prompt each shot gets, the itemised price, and a plan_id. Show it. 2) To change it, call with plan_id and a note (free, as often as they like); to re-price with other looks, gestures, brand, engine, layout or music, call with plan_id and those fields. 3) Call with plan_id and confirm:true to make it. It takes minutes and lands in the gallery. Two things it makes: an edited film (default), or talking_only (just the avatar saying the words to camera, no shots, no cards, no music unless asked). Choices, all the creator's: looks (two or more look ids from cadabra_avatar_looks: the same face changes clothes and room each time it returns to camera, one take per look, the voice stays one reading), gestures (natural hand movement on a photo avatar, same price, the face unchanged), brand (the saved brand kit: logo in a corner, its colours on the cards), document_text (the film states only its facts). A script is spoken word for word. Hebrew speaks to the audience in the plural. captions:false leaves the captions off. Every price line says what it is made of. The price is a cap: if the reading comes out longer, it stops and asks before going on.
cadabra_avatar_look
ChatGPTGenerate a new LOOK for an avatar the creator already owns: the same face, in a different outfit, a different room, a different framing β and it can speak, like any other look. π¨ THIS IS WHAT MAKES A FILM INSTEAD OF A TALKING HEAD. A trained twin otherwise wears whatever it was filmed in forever, so any script that needs the person in pyjamas and then in a studio looked impossible. Generate one look per situation, then pass each look_id to cadabra_avatar_speak for that segment. The identity is carried by the avatar's existing look used as reference, not by your words β so describe the CLOTHES, the PLACE, the LIGHT and the FRAMING, and never re-describe the face. About 13 credits per look. Set confirm:true to spend credits.
cadabra_avatar_looks
ChatGPTFree. Every look one of this creator's own avatars can be filmed in: the same face in another outfit and another room, each with its picture. Show the pictures before using a look. Pass look ids to cadabra_avatar_film in looks (several looks: the avatar changes clothes and room each time it comes back to camera) or one to cadabra_avatar_speak. A new look from words is cadabra_avatar_look. Get the avatar id from cadabra_my_avatars.
cadabra_avatar_scene
ChatGPTThe person the creator already built, saying what they wrote, inside a scene they describe. Not a talking head on a background: a real shot, with a camera, a place and light, and the mouth matching the words. About 25 credits for five seconds and 45 for ten, against 89 for the shot this replaces β which did not speak at all. Give it the avatar by name (cadabra_my_avatars lists them and prints the photo beside each), the script in the creator's own language, and the scene in English. It uses a voice from their own library, so Hebrew comes out Hebrew: name one, or clone theirs once with cadabra_clone_voice and it is theirs forever after. π¨ The voice is made FIRST and the shot is cut to its real length, so the words always fit the picture. Keep the script to what a person can say in the time: about three words a second. A script that runs longer than fifteen seconds is refused before the expensive half, with the voice already in hand.
cadabra_avatar_speak
ChatGPTGive an avatar the creator ALREADY owns a new script and get a new film. This is the cheap half of the studio and the reason a build was worth paying for: no rebuild, no re-upload, no new photo. Get the id from cadabra_my_avatars. Price is per second of finished video: about 0.47 credits/sec off a photo avatar, about 0.98 off a trained twin, minimum 2. Those are rounded for reading β the real rates carry full precision, so quote from cadabra_quote rather than multiplying these yourself. π¨ The charge settles against the REAL length of the film once it exists, so a script longer than the estimate costs more than the quote said, never less. Say that when you quote. π§ OFFER TO LISTEN FIRST: listen_first:true reads the script (or its first sentence) in the character's voice for about 1 credit and makes no film, so the creator hears the voice before paying for the render. Voice: a script is READ FIRST in the avatar's own voice, the one attached with cadabra_attach_voice (usually the creator's clone from cadabra_clone_voice), then the avatar lip-syncs to it. Pass voice_name to use a different voice this once, any saved or library voice. The read is quoted with the render, about 1 credit per 750 characters. π¨ An avatar with no voice attached and no voice_name is refused before any charge: ask the creator which voice, never pick one for them. π¨ The provider's own voice record carries the avatar's NAME and a stranger's voice; it is never used. Or send an audio_url you already made. Set confirm:true to spend credits.
cadabra_brand_kit
ChatGPTFree. Read or save the account brand: name, look note, colours, logo URL, and the LOOK IMAGE that locks the style. The film brain uses it when the job is actually that brand. π¨ Read this before generating anything for a brand you have worked on before, and when look_image_url is set, pass it as a reference image on every still and every shot: that is what makes the tenth image match the first.
cadabra_brand_words
ChatGPTFREE. The two pieces of writing a business needs before any post exists. Β· brand β what this brand SOUNDS like: its voice, what it promises, who it is talking to. Run it once and reuse it. Β· ads β the advertisement itself: the line that stops the scroll, the body, the call to action, written to an objective. π¨ Not the same thing as cadabra_social_copy, which writes the caption for ONE finished post, and not the same as cadabra_social_month, which fills a calendar. Reach for this when somebody has a business but no words yet.
cadabra_bridge_shot
ChatGPTWhen two shots do not connect (a cut that jumps in space, time or light), make the short shot that carries the eye across: the last frame of the first clip is read from the clip itself, and a 2 to 4 second shot is generated from it whose only content is the movement into the second clip, same light, same lens, no new character and no new place. Then cut it between them with cadabra_assemble or cadabra_edit_film. Price is one short video shot on the engine chosen (default seedance, Seedance 2.0, the one engine that takes 2 to 3 seconds; seedance-25 snaps to 5). Without confirm:true you get the quote only.
cadabra_camera_fx
ChatGPTFree. Lists the shot types (33: where the camera stands, its framing, angle and lens, taken as shot by cadabra_generate_image, cadabra_generate_video and cadabra_generate_kling), the camera moves (70), visual effects (114) and look options (41: camera body, lens, era, grain, still looks) that cadabra_generate_video and cadabra_generate_kling take as camera_moves, vfx and look. They ride the prompt, so the price never changes. Offer them by their Hebrew or English label and pass the id.
cadabra_campaign
ChatGPTTurn one finished film into a launch package with nothing new generated: a 6-second hook, a 15-second Reels version, a 30-second story version, and two alternative openings (the strongest frame first; the closest face first). Each is a re-edit of the film through the same editor cadabra_edit_film uses, the original untouched, every cut landing in the gallery. Price is five edit fees, each 3 credits base and 1 more per extra 15 seconds of the film's runtime; without confirm:true you get the quote only. Every note forbids black padding, so a short film gives shorter versions, never black frames.
cadabra_campaign_analyst
ChatGPTFree. Hand it the campaign's results, a Meta Ads export as CSV (English or Hebrew headers), pasted numbers, or rows, and it COMPUTES per ad: click rate, cost per click, cost per result, conversion rate, return, the share who watch past 3 seconds and the share who finish, then says SCALE, STOP, KEEP or WAIT FOR DATA against the account's own average, and where each ad loses people with the Cadabra tool that fixes that part (a new opening, a shorter middle, a new end card). Never do this arithmetic yourself; call this. It never moves budget or touches their ad account. The thresholds are rules of thumb and the answer says so.
cadabra_caption_emphasis
ChatGPTFree. For each caption line, the one word a viewer's eye should land on (to set bigger or in colour) and at most one emoji that truly fits, or none. The same pass as the website's captions studio. Use the words as captionEmphasis when asking cadabra_edit_film or cadabra_assemble for captions, or to style an SRT by hand. Nothing is saved or charged.
cadabra_catalog
ChatGPTUse this when they ask what Cadabra can do, which engines exist, or what things cost in general. Not for the price of one specific job (cadabra_quote). List everything available: image models, video engines, what each is best for, their features, credit prices, AND the full callable_tools array of every cadabra_* action this server will actually run. If a name from that array is missing from your tool list, run it with cadabra_invoke. Call this whenever the user asks what they can create, which models exist, what a model does, or how much things cost.
cadabra_change_voice
ChatGPTTake an existing recording and re-voice it with a different voice, keeping the original delivery: the timing, the pauses, the emphasis, the emotion. This is how you keep a performance you like but change who it sounds like. 2 credits. Different from cadabra_generate_voice, which reads text with no performance to copy. Pass voice_name, never ask the user for an id. Set confirm:true to spend credits.
cadabra_character_sheet
ChatGPTOne sentence in any language, one character reference sheet out: four full-body turnaround views and eight expressions of the SAME character on one page, the thing you lock a face with before any film is made. The character is first DESIGNED in English by the same brain the films use (free), so the engine is told what the creature actually is instead of guessing from three words, and that design is what the sheet is built from. Attach reference_image_url and the photo becomes the identity. π¨ No engine draws a Hebrew letter, so a prompt that comes back Hebrew is refused before anything is charged. Without confirm:true you get the English prompt and the price only.
cadabra_check_balance
ChatGPTUse this when they ask how many credits are left, and after any failure, before saying anything about money. Return the signed-in Cadabra user's remaining credits.
cadabra_cinematic_templates
ChatGPTCadabra's ready-made cinematic scenes, the same library the website shows. Call with no arguments for the index, or with template_id to get that template's FULL prompt, duration and aspect, which you then pass to cadabra_generate_video or cadabra_generate_kling. Free, nothing is charged. The prompts are long by design, which is why only one is returned at a time.
cadabra_clean_audio
ChatGPTTakes a recording made in the real world and returns the speech alone: no room, no traffic, no air conditioner, no music underneath. 2 credits per started minute, billed off the file's real duration read server-side, 20 minutes a file. Different job from cadabra_separate_audio, which hands back BOTH halves so the place the shot was recorded in can be kept. Reach for this one when the background is not worth keeping and the voice has to be usable on its own: a phone recording of a script, a voice memo about to be cloned, narration recorded in a room that hums. π¨ Clean it BEFORE cloning a voice from it. A clone learns whatever it is given, so a voice trained on a mix learns the music with the person, and the voice drifts as the background does. That was a real film: the voice at the start was not the voice at the end.
cadabra_clip_ideas
ChatGPTFREE. Call this FIRST whenever someone wants a music clip, a song video, a lyric video or any video for a song, before any shot is planned or priced. It returns the creative playbook viral clips are built on (a device library: literal lyrics, escalation to chaos, dead serious about nothing, the wrong singer, genre swap, scale swap, one-take journey, the world grows, the reveal and more; comedy craft; what makes strangers share and comment; local Israeli gold; the boring clichΓ©s never to pitch) and asks you to pitch THREE concepts in three lanes (hysterical, jaw-drop, heart), each with a logline, the first frame, the escalation ladder, the screenshot frame and the final button. A clip is shared for its idea: never skip straight to a singer in a field.
cadabra_clone_voice
ChatGPTTurn a recording of a real voice into a voice you can then speak any text with. 5 credits. π¨ Only clone a voice the user owns or has permission to use. Thirty seconds of clean speech beats five minutes of noisy. Set confirm:true to spend credits. π¨ HOW THE RECORDING GETS HERE, because this is where people get stuck: you need a URL. If you can read the file's bytes, cadabra_upload_file makes one. YOU USUALLY CANNOT (audio attached in a chat client is not readable bytes), so open cadabra_upload_file with no arguments (an upload button in the chat); only if no button appears, https://livinginmovie.com/mcp/upload (same account, drop the recording, paste the link back), straight away. Never reply 'upload the recording' without it.
cadabra_compare_engines
ChatGPTRun ONE prompt through two or three engines at once and put the results side by side, so the user picks the look instead of taking the first guess. Each engine is charged separately and each is independently refund-safe. Use this when the look matters and nobody is sure, not as a default: it costs the sum of the engines. Set confirm:true to spend credits.
cadabra_compose
ChatGPTFree. YOU write the motion graphic, we render it, and you get back TWO things: the film, and the editable source it was rendered from. Use it for the graphics a video engine cannot make and an EditPlan has no field for: a lower third with someone's stats, a stat card, an animated title or kinetic type, a chart that draws itself, a price card, a scoreboard, a progress bar, a map route, an end card, a logo sting. π¨ CHANGING ONE IS A SENTENCE, NOT A NEW RENDER. When the user says "make the name bigger", "put my colours on it", "the height is 1.90 not 1.82", "lose the third stat": send this tool the source_url from last time ON ITS OWN. The composition comes back as text. Change the one thing they asked for, leave everything else exactly as it is, and send the whole document back as html. Both steps are free. NEVER answer a change by writing a new composition from scratch, and never make the user pay to generate anything again because a word was spelled wrong. That round trip is the entire reason this tool hands back a source file. π¨ WRITE IT SO IT CAN BE CHANGED LATER. This is the whole reason the tool exists, and a composition that ignores it is worth no more than a flat render: - NAME EVERY LAYER. Put data-layer="Player stats" on each meaningful element, in words a person would use, never div1. The names are what someone reads when they come back to change it. - KEEP TEXT AS TEXT. Every number, name and title lives in the HTML as characters. Never bake a word into an image, and never render type you were given as a picture. - KEEP ARTWORK REPLACEABLE. Each image is one <img src> or one CSS url, in one place, so swapping it is swapping a link. - GROUP WHAT BELONGS TOGETHER in one container with its own data-layer, so the whole block moves as one. STRUCTURE THE RENDERER REQUIRES. The root element must carry data-composition-id, data-duration (seconds), data-width and data-height. Without them there is no timeline and nothing renders. Animate with CSS animations or GSAP. Everything must be inline in the one HTML you send: styles, keyframes, fonts as a link, and images as https URLs or data URIs. Nothing is fetched from a private address. HEBREW AND RTL: set dir="rtl" and the bidi, the punctuation and the line breaking come out right, because this is a browser and not a template engine. π¨ But NAME A FONT THAT EXISTS. The renderer carries the Noto family, so font-family: "Noto Sans Hebrew", sans-serif is always safe, and any Google Font linked in the document works too. Asking for a font that is not there does not fall back to a readable one: every Hebrew letter renders as an empty box while the numbers and the layout come out perfect, and the file returns as a success. WHAT IT IS NOT: not a film. It does not cut clips, hold a face, speak or carry music. A film is cadabra_make_film, and changing a finished one is cadabra_edit_film. This makes the graphic that goes ON a film. IF IT FAILS you get the renderer's own error back, usually naming exactly what broke. Fix that line and send it again; a failed compose costs nothing.
cadabra_concepts
ChatGPTTurn a rough thought into five DIFFERENT film ideas, each with the reason it would work and the reason it might not. Free, nothing is generated. Use this before cadabra_make_film whenever the user is not sure what they want: choosing between five costs nothing, and regretting a film costs credits.
cadabra_course
ChatGPTUse this when a creator has ONE long recording (a Zoom class, a webinar, a lecture, up to six hours) and wants it as a course they sell on their own site. Not for short clips of it (cadabra_find_moments and cadabra_cut_reel make reels from the same recording). The deliverable is FILES for their own platform (Teachable, Kajabi, their site): one video per lesson, an SRT beside each, a description, and one zip. Nothing is published here. Walk it in order: 1. action start + src: quotes the course price (9 credits per started hour, charged once, covering the transcript, the chapters, the cut and the packing). With confirm:true after the creator's yes it starts and returns course_id. Say that it takes a while and the chat can be closed. lessons asks for an exact number of lessons; language he or en. 2. action status + course_id (free): while listening, how far along; then the CHAPTERS found (each with its times, its kind, and what housekeeping at its head will be cut: greetings, 'can you see my screen'). π¨ Show them to the creator before cutting: this is the screen they edit. 3. action cut + course_id (free, already paid): cuts the lessons. To change them first, pass chapters [{title, start, end}] in seconds: rename, merge two (one entry spanning both), drop one (leave it out), split one (two entries). Cutting again later is free too. 4. Optional, each quoted before anything runs, with confirm:true after a yes: action opener (a five-second opener with a wordless sting and the course title, spliced into every lesson right after the teacher's first sentence; with portrait_url the creator's own face is held perfectly still) and action covers (one scene for the course, then every lesson's number and title set on it by GPT Image 2.5, which spells Hebrew; portrait_url puts their face in it, title_color and accent_color are their brand). π¨ Lesson titles and the course title are the creator's language; prompts sent to engines are English and are shown before any spend. Every paid part refunds itself if it produces nothing.
cadabra_craft_edit_prompt
ChatGPTFree, nothing generated or charged. Builds the prompt for a video-to-video EDIT: a clip the creator already has, changed in one place. This is NOT cadabra_craft_prompt β that one writes a new shot, and describing an edit as a scene is precisely how an edit comes back as a different film. Use it before running cadabra_generate_video with engine "seedance-25" and video_mode "edit" (or engine "flux3" with video_edit). The returned prompt declares the source clip the master, locks the camera and the performance frame by frame, then lists ONLY the changes, and names every attached picture as @image1, @image2 so the engine actually uses it. Give each change as its own item, and say what FILLS the space a removed thing leaves β a removal with nothing behind it is the most common way an edit comes back with a smear where the object was. The tool warns you when an item looks like it forgot that.
cadabra_craft_prompt
ChatGPTFree, nothing generated or charged. Send a rough idea or a draft prompt through Genie's own creative-director brain, the same award-craft, anti-cliche, colour and identity doctrine cadabra_make_film and cadabra_story already use, and get back one ready-to-use, upgraded prompt for a SINGLE standalone shot. Use this before cadabra_generate_video/cadabra_generate_image whenever you are picking the engine yourself and want the same craft quality make_film gets automatically. Name the engine or model if you already know it (locks the prompt to that engine's real vocabulary and limits); leave it blank to let the brain suggest one. If the idea genuinely cannot land in one shot (a reveal, a before/after), the brain may hand back more than one; both are shown honestly, never forced down to a single step. Pass the returned prompt(s) into cadabra_generate_video or cadabra_generate_image yourself; this tool never generates anything on its own.
cadabra_create_avatar
ChatGPTBuild a REUSABLE avatar (Cadabra Avatar Studio). Paid once, then it lives on this creator's shelf forever and every later film costs only the render. kind:"photo" β one still of a clearly visible face. 13 credits to build, ready immediately, renders at about 0.47 credits/sec. Ask cadabra_quote rather than quoting prices from memory; the studio owns them and they move. π¨ kind:"video" (a trained digital twin) IS CLOSED AND WILL BE REFUSED. Not a price problem: a twin occupies a paid SEAT on the provider's plan that costs us every month, so it cannot be self-serve. Do not offer it, do not describe it as available, and do not ask the creator to film themselves for it. If someone asks for a twin, tell them the trained twin is not open at the moment and give them the two things that ARE: Β· a photo avatar β one picture, built in seconds, speaks any script forever; Β· a real talking clip of themselves β cadabra_clone_voice on ~75s of their speech, then cadabra_generate_voice and cadabra_lipsync over their own footage. No avatar, no consent wait, and it is genuinely them. Anyone who already owns a twin keeps it: cadabra_avatar_speak still works on it and is not affected. π¨ A TWIN NEVER SPEAKS IN THE CREATOR'S OWN VOICE, whoever tells you otherwise. The provider mints a voice record labelled with the avatar's own name and the voice inside it is a stranger's: a man who filmed himself for five minutes was handed a woman's voice carrying his name. Their real voice only ever comes from cadabra_clone_voice (5 credits, about 75 seconds of clean speech), attached with cadabra_attach_voice. Never promise a creator the twin sounds like them. π¨ The picture has to reach us as a URL. If you cannot read the bytes (an attachment in a chat client usually is not readable bytes), open cadabra_upload_file with no arguments (an upload button in the chat); only if no button appears, https://livinginmovie.com/mcp/upload, in the same breath, rather than the word 'upload' on its own. FOR WHEN THE TWIN DOOR REOPENS, and not a reason to ask for footage now: 15 seconds is the provider's hard floor and a shorter take is refused before charging, two to five minutes at 1080p or better is what they actually want, duration_seconds/width/height bring the warnings back before the money does, and a trained twin cannot speak until the filmed person completes a consent link that dies after 24 hours. Call cadabra_my_avatars FIRST β never build a second copy of a face they already own. Set confirm:true to spend credits.
cadabra_create_character
ChatGPTSave a character under a name, from reference images the user already owns, so later shots can point back at it and keep the same look instead of re-attaching photos every time. This is the same training the Seedance 2.5 studio Train-character tab runs. Registers with Seedance (reuse @name on engine seedance with omni_reference:true, or on engine seedance-25 which loads the sheet into Omni like the cube) and Kling (element_ids on cadabra_generate_kling). It does NOT carry over to other image models or other video engines. Set confirm:true to spend credits (7: 6 for the Seedance save, 1 for the Kling element).
cadabra_creative_set
ChatGPTProduce a SET of images that share one look: a launch set, a brand set, a character set. π¨ The difference from generating N images: the first image becomes the look anchor, and every image after it is generated WITH that anchor as a reference, so the light, palette and treatment carry across the set instead of drifting. Give the shots you want, or let it use the launch default (hero, close detail, in use, context). Set confirm:true to spend credits.
cadabra_cut_reel
ChatGPTUse this when a moment from cadabra_find_moments should become a reel ready to post. Not for footage made in Cadabra (cadabra_assemble) and not for changing a finished film (cadabra_edit_film). Everything is decided by rules, not guessed: the cut starts and ends on whole sentences; hesitations (um, ΧΧΧ) and pauses longer than half a second are cut out, never a word the speaker meant; captions come from the words already timed, two or three per page, centred inside every platform's safe zone, in the recording's language; a wide shot is framed on the face of whoever is TALKING (faces measured on our own box, the frame moves only on a cut so it never drifts), with no borders; the hook title holds for the first two seconds; the sound is mastered at -14 LUFS; 1080x1920. After the render it is checked (size, length, a face inside the frame) and anything that fails is said plainly. π¨ IT OPENS ON A HOOK (Sela 2026-09-26): the first span is the strongest FACT, QUESTION or claim in the recording, even if it comes later (put that span first); hook_title is a question or a hard number the first line answers; a B-roll clip from 0s is the visual hook; hook_sound and hook_zoom are on by default. It ENDS on the speaker's punchline (the last span). Price: the edit render by runtime (the same as every edit), quoted first. Needs the words from cadabra_find_moments on the same recording. Each option can be switched off: remove_silences, remove_fillers, captions. A recording of any length works: over eight minutes, only the reel's spans are cut out of it first (free). With an avatar or B-roll the confirmed reel is made in the background (a few minutes): it returns a job_id, and cadabra_cut_reel with that job_id returns the reel.
cadabra_delete_avatar
ChatGPTPermanently remove one of this creator's own saved avatars. π¨ It cannot be undone and the build credits are not returned, so only ever call this when they asked for it by name. Get the id from cadabra_my_avatars.
cadabra_delete_creation
ChatGPTPermanently remove ONE item from this creator's own gallery. Free. π¨ THERE IS NO UNDO AND NO TRASH. The row is gone. So: take the id from cadabra_list_creations, tell the creator exactly WHICH item you are about to delete in words they can check (what it is, when it was made), and only send confirm:true after they have said yes to that specific thing. Never work from a description. "Delete the flower one" is not something anybody should be able to say to an irreversible action; find it, name it, confirm it, then delete it. It can only ever touch this creator's own gallery; the scoping is the studio's, not this tool's.
cadabra_depth_map
ChatGPTTurn a photograph into a per-pixel depth map. 1 credit. What it is for: convincing fake depth of field, a parallax move on a flat photograph, and as a mask that knows what is in front of what. Returns an image where near is bright and far is dark.
cadabra_design_voice
ChatGPTDescribe a voice in words and get three of them to listen to, then save the one you want so it can be used forever after with cadabra_generate_voice. 3 credits for the three previews, saving is free. The three clips also land in the audio gallery (cadabra_list_creations type:audio), so a later search can find them. The description must be between 20 and 1000 characters and must describe an ADULT voice by its sound: 'a warm Israeli woman in her thirties, low and unhurried, slight smile in the voice' works, 'nice voice' does not. Do not mention children, kindergarten, gan, or minors. ElevenLabs blocks those and refunds. THE LANGUAGE THE THREE VOICES SPEAK COMES FROM preview_text, NOT FROM THE DESCRIPTION. The description is written in English, as every prompt here is, and writing "Israeli Hebrew" inside it does not make the previews speak Hebrew: it makes them read the English sample in an invented accent that a Hebrew speaker hears as a foreign language. For a Hebrew voice, pass preview_text in Hebrew, ideally the user's own script, and pass language:"he". Same for any other language. Set confirm:true to spend credits. To SAVE one afterwards, call again with generated_voice_id and voice_name.
cadabra_develop
ChatGPTFree. The rooms that come BEFORE anything is generated, where the thinking is cheap and the mistakes cost nothing. mode "document": a REAL document becomes a film you can shoot. Paste a treatment, a client deck, a chapter, a research paper, a proposal (up to 120,000 characters, so a whole chapter fits) and it comes back mapped: what the document actually says, the evidence and VERBATIM quotes it contains, what it deliberately does NOT cover, plus a world and a cast built to carry it. This is the one to use when the creator arrives with material rather than an idea. Pair it with cadabra_read_pdf when the material is a PDF. mode "bible": a treatment or story becomes a compact project bible: logline, one style lock in English, up to four characters and four locations described by LOOK. This is the cheapest way to hold a face and a place steady across a whole film, and it belongs before shot one, not after the second shot drifts. mode "script": one line becomes a shootable 30-60 second script, 3 to 5 shots, one named character reused in every shot. mode "research": a sentence, ideally with a URL or two in it, comes back with what is actually out there. It READS the links you paste. Use it before writing an ad in a category the creator knows better than you do. mode "expert": one of five departments answers on the brief in their own discipline: strategist, screenwriter, cinematographer, director, animator. Ask when a brief is thin and you need the question behind it, not more adjectives. mode "concepts": THREE NAMED DIRECTIONS from the creator's own words, before any detail is asked. Five ideas are written and a cold critic keeps the best three, ranked, each with a title, a one line logline, the opening (the first three seconds), the turn, and a score; the first is the recommended one with its reason. Show all three, recommend the first, and let the creator pick or keep their own. π¨ What they pick is CARRIED FORWARD, never decoration: it answers the turn (unless they already gave one), its opening becomes the first beat, and on a trailer its title becomes the title card while its turn is WITHHELD (a trailer promises, it never reveals). Then draft the beats of the form, let them rewrite any line, and only then price. All of it is text. Nothing is rendered and nothing is charged.
cadabra_edit_effects
ChatGPTFree. The named effects the editor lays on a cut: viral moments that land on a word (a 3D title, a price drop, a stamp, time freeze, a shatter), text hooks set inside the shot (a hook title pulled into place, handwriting, words pulled out of a laptop, a line revealed between open hands, a swipe), moments made on the creator's own footage (breaking out of the frame, another world behind them, a clone, words behind the head), hand-drawn doodles, caption looks and film looks. Each comes with the sentence to put in cadabra_edit_film's note (or cadabra_assemble's brief), in English and Hebrew; adding one to a cut costs the edit's usual price and nothing more.
cadabra_edit_film
ChatGPTChange a film that is ALREADY finished, by describing the change in words. Hand over the film's URL and a note: warmer, hold the beat before the punchline, steady the shaky shot, smooth the slow motion, fix the face that went red, put captions on, take the music down under the voice, deliver it louder. The editor re-cuts that film and returns a new one; the original is untouched. Use this instead of cadabra_assemble whenever the source is one finished film rather than a pile of clips, and instead of generating anything new: a re-edit costs a few credits where a re-shoot costs many. Price is the edit fee, 3 credits base and 1 more per extra 15 seconds of runtime, measured from the film itself. Set confirm:true to spend; without it you get a quote only.
cadabra_edit_options
ChatGPTCall this FIRST whenever someone wants to change a video they already have β "edit my film", "fix this clip", "ΧΧ’Χ¨ΧΧ Χ‘Χ¨Χ" β before cadabra_pick_engine or any generate tool. "Edit" covers two different products at very different prices: changing what is INSIDE the frame (generative, billed per second of the clip) and changing the CUT (Remotion, from 3 credits for the whole film). This returns the routes for the kind they actually mean, priced on THEIR clip instead of as a per-second rate. It covers every edit route the product has, in three families: change what is inside the frame (FLUX-3, named effects, Seedance 2.5, Gemini Omni) Β· change the film (re-edit from a note, which also carries steady / smooth slow motion / skin tone / loudness / reverse / remove shot N / a score in cues / a grade over one stretch / emphasis on a word, and assemble) Β· add to it or convert it (graphics and titles, reframe to another aspect, upscale, remove the background, split the audio, lip sync, captions and translation). Pass video_url whenever you have one: the studio measures the file itself and every number comes back real. Free, charges nothing, runs nothing.
cadabra_editor_manual
ChatGPTFREE. The complete manual of Cadabra's render engine, the same one our edit brain works from. Start with section "overview" (a map of every capability: picture, transitions, titles, captions, MOTION DESIGN with graphics, charts, chat, mockups, logo reveals, 411 animated stickers, particles, light leaks, film looks, camera moves, sound, end cards). Then read "shape" for every EditPlan field and its allowed values before writing a plan for cadabra_render_plan. Other sections: doctrine (how to cut), story, rules, recipes (named shot recipes), craft (feature-film editorial craft), motion (motion-graphics principles), limits (what the engine cannot do; a key outside the shape fails the whole render).
cadabra_eleven_v4_scene
ChatGPTSeveral characters talking in ONE audio file on Eleven v4: each line in its own voice, the feeling carried from one speaker to the next, directed by tags inside each line. Hebrew supported, up to 40 lines and 10,000 characters. cadabra_generate_dialogue is the eleven_v3 version and stays as it is. π¨ To redo ONE line, call cadabra_eleven_v4_speak with that line and that speaker's voice: v4 keeps it the same person. Pass voice_name per line, never ask for ids. Priced at 1 credit per 750 characters across all lines. Set confirm:true to spend credits.
cadabra_eleven_v4_speak
ChatGPTThe Eleven v4 speech model as its own tool. 90+ languages, Hebrew included, up to 10,000 characters in one read. It performs the text the way an actor would, directed by tags written INSIDE the text, and it keeps the speaker the same person across every take, so asking for another take of a line never drifts to a different voice. Use it when the creator asks for v4, for the most acted, emotional read, or for a long narration that has to stay steady. cadabra_generate_voice is the eleven_v3 tool and stays the default; this one never replaces it. π¨ A new take of ONE line of a longer script: call this tool with just that line and the same voice. π¨ NEVER ask the user for a voice id. Pass voice_name; a wrong or ambiguous name returns the candidates, free. Any voice speaks any of the 90+ languages: the creator picks the voice, and it is never refused for the language of the text. Priced at 1 credit per 750 characters, the same as v3. Set confirm:true to spend credits.
cadabra_entities
ChatGPTFree. The account's untrained reference library: saved character photos (or text-only appearance notes with no photo yet), saved locations/sets, and saved style/tone looks. This is NOT the paid trained-identity system (cadabra_create_character/cadabra_list_characters, which trains a real @name Seedance Omni + Kling character); this is the free, casual 'keep this photo/note around and reuse it' shelf the website's Director's Room cast/world steps are built on. Use it to save a reference once and pull it back into image_url/reference_image_urls/cast on later generations, instead of re-uploading or re-describing the same person, place or look every time. action list|save|update|delete.
cadabra_export_xml
ChatGPTTurn a list of clips into a Premiere / Final Cut XML sequence and hand back a link to the file. FREE. π¨ This is how somebody finishes in their OWN editor. Until now a creator who wanted to take a Cadabra cut into Premiere had to rebuild the timeline by hand, clip by clip. Offer it whenever they mention their editor, a client handover, or wanting to keep working on it elsewhere. The clips are listed in order with their real lengths; the XML points at the same URLs, so their editor pulls the media straight from the gallery.
cadabra_extend_image
ChatGPTTakes a picture and gives it MORE PICTURE: the same photo at a different shape, with the new space painted in to match rather than cropped, letterboxed or stretched. 2 credits. This is the answer to the most common thing a creator hits: a portrait photo that has to run as a 16:9 cover, a square product shot that has to fill a 9:16 reel. Cropping throws away the subject, bars look like a mistake, stretching is worse. This keeps every original pixel and invents only the room around them. The subject is never scaled down to make room, and the original is never touched. Say where the picture should sit when it matters: a portrait becoming a landscape usually wants anchor "top", so the head stays high instead of floating in the middle with air above it.
cadabra_film_status
ChatGPTUse this when a job id is waiting: a film, a render or any long generation. Not for listing past work (cadabra_list_creations). Check a job started with cadabra_make_film or cadabra_create_character. Pass the job id it returned, or nothing to list recent jobs. Free. Pass wait_seconds (up to 45) to wait for it right here, and call again until it is done, so the creator gets the file in this conversation. The work also keeps running if the chat closes. A stopped film can be continued with cadabra_make_film + resume_job_id.
cadabra_finalize_draft
ChatGPTRender a Seedance 2.5 DRAFT (cadabra_generate_video engine seedance-25 with draft:true) again as the VERY SAME shot at 1080p: its composition, subject, motion and sound. This is not an upscale and not a new take, which is the whole point: test the shot cheaply at 480p, pay for 1080p only for the take they liked. Valid for 24 hours after the draft. Priced as the same length at 1080p; quote first, then confirm:true. If it fails, the credits come back.
cadabra_find_moments
ChatGPTUse this when someone has a LONG recording (a podcast, a talk, an interview, a lesson, a vlog) and wants the short clips worth posting from it. Not for a clip that is already short (cadabra_edit_film), and not for making new footage. It times every word (the same transcription cadabra_transcribe sells, at the same price, 1 credit per started 10 minutes), then finds the standalone moments for free: each one starts on a sentence and ends on one, keeps to the length asked, never overlaps another, and carries an honest score out of 100 in four named parts (hook, standalone, payoff, energy), a hook title in the recording's language, and why a stranger keeps watching. A flat recording is told it is flat. Without confirm it returns the price, free. Then show the moments with their scores and let the creator pick. Each moment comes with the exact cadabra_cut_reel call that turns it into a publishable vertical reel (the words are kept, so the reel never pays to hear them again); that reel is quoted by its own tool. For several reels at once, put those calls in a cadabra_mission.
cadabra_fix_words
ChatGPTA flubbed word, a wrong number, a name said wrong: replace just those words inside a video or an audio recording without re-recording. The words are timed and the phrase is FOUND in them; the corrected words are spoken in the speaker's own voice (a saved voice by name, or cloned from this very recording) and laid into exactly that stretch, matched to the room's level and faded so there is no click. On a video the new words are fitted to the old stretch's length so everything after stays in sync, and with lipsync:true only a short window around it is lip-synced so the mouth says the new words. π¨ ONLY on the creator's explicit ask, and only the words they name: an attached recording is theirs and is never rewritten on our own initiative. Without confirm you get an itemized quote (timing the words, a voice clone when no voice is given, the new speech, the lip sync); nothing is charged if the phrase is not found.
cadabra_gemini_design_voice
ChatGPTInvent a voice for cadabra_gemini_speak from a written description and get a preview to listen to. It costs credits (Sela 2026-09-28: design is never free): call without confirm for the quote, show the price, and confirm:true only after the creator's yes. The preview lands in their gallery; a failure is refunded. Describe it in English: age, the language and accent it speaks natively ("native Israeli Hebrew, a Tel Aviv accent"), pitch, pace, the feeling. π¨ For a Hebrew script this is the proven route: a voice designed as a native Hebrew speaker. Play the preview to the creator before using the voice. Hand back the voice_... id and keep it in the conversation: it is not in cadabra_gemini_voices and nothing else remembers it for them.
cadabra_gemini_speak
ChatGPTGoogle's official Gemini 3.8 Flash TTS. A narration in one voice, or a SCENE of two speakers performed in one take with natural turn-taking, each line directed in plain words ("whispering, amused", "fast, out of breath"). Hebrew supported. 1 credit per 750 characters; the quote is free: call without confirm, show the price, and confirm:true only after the creator's yes. Lands in their gallery; a failure is refunded. Which voice tool: cadabra_generate_voice speaks in the creator's OWN cloned voice and the ElevenLabs library. This one is for Google's voices (cadabra_gemini_voices, free), a voice designed from a description (cadabra_gemini_design_voice, priced, quote first), and a two-person dialogue directed line by line. π¨ It turns TEXT into speech and cannot re-voice a recording: to keep a performance and change whose voice it is, use cadabra_change_voice. Direction goes in each line's style, never inside the spoken text. Up to 6,000 characters and two speakers per call; a longer piece is several calls.
cadabra_gemini_voices
ChatGPTFree. Google's official prebuilt voices for cadabra_gemini_speak, each with a gender, a pitch, a persona and a one-line description, so pick by what the script needs ("a calm storyteller", "an energetic host"). The ids it returns ("kore", "achird") go straight into voice or voices. A voice the creator designed with cadabra_gemini_design_voice is not listed here: keep its voice_... id from that call.
cadabra_generate_avatar
ChatGPTTurn a photo + an audio clip into a talking-avatar video (Cadabra Avatar Studio, powered by OmniHuman 1.5). The photo needs a clearly visible person/face β subject_check rejects images with none detected. Audio caps at 35s. Set confirm:true to spend credits.
cadabra_generate_dialogue
ChatGPTSeveral characters talking, in ONE audio file, each line in its own voice, with the turn-taking sounding natural. Hebrew supported. Up to 40 lines. Priced like speech, about 1 credit per 750 characters across all lines together. Use this instead of generating each line separately: separate lines do not sound like a conversation. Pass voice_name per line, never ask the user for ids. Set confirm:true to spend credits.
cadabra_generate_image
ChatGPTUse this when they want a still: a picture, a poster, a product shot, a character master or sheet. Not for moving pictures. Generate an image with Cadabra. Default model nano-banana (fast, cheap). look animation selects gpt-image-2 unless they named another model, and stamps production-CG laws (empty hands, no ink on skin). gpt-image-2 and both gpt-image-2.5 tiers charge +1 credit per reference image attached (on top of the base resolution/quality price) β tell the user this before attaching many references, e.g. a character sheet built from several photos. π¨ The 2.5 tiers sit just above gpt-image-2 for the same cell (2K high: flare 5, sunburst 8, against 4), so reach for sunburst when a real face, product or logo must survive an edit, when a mask_url marks the one area to change, or for dense small text, and say the credit number before you run it. Set confirm:true to actually spend credits β without it you get a price quote only. Failed generations are auto-refunded.
cadabra_generate_kling
ChatGPTGenerate a video with Kling O3 (Kling Omni V3) β text-to-video, image-to-video, or reference-to-video (transform an existing video). Optionally reuse a trained character via element_ids (see cadabra_create_character). Set confirm:true to spend credits. Failed jobs are auto-refunded. π¬ THE CODES RUN HERE TOO. A slash code (/fpvdrone /glambot /liquid /bullettime β¦) with mode image-to-video and their picture as image_start writes the whole English shot, compiled for Kling's own strength β physical simulation, weight, contact, cloth and liquid β rather than for a Seedance prompt. cadabra_video_codes lists all of them.
cadabra_generate_kling_4k
ChatGPTGenerate a premium 4K video with Kling 3.0. Text-to-video or image-to-video, or Omni mode with up to 4 reference images (reference them in the prompt as <<<image_1>>>, <<<image_2>>>, etc.). Set confirm:true to spend credits.
cadabra_generate_music
ChatGPTUse this only when they ask for the older V5.5 music model by name. A new song or score goes to cadabra_generate_music_v6, the default. Generate a full song from a prompt, or from your own lyrics (custom mode). Flat 1 credit. Set confirm:true to spend it.
cadabra_generate_music_v6
ChatGPTWrite a full song on Suno V6, the newest Suno. 1 credit flat, and flat really means flat: a 20-second jingle and a 6-minute track cost the same, so ask for the length the film actually needs instead of trimming to save money. Every run returns TWO different takes of the same brief for that one credit, so play both before deciding. Simple mode (custom_mode false): describe the song in a sentence and Suno writes the words and the music. Custom mode (custom_mode true): style and title are required, and your lyrics are sung as written, which is the only way to control the words. duration (10-360s) works in custom mode only, and V6 honours it. π¨ Suno sings Hebrew, unlike the video engines: Hebrew lyrics in custom mode come back sung in Hebrew. π¬ A SONG WRITTEN FOR A VIDEO: send video_url in simple mode and Suno watches the clip and writes the words and the music to fit it (their limits: one clip, mp4/mov/webm, up to 100MB and 241 seconds). Say the language and the feel in prompt. Simple mode does not take a length, so lay it under the clip with cadabra_assemble and let the edit end it. Three models: V6 (best all round), V6_MINI (faster, lighter), V6_WILD (stranger, bolder). Afterwards cadabra_music_v6 extends it, covers it, splits the stems, makes a video of it or turns the singer into a reusable persona. Set confirm:true to spend the credit.
cadabra_generate_sfx
ChatGPTGenerate sound effects and ambient audio. Priced ~3 credits/minute of AUDIO PRODUCED (minimum 2), refined to the exact length after generation. A 5-minute worst case is reserved up front and refunded down, so the reservation is not the price β quote the estimate, not the hold. Set confirm:true to spend credits. π¨ NOT THE NARRATION TOOL. This speaks with a generic preset voice. Any spoken line β narration, voiceover, dialogue, Hebrew or English β belongs in cadabra_generate_voice, which uses the account's OWN saved and cloned voices (call cadabra_list_voices first, it is free) and costs 1 credit per 750 characters, cheaper than this. A user asked for a specific saved voice by name, was brought here instead, saw the 15-credit reservation, and shipped a film with no voiceover at all. Come here for a door slam, a dragon wing, a room tone. Not for a voice.
cadabra_generate_video
ChatGPTUse this when they want ONE clip, named an engine, or are working shot by shot. Not for a whole film from an idea (cadabra_make_film), and not for changing a finished film (cadabra_edit_film). Generate a video. confirm:true to charge, else quote only. Failed jobs auto-refund. π¬ A PICTURE + A CODE = A SHOT. If the creator wrote a slash code β /fpvdrone /tvcommercial /slowmotion /360 /dollyzoom /spiralshot /glambot /ugc /thirdpersonview /pov /gimbalshot /flyover /zoomout /orbitshot /macrozoom /macroreveal /earthzoom, or the Hebrew name of one β send it as code with their picture as image_url and NOTHING else. The code writes the whole English prompt, locks the identity in that picture and picks the engine, length and frame. Their own words, if they wrote any, go in prompt and become the subject line. cadabra_video_codes lists all seventeen with what each one wants attached. Legal engine values: seedance-25 (Seedance 2.5), ltx-25 (LTX-2.5), seedance (Seedance 2.0 cube), seedance-4k, flux3, minimax-h3, gemini-omni, grok-imagine, happyhorse, wan-3. If they named 2.5 send seedance-25. If they named 2.0 or Seed 2 send seedance. 15s on seedance-25 bills as 20 then cut; do not switch to seedance to keep 15. Default engine seedance (2.0). Feature animation: get the prompt from cadabra_animation_craft kind seedance-25. One block equals one shot. Whole seconds, last timecode never above 30. Production CG look, no film stock. Dialogue only in braces. π IN CHATGPT, FILES THE USER ATTACHED GO STRAIGHT IN: their photo as image_file, a song or voice as audio_files, extra pictures or clips as reference_files. ChatGPT fills them in; there is no upload step and no upload page. They are copied to Cadabra's CDN on the quote call and their URLs come back in the quote, so the confirm call can send those URLs or the same files. π¨ IF CHATGPT SAYS "INTERNAL ERROR" ON A CALL WITH ATTACHED FILES, THE CALL NEVER REACHED CADABRA and nothing was charged (ChatGPT's own file hand-off failed; measured 2026-09-28 with three references). Do not retry the same call. Put each file through cadabra_upload_file (chat_files, one file per call), then call this tool again with the returned links in image_url / reference_image_urls / audio_urls and no *_file fields. The prompt, engine and price stay the same.
cadabra_generate_voice
ChatGPTUse this when words must be spoken: narration, a character's line, Hebrew speech. Not for singing (cadabra_generate_music_v6). π¨ THE narration tool, and it runs on eleven_v3 with ALL of v3 exposed: audio tags inline in the text, plus stability, style and similarity_boost. A narration that comes back flat is almost always a missing [tag] and a default stability, not a wrong voice. For an advertisement, reach for stability 0.3 and tag the lines. Turn text into spoken audio in the account's own saved and cloned voices, Hebrew included (it is the only Cadabra voice tool that supports Hebrew). Use it for every spoken line: narration, voiceover, dialogue. π¨ NEVER ask the user for a voice id. Pass voice_name. If the name is wrong or ambiguous you get the candidates back, free, with nothing charged. π¨ HEBREW AND THE VOICE. Every library voice can read Hebrew, but one whose own language is not Hebrew pronounces it with that language's accent, and a native listener hears it. When that applies, Cadabra says so before anything is charged and the creator decides. A CLONE of someone who speaks another language is refused for Hebrew unless the creator insists (use_this_voice_anyway). To choose, call cadabra_list_voices and pick by the job (advertisement, conversational), or take the creator's own Hebrew clone for a native read. Priced at ~1 credit per 750 characters. Kept 7 days. Set confirm:true to spend credits.
cadabra_generate_world
ChatGPTWorlds (Marble / World Labs): build a walkable 3D set from a prompt or a plate. Returns a camera still (not the odd 360 thumb). Video is NOT included. After this lands, call cadabra_generate_video with that still as image_url (engine ltx-25 or seedance-25). Two charges: 16 credits for the world, then the video engine. Set confirm:true to spend. Failed jobs are auto-refunded.
cadabra_get_edit_plan
ChatGPTFREE. Returns the full EditPlan JSON behind a past edit (its project_id comes back from cadabra_assemble, cadabra_edit_film and cadabra_render_plan, and cadabra_my_projects lists them). Use it to change ONE thing precisely: move a title, swap a clip, retime a caption, add confetti on the price, then send the changed plan to cadabra_render_plan. That is a change to THAT cut, not a new edit that might redo everything else.
cadabra_hebrew_lyrics
ChatGPTFREE. Before a Hebrew song is generated: the lyrics come back with full niqqud, a version ready to paste into the song engine (English section tags, niqqud on every risky word, names and foreign words split into sung syllables like ΧΧΦΉ-ΧΦ΅Χ¨), and the list of words most likely to be sung wrong. Unvowelled Hebrew is ambiguous to a singing engine, which is why names and look-alike words come back mispronounced. The words are never changed, only helped. Run it on every Hebrew song before cadabra_generate_music_v6.
cadabra_home
ChatGPTUse this when the user opens Cadabra, says hello, asks what Cadabra can do or where to start, or asks for the menu or home screen. It shows the Cadabra home screen right here in the chat: what they can make (a finished film, a video, a face that talks, an image, an ad, a series, a clip for a song), quick tools, real work made on Cadabra, their balance and their latest creations. Free: nothing is created or charged. Each tile writes the user's request into the chat in their words. After it opens, do NOT list its contents again in text: say one short welcoming line in their language and wait for them to tap something or say what they want.
cadabra_identify_speakers
ChatGPTSplit a recording by speaker: who talks, when, and what they said. 1 credit. Use it on an interview or a podcast before cutting, so a cut never lands in the middle of someone's sentence. Set confirm:true to run it.
cadabra_import_link
ChatGPTFree. Give it a link to a product page, a shop or a brand's site and it returns what that page publishes about itself: the product name, the description, the pictures, the logo and the brand colour. π¨ Call this FIRST whenever someone pastes a URL or says 'here is my product / my shop / my site' β it saves them uploading anything. Feed the images straight into cadabra_generate_image, cadabra_make_film or cadabra_brand_kit. It reads only the public page (the same data a link preview shows), never anything behind a login.
cadabra_inspect
ChatGPTFree. Point this at any video, film or audio file and find out what is actually in it BEFORE you plan anything with it. Three readings, and you choose which: - duration: how long it really runs, measured from the file itself rather than from what anyone remembers. - frames: stills sampled across the runtime, returned AS PICTURES you can look at. Use them to see what is in the shot before writing a caption over it, to check a face is in frame, to judge whether two shots cut together, or to find the moment worth holding on. - beats: where the pulse falls in a piece of music, in seconds, with the bpm. π¨ Look before you edit, not after. A cut planned without seeing the frames and hearing the beats is a guess, and a guess costs a re-edit. This read is free; the re-edit is not. π¨ Use it on your own output too. After a shot lands, inspect it. That is how you catch the shot that came back wrong before building four more on top of it, and it is the only way to tell a user honestly what their film looks like instead of repeating their prompt back to them. π¨ The beat list is the PULSE, not a list of cut points. A cut usually lands every 2, 4 or 8 beats; cutting on every one is a strobe, not an edit. Note also that a fast bpm here can be the double-time reading of a slower track, so check the gap between beats against how the music actually feels. Beats run ffmpeg over the whole track and can take a couple of minutes, so ask for them only when a cut really has to land on music.
cadabra_invoke
ChatGPTRun ANY Cadabra studio action by its cadabra_* name, including tools your client did not load into this chat. Cadabra registers the full studio on tools/list: worlds (cadabra_generate_world), Seedance 2.5 (cadabra_generate_video engine seedance-25), storyboard, assemble, voices, film, marketing. Some clients keep only the first ~17. If a catalog name is missing from your callable set, call THIS: action is the exact tool name, params is that tool's arguments (include confirm:true to charge). Quote-before-charge still applies. Never tell the user Cadabra cannot do a catalog capability because the tool is missing from your list. Call cadabra_catalog for the full callable_tools array.
cadabra_lipsync
ChatGPTRe-sync a video's mouth movement to a DIFFERENT audio track. Three models: veed-2 (default, the only one that handles a video that CUTS between shots), sync-3 (highest fidelity, dearest), sync-2 (cheapest). Needs pre-hosted video_url and audio_url. Durations are measured server-side, so you do not need to know them; video_seconds/audio_seconds can only raise a measured length, never lower it. Price follows the longer of the two.
cadabra_list_characters
ChatGPTList the characters saved on this account, with exactly how to reuse each and the consistency sheet each one was trained on. Kling characters carry an element_id for cadabra_generate_kling. Seedance characters are the same library the 2.5 studio Train-character tab writes: reuse with @name on cadabra_generate_video, engine seedance (omni_reference:true) or engine seedance-25 (the tool loads the sheet into Omni, same as clicking the face in the cube). Free. Call this whenever the user mentions a character by name, asks what characters they have, or asks to SEE a character. π¨ Copy the name character-for-character: a character saved as "Yonatan" is NOT reachable as "@ΧΧΧ ΧͺΧ".
cadabra_list_creations
ChatGPTUse this when they ask for something they already made: the gallery, a past result, its link. Not for a job still running (cadabra_film_status). The user's gallery, searchable, with the PROMPT each thing was made from. Free. This is how you continue from past work instead of starting from a blank page: search for what they are referring to ('the coffee one', 'yesterday's reel'), take its URL as the input to the next tool, or take its prompt as the starting point for a variation. Whenever the user says 'that video', 'the last one', or 'like the one I made', come here first rather than asking them to paste a link. π¨ A film from cadabra_make_film is NOT here until it is finished. The gallery holds each shot as it lands; the assembled film is a JOB, and it appears here only once assembly completes. Never conclude from this list that a film failed β this listing names any film job it can see, and cadabra_film_status is the only truth about one. Telling a user their paid film 'was not created' while it is still assembling is the worst thing you can do with this tool.
cadabra_list_voices
ChatGPTThe whole voice shelf: this creator's OWN cloned voices first, then the ElevenLabs library (v3 only). Returns voice_id plus name, gender, accent, age and use case. Free. π¨ ANY LIBRARY VOICE CAN READ HEBREW. The language on a library voice is the language its DEMO was recorded in, not a limit, and ElevenLabs' verified list is the set they certified rather than the set the v3 model can speak. Not one library voice is certified for Hebrew. A voice whose own language is not Hebrew reads it with that accent; cadabra_generate_voice warns before charging when it applies, and the creator's own Hebrew clone is the native-sounding choice. So do NOT search "hebrew" and pick from what comes back: that returns almost nothing but this account's own clones, which is why the same three voices narrate everything and why creators keep being told their voice cannot speak their language. PICK BY THE JOB, NOT BY THE LANGUAGE. Search a use case ("advertisement", "conversational", "social_media", "narrative_story"), a gender, an age or an accent, or pass no search at all and read the whole shelf. For an ad, "advertisement" and "conversational" are the voices that sound like a person; "narrative_story" is the trailer-and-audiobook end of the shelf and will sound heavy on a 20-second commercial. A cloned voice is the one exception where the language IS real, because it carries the mouth of the person who was recorded.
cadabra_magic_angles
ChatGPTTake one photo and re-shoot the SAME subject from different cinematic camera angles, keeping identity, outfit, lighting and scene identical and moving only the camera. 2 credits per angle, up to 4 angles per call. Set confirm:true to spend credits.
cadabra_make_campaign
ChatGPTGenerate several image OR video ad variants in parallel from a list of prompts (or one prompt Γ count). Because failed generations are auto-refunded, you only pay for the variants that actually land β explore freely. This is the refund-safe answer to 'give me 5 ideas'. Optionally pass a preset for ready-made framing/aspect-ratio defaults instead of specifying them yourself.
cadabra_make_film
ChatGPTUse this when someone wants a finished film, ad, reel or story from an idea and has not named an engine: it plans, shoots, edits and captions, and its preview is free with one total price. Not for one clip in an engine they named (cadabra_generate_video), and not for a job of many deliverables (cadabra_mission). Turn one idea into a complete, EDITED film, not a single clip. Plans the shots, generates every one, and assembles them (cuts, transitions, captions, music) into one finished video. Choose the length: short 20-30s (default), standard 35-45s, long 50-90s, or feature 2-3 minutes. The run is detached, so a longer film means a longer wait and a bigger bill, never a lost film. Photos keep their job: product_image_url is the real product, character_image_url / character_image_urls are people who must stay the same (up to 8), look_image_url is grade and palette only. reference_image_url is the old single-still slot if you do not know the role. Set confirm:true to actually run it (spends real credits per shot plus a small edit fee); without it you get a free plan preview only.
cadabra_make_thumbnail
ChatGPTDesign a scroll-stopping cover image for a video or post. Cadabra's thumbnail director writes the concept (free), then each concept is rendered as an image. 1 credit per concept, 2 when a reference photo is attached. Returns the artwork plus the headline text written for it. Note: on the website the headline is burned onto the picture by the browser; here you get the image and the words separately, so you can place them yourself. Set confirm:true to spend credits.
cadabra_marketing_workflow
ChatGPTFree. The brand-visuals pipeline as pickable steps, each mapped to a tool in this chat. Run only the steps the user asked for. Stay in this chat.
cadabra_midjourney_edit
ChatGPTThe six edit modes of the website's Midjourney V8.1 studio. 1 credit each, quote first, auto-refunded if it fails. - variation / remix: a new take on ONE of the four images of a Midjourney grid made earlier. Needs source_task_id (the taskId cadabra_generate_image returned for that midjourney run) and image_number 0-3 (top-left 0, top-right 1, bottom-left 2, bottom-right 3). strength "subtle" stays close, "strong" (default) moves further. remix also takes a new prompt. - edit (canvas): repaint one grid image from a prompt. Needs source_task_id, image_number and prompt. - retexture: keep the shapes of ANY picture and restyle its surfaces from the prompt. Needs image_url and prompt. - remove-bg: cut the subject out of ANY picture. Needs image_url. - upload-paint (mask): repaint part of ANY picture. Needs image_url and prompt; mask_url is an image whose opaque area is repainted (without it the whole frame is repainted). Say the price (1 credit) before you call with confirm:true.
cadabra_mission
ChatGPTUse this when the creator asks for a JOB with several deliverables (a week of posts, a film plus three reels plus stills, a campaign in three sizes), or says 'just do it all'. Not for one clip or one picture (use that tool), and not for one film from an idea (cadabra_make_film). HOW IT WORKS. action "plan": you write the items, each one a call to an existing cadabra_* tool with its full arguments, written at full craft (use cadabra_recipes, cadabra_craft_prompt and the tool's own rules; every prompt is the best you can write, never a placeholder). An item can use an earlier item's result by writing "{{m1}}" where a link goes (a master still feeding a shot). Free: every item is priced by its own tool, every prompt comes back crafted, EXACTLY as it will render, and nothing runs. The mission id returns at once; the pricing runs on our side, so call action "status" until the board is ready (the card shows it too). π¨ THEN SHOW THE BOARD: every item, its price, its prompt, the total. Nothing runs until the creator says yes. action "approve" with confirm:true and item_ids (or all:true) runs only what they approved, on our side, even if they close the chat. Every approved price is a CAP: an item that would cost more when its turn comes stops and asks again. An item whose result feeds a later one is a CHECKPOINT: the mission stops after it so they see it before paying for what uses it; approve again to continue. A failed item stops the mission; nothing is retried without a new yes. action "cancel" stops what has not started.
cadabra_moodboard
ChatGPTFree, no credits. Build a moodboard from reference images and get back a style_code (like mb1a2b3c4d) that carries the palette, lighting, lens and mood, never the subject. Paste that code directly into a prompt on cadabra_generate_image/cadabra_generate_video: engines nano-banana, seedream, gpt-image-2, wan, grok-imagine, minimax-h3 and happyhorse all read it automatically and strip it before generating, no extra step needed. Other engines need the DNA read out in words instead (action:"get", then describe the palette/lighting/lens yourself in the prompt). Flow: action:"create" (name it) -> action:"add_images" (2+ real photos; more angles/moments than one gives a richer read) -> action:"analyze" (the actual vision pass, capped at 10 images) -> action:"get" to read the style_code and full DNA back. action:"list" shows your boards, action:"delete" removes one.
cadabra_morph
ChatGPTCast one Cadabra Morph spell (ids from cadabra_morph_spells) on the creator's own photo: a picture (GPT Image 2.5, 1 credit plus 1 per attached photo) or a 5-second vertical film (Seedance 2.5, priced by resolution). Every brief locks the person's real face at its start and again at its end, with their age, body and hair. Without confirm:true it returns the quote and the exact brief; with it, it spends and returns the result, which also lands in the gallery. Open to accounts with a package and at least 1 credit; anyone else is told so and nothing is charged. Use only photos of the creator themself, their pet, or someone who agreed to it.
cadabra_morph_mix
ChatGPTCadabra Morph Mix: one clip the creator filmed, restyled in 2 to 4 styles (style ids from cadabra_video_effect's style list), then cut between them on every snap or clap with the original voice underneath, as one reel. Each style is priced like a restyle of the whole clip on Seedance 2.5; the joining is free. Without confirm:true it returns the quote only. Takes minutes; the result also lands in the gallery, and each style on its own as well. Open to accounts with a package and at least 1 credit.
cadabra_morph_spells
ChatGPTList Cadabra Morph's ready spells cast from a photo, free: "you, in a picture" (an action figure in its box, a magazine cover, a marble statue, a Renaissance portrait, a crochet doll and more), "you, in a film" (a 5-second vertical clip), and every animation style as a still. Each line gives the spell id, what it makes, what to attach and the price. To restyle a VIDEO the creator filmed, use cadabra_video_effect with effect restyle and a style instead.
cadabra_motion_control
ChatGPTKling 3.0 Motion Control: the character in a photo performs the movement of a reference clip (a dance, a gesture, a walk), as a new video. Priced by the reference clip's full length (3 to 30 seconds), measured from the file. orientation "image" (default) keeps the face true to the photo; "video" follows the clip's framing more closely. Without confirm:true it returns the quote only. Use a photo of the creator themself, a character they made, or someone who agreed to it.
cadabra_music_edit
ChatGPTTakes a song they already have and changes ONE window of it. 4 credits per started minute of the song, billed off its real length. π¨ THIS IS THE THING THAT WAS IMPOSSIBLE UNTIL NOW. Every music engine here writes a whole song from nothing, so "the chorus is perfect, the middle drags" had exactly one answer: generate again, pay again, and lose the part they loved β because a new generation is a new song. Here the audio outside the window is the ORIGINAL, not a re-performance of it, and only the window is played again. It works on ANY song they have: one made here, one from another engine, their own recording off a phone. The seconds just before the window are used as the reference, so the new part comes back in the same key and the same room instead of sounding spliced in. The window must be between 3 and 120 seconds, and the whole song at most 10 minutes. Say what the new section should BE ("drums drop out, just piano and voice"), not what is wrong with it.
cadabra_music_from_video
ChatGPTHand it the finished film and it writes the score FOR that film: the music is composed against the picture, so the swells land on the cuts instead of a track being laid under them and hoped for. 2 credits per started minute of video, billed off the real length. Up to 10 minutes. Use it when a cut is locked and needs a bed that belongs to it. For a song that stands on its own, or one whose length and lyrics you want to control, cadabra_generate_music_v6 is the cheaper and better tool.
cadabra_music_v6
ChatGPTEverything you do TO a song on Suno V6, plus free lyric writing. Pick one action: β’ extend β carry an existing track on for longer (1 credit) β’ cover β upload any recording and have it re-performed in a new style, melody kept (1 credit) β’ add_vocals β put singing over an instrumental you upload (1 credit) β’ add_instrumental β put a band behind a vocal you upload (1 credit) β’ stems β split a track into its parts. separation 'separate_vocal' gives vocals + backing (1 credit), 'split_stem' gives up to 12 instrument stems (3 credits), 'split_stem_advanced' pulls one named instrument (1 credit) β’ music_video β an MP4 of the song you can post (1 credit) β’ wav β the same song as a studio-quality WAV (free) β’ timed_lyrics β the lyrics with a timestamp on every word, which is what karaoke captions are made from (free) β’ persona β save this track's singer and style as a reusable voice, then pass the id back to cadabra_generate_music_v6 as persona_id (free) β’ lyrics β write lyrics only, no music, no track needed (free, 60 drafts a day) π¨ wav, timed_lyrics and persona are free because they only ever run on a track YOU generated here and each reaches Suno once per track; they need source_task_id and audio_id from that song's result. extend and stems can also work on a track from this studio; cover, add_vocals and add_instrumental take an uploaded file instead (cadabra_upload_file with no arguments (an upload button in the chat); only if no button appears, https://livinginmovie.com/mcp/upload). Set confirm:true for the paid actions.
cadabra_music_video_cut
ChatGPTThe final cut of a music clip, built in code, not guessed. Before planning ANY music clip, read cadabra_editor_manual section 'music-clip' (free): how international-level clips are made and how to lead the creator there. Every every singing shot starts 0.3 s before its line (the slice it was made from starts there too) so the mouth lands on the words; the gaps are filled with B-roll cut to whole bars at the song's own tempo; every clip is muted and the song carries the sound, mastered loud for social. Pass range_start/range_end for a vertical 15 to 30 s cut from the best part (usually the chorus). Same edit fee as cadabra_assemble. Without confirm you get the quote, the runtime and the shot order.
cadabra_my_avatars
ChatGPTFree. List the saved avatars on this creator's shelf β the ones they already paid to build, in the Cadabra Avatar Studio or here. An avatar is bought ONCE and then speaks as often as they like, so ALWAYS call this before offering to build a new one: rebuilding an avatar they already own charges them twice for the same face. Returns each one's id (pass it to cadabra_avatar_speak), whether it carries its own cloned voice, whether it is still training or still waiting on consent, and THE PHOTO IT WAS BUILT FROM. π¨ AN AVATAR DOES ONE THING: IT SPEAKS. A script becomes that person saying it, on cadabra_avatar_speak. Putting that person in a SCENE is a different job on different engines, and filming the avatar itself in a scene is CLOSED because it cost 89 credits a shot. So when someone who already owns an avatar asks for a cinematic scene, nothing has to be built and nothing has to be uploaded: take the photo printed beside that avatar and pass it as image_url to cadabra_generate_video (about 18 credits for 5 seconds at 720p). They already gave us that picture once. For a whole series with the same face, cadabra_create_character on the same photo (6 credits) holds it across every shot.
cadabra_my_projects
ChatGPTList the creator's EDITABLE PROJECTS β the timelines left behind by their edits, assemblies and Genie films β newest first, each with a link that opens it on a real timeline in the studio editor. Use it whenever they ask for an earlier cut, want to change one by hand rather than by description, or say something like "open the project from yesterday". Free, reads only.
cadabra_my_storyboards
ChatGPTFree. The storyboards saved on this account, newest first, and every scene inside one of them: order, description, dialogue, shot type, camera angle and movement, transition and duration. The connector could always BUILD a board and never open one that already existed, so "carry on with the board from last week" had no answer but to start again. Call it with no arguments to list, or with a board id to read it scene by scene and pick up where they left off.
cadabra_my_voices
ChatGPTEvery voice THIS user trained or designed, whether it is currently awake or asleep, and the way to wake a sleeping one. Free, both actions. Use it before telling anyone a voice of theirs is missing. cadabra_list_voices lists what is live at the provider right now; this lists what the user OWNS. A voice can be asleep because the platform holds a limited number of live voices at once and the least recently used ones step aside β the voice itself is kept, with its own recordings, and waking it rebuilds it exactly as it was. Nothing is lost and nothing is charged again. action:"list" returns each voice with a warm flag. A warm voice carries a voice_id you can pass straight to cadabra_generate_voice. A cold one carries only a library_id: call this tool again with action:"wake" and that library_id, then use the voice_id it returns. Waking takes a few seconds and is free.
cadabra_pack_files
ChatGPTFree. Turn 2 to 30 public file URLs into ONE batch link. Pass that single link wherever reference images, Seedance 2.5 Omni stills, or a cast list are accepted. The tools expand it. This is how you stop pasting thirty links. Nested batch links are opened first, then re-packed.
cadabra_personas
ChatGPTFree. Turn a film that worked into a named, reusable STYLE: its preferred engine, aspect ratio, prompt vocabulary and credit ceiling, distilled into one addendum block. Save it once, then pass its persona_id on cadabra_make_film to auto-apply that exact style to a new idea, instead of re-explaining your look every time. Typical flow: right after a film you loved, call action:"synthesize" with the same sentence and the plan object cadabra_make_film returned, review the draft it hands back, then action:"save" to keep it. There is also a public community feed (list_community/clone) of other users' shared specialists, but that is a bonus, not the point: the main value is YOUR OWN reusable look.
cadabra_pick_engine
ChatGPTChoose the right model for a specific job, with the reasoning and the price, instead of guessing. Returns what each engine is best at, what it BREAKS on, and the hard rules that decide it (Hebrew text in an image, Hebrew speech, keeping a real person's face, iterating cheaply, 4K for a hero shot). Free. Call this before generating whenever the user has not named an engine, then tell them what you picked and why, and offer the cheaper alternative with what it costs them.
cadabra_plan_cuts
ChatGPTFree. Given already-made clips, decide order and in/out points. A batch link from cadabra_pack_files is fine. Then cadabra_assemble with confirm:true to cut them.
cadabra_portrait_looks
ChatGPTFree. 44 finished looks for a picture of a REAL PERSON, each one a full production prompt with the wardrobe, lighting, lens and grade already written: the professional headshots (LinkedIn, CEO, lawyer, doctor, realtor, speaker, dating), the fashion ones (editorial, black and white, noir, golden hour, 70s, street), and the viral ones people ask for by name (action figure in the blister pack, animation still, brick minifigure, trading card, movie poster, magazine cover, superhero, cyberpunk, samurai, royal). Call with no arguments to see the shelf. Call with a look to get the prompt, then run it yourself on cadabra_generate_image with the person's photo in reference_image_urls and raw_prompt:true β the prompt is already crafted and the free craft pass would only rewrite it. Without a reference photo this makes a stranger in a nice suit, so ask for the photo before running anything. The prompts have male and female variants because wardrobe and styling differ. Use what you already know about the person from the conversation; ask only if you genuinely do not know.
cadabra_prompter
ChatGPTThe Cadabra Prompter itself, the same brain the website's Prompter chat runs on, with all of its modes reachable from here. Β· prompt β the conversational prompt engineer, its own default. 1 credit. Β· screenwriter β a COMPLETE scene back on turn one instead of follow-up questions. 1 credit. Β· lyricist β lyrics written to be sung rather than read. 1 credit. Β· identity_lock β the description that keeps one person the same person across shots. 1 credit. Pair it with cadabra_create_character, which trains the face; this writes the words that hold everything else. Β· merge β two ideas folded into one prompt. 1 credit. Β· assistant β the general Cadabra assistant. 1 credit. Β· avatar_check β whether a photo will survive as an avatar, asked BEFORE paying to train one. 1 credit. Β· learn_style β reads work the creator likes and names their style back to them in their own vocabulary. FREE. Β· edit_brief β turns a rough note into the brief the editor actually reads, for cadabra_edit_film or cadabra_assemble. FREE, for the same reason the edit itself is: the credit is taken by the render, and charging someone to phrase their own request better only teaches them not to. Nothing is rendered and nothing is generated here: every mode returns TEXT, which you then run with the right tool. Free modes run straight away. A mode that costs a credit needs confirm:true, and without it you get the price and the description back for nothing. π¨ The Prompter is for paying customers with a live balance: a paid account AND at least one credit. Signup-gift credits do not open it. If the account does not qualify you get one clear sentence back and nothing is charged; relay it and do not retry.
cadabra_pronunciation
ChatGPTFree. A speech engine reads what is written, so a brand, a family name or a foreign word comes out as a guess, and in Hebrew reliably the wrong guess. Teach it once here and every narration this creator makes from then on says the word right, with the script still written the way a human writes it. π¨ THIS REPLACES THE HABIT OF MISSPELLING THE WORD IN THE SCRIPT. Bending the spelling to force a sound works for exactly one read, has to be redone in every script forever, and poisons the text the moment anyone copies it into a caption, a post or a subtitle, where a human reads the deliberate typo. action "teach": word is how it is WRITTEN, said_as is how it should SOUND, written the way it would be spelled to get that sound ("Cadabra" said as "Χ§ΧΧΧΧ¨Χ"). action "list" shows what they have taught. action "forget" removes one word. Reach for it the moment a creator says a name came out wrong, and before a long narration that is full of names.
cadabra_proofread_script
ChatGPTFree. Run this on any text that is about to be read aloud by cadabra_generate_voice or cadabra_generate_dialogue. A speech engine voices what is written, letter for letter: a typo becomes a mispronounced word, a number becomes the wrong number out loud, and the mistake is only heard after the credits are spent and the MP3 is in the creator's hands. This returns the corrected text, a short list of what was changed, and β when the writing is chatty rather than spoken β an alternative written in narration register. Up to 5000 characters. Hebrew included.
cadabra_publish
ChatGPTPublish a finished film to the creator's OWN connected Instagram or Facebook. Free. Two steps, always: call it once WITHOUT confirm to show exactly what will be posted and where, and only call it again with confirm true after the creator has read that and said yes. Never call it with confirm on your own initiative, and never with copy they have not seen. The account is whoever owns this connection; you cannot post anywhere else. Connecting an account happens once on the website.
cadabra_qa_shot
ChatGPTFree. Hand it a frame from a shot you just rendered (cadabra_inspect gives you frames) and, when there is one, the reference photo the person was supposed to match. It answers with the one thing that is wrong, or with nothing. Pass an end frame too whenever you can. The two failures that actually reach a creator β the character changing halfway through, a camera move that never resolves β happen ACROSS a clip and are invisible in a single frame. π¨ It is advisory and it fails open: silence means no problem was found, never a guarantee. It does not block anything and it never re-renders on its own. Use it to catch the shot before the creator does, then tell them what you found and let them decide.
cadabra_quote
ChatGPTUse this when they ask what something costs before deciding. Not needed before a generate call: every paid tool quotes itself when called without confirm. Estimate the credit cost of an image or video BEFORE generating. Always quote first, show the price, then call the matching generate tool with confirm:true. Nothing is charged by this tool.
cadabra_read_pdf
ChatGPTRead the text out of a PDF the creator points at: a brand book, a treatment, a brief, a script. FREE. π¨ Reach for this the moment somebody says "here is our brand guide" or "the brief is attached" instead of asking them to paste it in. Feed what comes back into cadabra_brand_kit, cadabra_make_film or cadabra_prompter. Takes a URL, not bytes: a real brief is hundreds of kilobytes and does not belong inside a tool call. Up to 20MB.
cadabra_recipes
ChatGPTCadabra's ready-made recipes: proven ways to build a viral reel, a product ad, a talking character, a consistent series, a cinematic scene, a feature-animation film, or a festival short from three cards (character, place, object). Call with no arguments to list them, or with a recipe name plus a topic to get the full direction, which you then follow step by step. Free. Use this whenever the user asks what they can make, or asks for one of these things without giving a detailed brief.
cadabra_reframe_video
ChatGPTTurn one long clip (up to 5 minutes) into a different aspect ratio in one shot. Finds every real scene cut inside it and reframes each segment to the new shape β the subject stays composed, this is not a center-crop β then hands back one clip with the original audio reattached. 0.3-0.6 credits per second by output quality (480p/580p/720p), minimum 2. Set confirm:true to spend credits.
cadabra_remix_voice
ChatGPTTake a voice that already exists and shift it: older, warmer, faster, angrier, a different accent. Returns three variations to listen to; save the one they pick and it becomes a voice of its own, with the original untouched. 3 credits for the three, saving is free. Use this instead of cadabra_design_voice when the creator says "this voice, but ..." β designing from scratch gambles on a stranger, this keeps the identity they already approved. Give the source voice by NAME; never ask the user for an id. The description says what CHANGES, in English, 20 to 1000 characters ("the same voice ten years older, slower, a little gravel in it"). The language the previews SPEAK comes from preview_text, so for Hebrew pass a Hebrew line, ideally the creator's own script. Set confirm:true to spend. To keep one afterwards, call again with generated_voice_id and voice_name.
cadabra_remove_video_background
ChatGPTCut the subject out of a video and drop the background. Optionally composite a replacement background image behind them. Flat 4 credits. Set confirm:true to spend credits. A failed run is auto-refunded.
cadabra_render_plan
ChatGPTRender an EditPlan that YOU wrote, with every capability of the engine and nothing lost in translation: the exact trims, the title at 2.0s, the split screen from 4s, confetti on the price, a counter that lands, a chat thread, a phone mockup, the logo reveal at the end. Read cadabra_editor_manual (overview, then shape) first; a key outside the shape fails the render. To change a past edit, start from cadabra_get_edit_plan. Every clip src must be an https url the creator gave or made (uploads, gallery, results), never an invented one. The fee is the same edit fee as cadabra_assemble: 3 credits base, +1 per extra 15 seconds. Without confirm you get the quote, the measured runtime and a list of what the plan does; show the creator, then confirm:true.
cadabra_review_film
ChatGPTTake a FINISHED film and get director's notes on it: at most three, most damaging first, and only things an edit can actually fix (order, trims, caption size and position, a title, pacing, the ending, colour mismatch between shots, a shot that should go). Free. A clean film gets ZERO notes, and zero is a pass, not a failure. Run this before the user posts anything.
cadabra_separate_audio
ChatGPTSplit any recording into TWO files: the dialogue alone, and the environment alone β rain, room tone, traffic, a crowd, music. Use the environment file whenever a voice is about to be replaced (a dub, a new narration, a lip-sync, a translation): laying the new voice over a preserved bed keeps the world the shot was recorded in, instead of dropping the speaker into silence. Use the dialogue file when the speech has to be transcribed or re-used clean out of a noisy location. Runs on Cadabra's own machine, so the audio goes to no third party. 1 credit per started minute, billed off the file's REAL duration read server-side. Ceiling is 20 minutes per file. Set confirm:true to run it.
cadabra_series_guide
ChatGPTπ¨ Walk it as ONE proposal, not a questionnaire: every stage filled with its recommended answer, changeable with one word, then one price. The road to a micro drama or any series with recurring characters: the idea, three directions, language and format, the cast (one role at a time), the world, the season grid, the beats of episode one, locking faces, places and voices, making the episode, the review, the post, the next episode. Each call returns the map with where the creator is, what this stage does, what they bring and get, THE ONE QUESTION for this stage with three answers and one recommended, the craft that decides it (hooks, cliffhanger shapes, consistent faces and worlds, dialogue performed in the shot on Seedance, prompt shape, what moves views), and a price map in credits: what is paid once for the season, what every episode costs, the whole season, and their balance. Free, spends nothing. π¨ Call it FIRST whenever someone wants a series, a micro drama, episodes, or "a film like this one" with recurring people, and again at every stage with the choices so far, so the price map follows their choices. Speak to them in their language, one question at a time.
cadabra_social_copy
ChatGPTWrite the post that goes with the film: the caption, the hook line, the hashtags, shaped for the platform. Free. A film with no copy does not get posted, so offer this the moment a film is finished.
cadabra_social_month
ChatGPTFill a month of posting days with ideas, aware of what falls in that month and of what has already worked for this creator. π¨ This is the calendar; cadabra_social_copy writes the words for ONE post. Reach for this when somebody says they do not know what to post, or asks for a month, a plan, or a content calendar. It plans and it never publishes: nothing goes to a real account from here. cadabra_publish is the separate, confirmed step for that. Prices itself, so call once without confirm to get the number.
cadabra_song_phrases
ChatGPTFREE, for a music clip. Before planning ANY music clip, read cadabra_editor_manual section 'music-clip' (free): how international-level clips are made and how to lead the creator there. The song's voice is separated from the music, every sung line is timed, the chorus is found by the lines that repeat, and each line gets its own slice of CLEAN VOICE starting 0.3 s before it. Those slices are what every singing shot is made from: a lip sync fed the full mix with drums under it keeps the mouth half closed, and a sync run over a finished edit breaks at every cut. Use the chorus lines for the hero singing shots (3 to 10 s each). Pass the lyrics when you have them: they anchor the timing.
cadabra_sound_effect
ChatGPTA single sound, flat 1 credit, up to 30 seconds: a door slam, a whoosh, a sword, a notification, a crowd, rain. Describe the SOUND, not the story around it. π¨ THIS IS THE CHEAP ONE, AND IT IS THE RIGHT DEFAULT FOR A SHORT EFFECT. cadabra_generate_sfx runs a different engine that is priced per minute and reserves a large hold up front, which is the right tool for a long ambience or a non-Hebrew spoken line and the wrong one for a two-second door. A creator who wanted one sound saw a fifteen-credit reservation and shipped the film silent. loop:true returns a bed that joins to itself with no seam, for room tone, rain or traffic that has to run under a whole film. Without it the bed has a join every few seconds and the editor has to hide it. Not for narration or dialogue, ever: a spoken line belongs in cadabra_generate_voice, which uses this account's own voices.
cadabra_split_layers
ChatGPTSplit one image (PNG or JPG) into separate transparent layers: the base plus each element on its own PNG, ready to move, recolour or animate. Spends credits: the studio holds the most a split can cost (up to 17 images) and returns the unused part automatically when the job lands, so the final price follows how many layers came back. Without confirm:true it returns the quote only.
cadabra_story
ChatGPTTurn an idea into an actual story: the beats, in order, each one changing the situation. Free. π¨ Cadabra's rule is no story, no video. direct true: push on ordinary world, rule of three, plant/payoff, do not transcribe. lock_dialogue true: return only verbatim lines, no shot list. Run it before cadabra_make_film when the user cares about the result, and show the beats in the same message as the free preview and its price, so one yes approves both.
cadabra_story_rewrite
ChatGPT1 credit. mode hooks: 2 to 8 different opening hooks for one idea, to pick from before the story is written. mode beat: rewrite ONE beat of a story by a direction ("make it funnier", "end on a question"), keeping continuity with the beats before and after it. Without confirm:true it returns the quote only.
cadabra_storyboard
ChatGPTFree. Turn an idea into a shot-by-shot board: order, camera, action, duration, and a still prompt per panel. painted true: also return eight-panel colour-script sheets for cadabra_generate_image look animation. Show this to the user and get a yes before spending credits. Then generate stills with cadabra_generate_image and motion with cadabra_generate_video (engine seedance-25 if they named 2.5), or run cadabra_make_film for the whole cut.
cadabra_storyboard_board
ChatGPT2 credits: 1 to write the story, 1 to build the panels. Quote first (call without confirm), then confirm:true. The engine behind the website's 'AI Storyboard Board' studio. One idea in β a locked character_bible (with a named DNA marker), one style_lock, a character_sheet_prompt, and 4-12 per-panel image_prompts built to read as one production β richer than cadabra_storyboard's board (a real character bible + style lock, not just shot lines). Nothing is rendered here. Render the character sheet FIRST with cadabra_generate_image (locks identity before panel one), then each panel's image_prompt with that sheet's URL as reference_image_urls β same trick as a consistent cast. Pass reference_image_url when the user already has a real photo (a face or a product) to use as the identity instead of a generated one β the sheet step is then skipped entirely, so it isn't returned and shouldn't be generated. Quote each image with cadabra_quote before generating; show the user the board and get a yes before spending anything.
cadabra_talk
ChatGPTA real conversation with Cadabra's Genie, the same director the website has: a thousand-year-old character with taste and opinions, who asks the one question that unlocks an idea, tells you plainly when a version is safe but forgettable, and knows the six named productions (a short film, a character story, a cinematic trailer, a product ad, a viral reel, a micro drama), festival mode for a film meant to win, performance direction by physical beats, and how to diagnose a cut that does not hit. Free. Send the creator's sentence in their language, and after the first turn a short context of what is already settled, and you get the Genie's reply plus two or three next offers. π¨ It answers and directs; it does not spend. When the reply calls for a plan or a shot, run cadabra_make_film, cadabra_generate_video or the tool it names, with a quote first. π¨ When the Genie has pitched the film here and the creator said yes, pass that reply verbatim as cadabra_make_film's director_pitch, so the plan is the film they were promised and not a second story written from the idea alone. Use this for any open creative talk, for "ΧͺΧ©Χ€Χ¨Χ" / "make it better" on a described film, and whenever the creator wants a partner rather than a menu.
cadabra_transcribe
ChatGPTTurn a media file into text with word-level timing. Hebrew supported. 1 credit per started 10 minutes, billed off the file's REAL duration read server-side, so the price is exact and not an estimate. Ceiling is 3 hours per file. Set confirm:true to run it. π¨ ALWAYS PASS THE LANGUAGE YOU EXPECT. Auto-detect on Hebrew is where invented text comes from: a transcriber that is unsure will write a plausible sentence nobody said, and on a short or quiet tail it writes subtitle credits or a sign-off out of silence. Naming the language is free and it is the difference between a transcript and a guess. Two engines. The default reads word timings and speaker labels and does not invent a tail; engine "whisper" runs on Cadabra's own machine and is there as the fallback. Same price either way. π¨ And if the words are already known β a script, a narration you made, lyrics β do not transcribe at all. cadabra_align_script times the words you HAVE, which removes the guessing entirely.
cadabra_translate_captions
ChatGPTTranslate a caption track into another language while KEEPING each line's timing, so the subtitles still land on the words. Free. π¨ Use the aligned mode (the default) for anything that will be burned onto a video: a plain translation returns text with the timing thrown away, and captions whose timing drifted are worse than none. Get the source captions from cadabra_transcribe.
cadabra_translate_video
ChatGPTDub a finished film into other languages IN THE SPEAKER'S OWN VOICE, with the mouth moved to match. 30+ languages, several at once. Hand it a direct https link to the film; Cadabra measures its real length, because the price is per minute of the source and per language. π¨ That per-minute rate is an ESTIMATE until a real minute is measured; quote it as one. Set confirm:true to spend credits.
cadabra_trim_audio
ChatGPTCut an audio file down to a given number of seconds and hand back the trimmed file. FREE: it is a cut, not a generation, and the original is untouched. Use it whenever a track is longer than the film it has to sit under: a music bed that outruns the cut, or a narration whose last line the creator decided to drop. Trimming is always cheaper than generating the audio again, and on a narration it also keeps the take the creator already approved.
cadabra_upload_file
ChatGPTUse this when the user has a file of their own (photo, clip, song, voice) that another tool needs as a link. Call it with no arguments to open an upload button in the card. Not for a file that is already on an https link. Turn file bytes (photo, video, or audio) into public URLs for the other tools. Free, kept 7 days. One file or many (up to 30). Two or more files automatically return ONE batch link that stands for all of them. Pass that single link as reference_image_urls, Seedance 2.5 Omni refs, or a cast list. Do not paste thirty links. Needs real BASE64 BYTES. If you can read the file off disk, do that and call this tool. π IN CHATGPT, A FILE THE USER ATTACHED TO THE CONVERSATION GOES IN chat_files. ChatGPT fills in its download link itself; you never type one and you never need the bytes. Photo, song, voice recording or clip, all of them. This is the road for an attached MP3. π¨ IF YOU CANNOT READ THE BYTES AND chat_files IS NOT OFFERED TO YOU, FIRST CALL cadabra_upload_file WITH NO ARGUMENTS, WHICH OPENS AN UPLOAD BUTTON IN THE CHAT; IF NONE APPEARS, GIVE THEM THE LINK, NOT AN INSTRUCTION: https://livinginmovie.com/mcp/upload (same account): drop the file, paste the link back here. Attached pictures you can see but not reproduce, and audio or video of any kind, belong on that page when chat_files is unavailable. Saying 'upload the file' without that link is a dead end, and people have sat in it for hours.
cadabra_upscale_image
ChatGPTMake a STILL bigger and sharper. This is the image upscaler; cadabra_upscale_video is the separate one for footage. Β· fast β ESRGAN at 2x or 4x. 1 credit. The everyday answer. Β· crispy β Clarity, which rebuilds detail rather than only enlarging. Billed per megapixel of OUTPUT and the studio measures the real file, so the exact number comes back with the result. Reach for it on a hero still, not on a draft. Set confirm:true to spend.
cadabra_upscale_video
ChatGPTUpscale an existing video to REAL 4K with Topaz Starlight Precise 2.5, a restoration model: it rebuilds skin, hair, fabric and text on AI and camera footage while keeping faces faithful. The output always has 2160 pixels on the clip's SHORT side (a vertical 1080x1920 clip becomes 2160x3840, a 720x1280 one x3), measured from the file itself. Give the CLEAN original, not an export that was already compressed or filtered: an upscale cannot restore what a filter destroyed. Cadabra measures the clip's real length, because that is what it bills on. 1 credit per second, minimum 2, clips up to 600 seconds. Starlight takes about 1.2-2 minutes per second of video; quality "fast" (Proteus) is quicker and softer, same price. Set confirm:true to spend credits. A failed run is auto-refunded.
cadabra_video_codes
ChatGPTThe 61 slash codes that turn ONE picture into a shot, on 7 shelves: camera moves (/fpvdrone /gimbalshot /orbitshot /360 /flyover /spiralshot /craneup /crashzoom /dollyzoom /zoomout /whippan /handheld /lowtrack /snorricam /parallax), time (/slowmotion /bullettime /speedramp /timelapse /hyperlapse /freeze /floatfall), product (/tvcommercial /explode /assemble /liquid /shatter /unbox /levitate /macrozoom /macroreveal), people (/glambot /ugc /pov /thirdpersonview /herowalk /runway /reaction /dance /sport), world (/earthzoom /weather /daytonight /crowd /underwater /fire /bloom) and looks (/cinematic /noir /retro80s /documentary). Every one has a Hebrew name for the Hebrew keyboard and a line on what it wants attached. FREE. π¨ Behind each code is a full English production spec β identity lock, shot, camera body and glass, one move, light, grade, physics, sound and an explicit avoid list β written for the engine that will render it. π¨ Do not hand the creator the list and stop there: the code RUNS here. Their picture goes to cadabra_upload_file if it is not on a URL yet, then cadabra_generate_video with code plus image_url quotes the shot and confirm:true renders it (or cadabra_generate_kling for Kling O3). Call this when they ask what the codes are, when they type one you do not recognise, or when they have a photo and no idea what to do with it.
cadabra_video_effect
ChatGPTChange what is INSIDE an existing clip, by name instead of by phrasing. Effects: wardrobe (change what they are wearing), world (change the place around them), weather (change the weather), time-of-day (change the hour and the light), light (relight the shot without restaging it), sky (replace the sky and match the light to it), swap (put one thing in place of another), add (put something into the shot), color (change the colour of one thing), material (change what a thing is made of), remove (take something out of the shot), declutter (clear the background), restyle (restyle the whole picture), season (change the season). Each one carries a deliberate instruction that holds everything else still, which is what makes the result repeatable: the same ask gives the same kind of answer instead of depending on how the sentence happened to be worded. The output keeps the source clip's length and shape. Priced off the SOURCE clip's length and measured from the file itself, so a longer clip costs more; set confirm:true to spend, and without it you get a quote only. π¨ This cannot swap the person for one of your saved characters: the engine takes the clip and the words and no identity. For anything more than the face (build, wardrobe, the whole person), ask for a fresh render with that character instead.
cadabra_voice_tags
ChatGPTFree. Takes a plain line and returns the same line marked with the delivery tags the voice engine actually obeys β [whispers], [excited], [sighs], [laughs], a pause where the sentence needs one. Feed the RESULT to cadabra_generate_voice as the text. This is the difference between a script being read and a script being performed, and it costs nothing to try. Reach for it whenever a read came back flat, rushed, or evenly loud from beginning to end: the fix is almost never a different voice, it is the same voice told where to breathe. Up to 5000 characters, Hebrew included. Only tags the engine understands are used, so the output is always safe to speak.
cadabra_whats_new
ChatGPTWhat is NEW in Cadabra, newest first, with the tool each change belongs to. Free. Call this when the user asks what is new, what changed, or what you can do now that you could not before, and whenever they seem to be working from an older idea of what this connector can do. Everything it lists is callable in this chat right now. For the full current capability list rather than the changes, call cadabra_catalog.
cadabra_world_sheet
ChatGPTOne sentence in any language, one location reference sheet out: the same place from four angles at one time of day, with its light, its palette and the three things it is recognised by, the thing you lock a world with before any film is made. The place is first DESIGNED in English by the same brain the films use (free). π¨ A prompt that comes back Hebrew is refused before anything is charged. Without confirm:true you get the English prompt and the price only.
