jsoncut_abort_upload
ChatGPTAbort an owned unfinished upload and remove temporary parts. A committing or completed upload cannot be aborted; check its status instead.
jsoncut_add_canvas
ChatGPTAdd one procedural canvas background layer from a supported preset. Read the current project version first; on conflict reread before retrying. Use the returned version for the next mutation. Persists changes in the connected account. Retrying a successful mutation can duplicate or repeat the change.
jsoncut_add_caption_track
ChatGPTAdd one editable caption track from timed text cues and an explicit caption style. Read the current project version first; on conflict reread before retrying. Use the returned version for the next mutation. Persists changes in the connected account. Retrying a successful mutation can duplicate or repeat the change.
jsoncut_add_captions
ChatGPTAdd one editable, word-timed caption track from speech. source="auto" transcribes the project's speech (music and sound effects are left out); a clip id transcribes only that clip as it plays on the timeline (trim, speed and position included). Returns the new version and the spoken ranges. Uses AI credits. request_key: only to follow up a call that did not answer (its error names the key). Transcribes project speech with AI credits and adds a caption track to the saved project. Retrying adds another track.
jsoncut_add_component
ChatGPTAdd one editable component from the supported browser-window, kpi-card, notification or device-frame presets. Read the current project version first; on conflict reread before retrying. Use the returned version for the next mutation. Persists changes in the connected account. Retrying a successful mutation can duplicate or repeat the change.
jsoncut_add_custom
ChatGPTAdd one custom canvas-2d, SVG, DOM or WebGL layer with its code and optional inputs. Read the current project version first; on conflict reread before retrying. Use the returned version for the next mutation. Optional scene_id restricts this same operation to that scene: timing is scene-relative, targets must belong to it, and its boundary cannot be removed. Persists changes in the connected account. Retrying a successful mutation can duplicate or repeat the change.
jsoncut_add_media
ChatGPTAdd one image, video or audio layer from an already uploaded mediaId; does not fetch external URLs. Read the current project version first; on conflict reread before retrying. Use the returned version for the next mutation. Optional scene_id restricts this same operation to that scene: timing is scene-relative, targets must belong to it, and its boundary cannot be removed. Persists changes in the connected account. Retrying a successful mutation can duplicate or repeat the change.
jsoncut_add_repeat
ChatGPTAdd one repeat layer: a list of texts or pictures laid out as a grid, row, column or circle, the cells entering staggered. A wall of cards, logos or chips is one such layer, not many. Read the current project version first; on conflict reread before retrying. Use the returned version for the next mutation. Optional scene_id restricts this same operation to that scene: timing is scene-relative, targets must belong to it, and its boundary cannot be removed. Persists changes in the connected account. Retrying a successful mutation can duplicate or repeat the change.
jsoncut_add_scene
ChatGPTAdd one scene container with an explicit start, duration and track. Read the current project version first; on conflict reread before retrying. Use the returned version for the next mutation. Persists changes in the connected account. Retrying a successful mutation can duplicate or repeat the change.
jsoncut_add_scene_recipe
ChatGPTLay out one whole scene from a proven structure (hook, feature list, KPI, quote, steps, split media, lower third, kinetic words, call to action, logo reveal): fitted text, reading-order entrances and a slow background move as ordinary editable layers. Read the current project version first; on conflict reread before retrying. Use the returned version for the next mutation. Persists changes in the connected account. Retrying a successful mutation can duplicate or repeat the change.
jsoncut_add_shape
ChatGPTAdd one editable rectangular shape layer with color, dimensions, timing and optional corner radius. Read the current project version first; on conflict reread before retrying. Use the returned version for the next mutation. Optional scene_id restricts this same operation to that scene: timing is scene-relative, targets must belong to it, and its boundary cannot be removed. Persists changes in the connected account. Retrying a successful mutation can duplicate or repeat the change.
jsoncut_add_sound_effect
ChatGPTPlace a library sound (id from jsoncut_search_sounds) at a moment of the video: its loudest point lands on at (a cut, an entrance, a word), or align="start" begins it there. Effects share one track, open another only where two overlap and never duck the music; ambience runs until the end of the video. Returns the new version. A few accents per video, not on top of a spoken word. Copies a library sound into the project media and adds it to the saved project. Retrying adds another clip.
jsoncut_add_svg
ChatGPTAdd one SVG image layer. Prefer native text and shape layers for independently editable copy and controls. Read the current project version first; on conflict reread before retrying. Use the returned version for the next mutation. Optional scene_id restricts this same operation to that scene: timing is scene-relative, targets must belong to it, and its boundary cannot be removed. Persists changes in the connected account. Retrying a successful mutation can duplicate or repeat the change.
jsoncut_add_text
ChatGPTAdd one independently editable text layer with position, dimensions, timing and styling. Read the current project version first; on conflict reread before retrying. Use the returned version for the next mutation. Optional scene_id restricts this same operation to that scene: timing is scene-relative, targets must belong to it, and its boundary cannot be removed. Persists changes in the connected account. Retrying a successful mutation can duplicate or repeat the change.
jsoncut_add_variable
ChatGPTAdd one template field (text, number, color, boolean, media, HTML or list) whose default is the current content; bulk rendering and the API fill it per version. Read the current project version first; on conflict reread before retrying. Use the returned version for the next mutation. Persists changes in the connected account. Retrying a successful mutation can duplicate or repeat the change.
jsoncut_add_voiceover
ChatGPTSpeak a script with a voice and add it to the video timeline: one clip per paragraph (blank line), back to back from at; with captions=true also a word-timed caption track from the script itself (no second transcription). Subtitles are optional: add them only when they are wanted. Tags like [excited] or [whispers] steer the delivery and never appear in captions. Returns the new version, the clips and the spoken ranges with their text, so later layers can be timed to the words. Uses AI credits. request_key: only to follow up a call that did not answer (its error names the key); the call then returns that saved result instead of generating again. Generates speech with AI credits, stores the audio and adds clips and captions to the saved project. Retrying adds another voiceover.
jsoncut_align_layers
ChatGPTAlign two or more layers that share a parent to their combined bounds, or one layer to its canvas, group or scene. Read the current project version first; on conflict reread before retrying. Use the returned version for the next mutation. Optional scene_id restricts this same operation to that scene: timing is scene-relative, targets must belong to it, and its boundary cannot be removed. Persists changes in the connected account. Retrying a successful mutation can duplicate or repeat the change.
jsoncut_append_child_html
ChatGPTAppend a complete HTML fragment inside one parent layer. Read the current project version first; on conflict reread before retrying. Changes existing saved account data; may replace or remove existing content.
jsoncut_append_root_html
ChatGPTAppend a complete HTML fragment to the project root. Read the current project version first; on conflict reread before retrying. Changes existing saved account data; may replace or remove existing content.
jsoncut_apply_brand_kit
ChatGPTApply a workspace brand kit (from list_brand_kits) to a project: its four colors become palette roles, its heading and body fonts and logo become brand tokens, and existing layers are mapped onto them. Applying another kit later only swaps the tokens. link_template_fields=true adds the template fields brandPrimary, brandSecondary, brandBackground, brandText, headingFont, bodyFont and brandLogo, so bulk rendering and the API can render one design for several brands. Afterwards use brand.primary, brand.secondary, brand.background, brand.text and brand.heading, brand.body as color and fontFamily values in authoring actions. Replaces the project colors, fonts and logo with the brand kit and maps existing layers onto them.
jsoncut_apply_motion
ChatGPTApply one discovered motion preset to an existing layer, replacing its motion settings. Read the current project version first; on conflict reread before retrying. Use the returned version for the next mutation. Persists changes in the connected account. Retrying a successful mutation can duplicate or repeat the change.
jsoncut_apply_text_motion
ChatGPTSet editable text animation on one existing text layer. Omit exitEffect to preserve its exit animation; an entrance does not imply an exit. Read the current project version first; on conflict reread before retrying. Use the returned version for the next mutation. Persists changes in the connected account. Retrying a successful mutation can duplicate or repeat the change.
jsoncut_assemble_script
ChatGPTRough cut from a script (plain text; read a PDF first): finds each line in the speech of the project's files, takes the best take (cleanest; among equal ones the last) and lays the takes out in the script's order, up to hundreds of clips. Uses stored transcripts and transcribes nothing: files without one (transcribed in Studio) are listed as noTranscript. apply=false answers the takes and the lines not found; apply=true saves one version. apply=false only reports what would change; apply=true saves one project version. Uses stored transcripts and FFmpeg or analysis only: no AI model, no AI credits (copies of files count as storage).
jsoncut_author_project
ChatGPTApply bounded editable layers, optional scenes, SVG, custom blocks or keyframes. Prefer add_text/add_shape/add_media for designed compositions; use SVG as an intentional hybrid/atomic layer when appropriate. Text inside compact controls must share their bounds and be centered. Compatibility batch of the same explicitly registered operations: jsoncut_insert_library_item, jsoncut_add_scene, jsoncut_add_text, jsoncut_add_caption_track, jsoncut_add_shape, jsoncut_add_component, jsoncut_add_canvas, jsoncut_update_clip, jsoncut_update_canvas, jsoncut_apply_motion, jsoncut_apply_text_motion, jsoncut_group, jsoncut_add_media, jsoncut_add_svg, jsoncut_add_custom, jsoncut_patch_clip, jsoncut_patch_svg, jsoncut_patch_custom, jsoncut_ungroup, jsoncut_delete_clip. Prefer the individual tool for one change. The complete typed action union is declared in this tool schema; no additional operation is discovered or enabled at runtime. Changes existing saved account data; may replace or remove existing content.
jsoncut_author_scene
ChatGPTEdit one long-video scene by its canonical outline id with <=24 typed actions. Times are scene-relative. Send only one mutation call per response, then use its returned version. Compatibility batch of the same explicitly registered operations: jsoncut_insert_library_item, jsoncut_add_text, jsoncut_add_shape, jsoncut_group, jsoncut_add_media, jsoncut_add_svg, jsoncut_add_custom, jsoncut_patch_clip, jsoncut_patch_svg, jsoncut_patch_custom, jsoncut_ungroup, jsoncut_delete_clip. Prefer the individual tool for one change. The complete typed action union is declared in this tool schema; no additional operation is discovered or enabled at runtime. Changes existing layers; uses account-scoped reads and version checks.
jsoncut_auto_mix
ChatGPTMix the sound in one step: listens to the project (as jsoncut_inspect_audio), applies every measured repair (music under speech, levels) and sets the master to target_lufs (-14 for web and social) with a -1 dBTP ceiling. Returns what changed; inspect_audio again to hear the result. Applies the measured sound repairs and the master loudness to the saved project.
jsoncut_bind_audio_motion
ChatGPTMeasure an owned audio/video source and bind its real energy to a visual layer. Source limit 300s/250MB. Studio can edit/remove this binding; trim, offset, rate and loops follow the audio layer. Channel energy is pre-mix, not invented waveform data; onset markers are not a guaranteed musical beat grid. Scale/opacity amount must be -1..1, y -1000..1000. Rebind after replacing source media. Explicitly patch audioReactive=null to remove. Replaces motion settings on an existing layer.
jsoncut_bind_variable
ChatGPTConnect one layer slot (text, source, html, visible, repeat or a style property) to a template field, or disconnect it with variableId null. Read the current project version first; on conflict reread before retrying. Use the returned version for the next mutation. Optional scene_id restricts this same operation to that scene: timing is scene-relative, targets must belong to it, and its boundary cannot be removed. Persists changes in the connected account. Retrying a successful mutation can duplicate or repeat the change.
jsoncut_bulk_render_template
ChatGPTQueue up to 500 independently validated rows against one template version. Queues multiple renders using existing allowance; optional webhook delivery cannot be recalled. Retrying can duplicate the batch.
jsoncut_camera
ChatGPTMove the camera on one video, picture or scene (push, pull, punch-in on a spoken word, zoom-to a detail, pan, whip, shake, settle); expands into editable keyframes. Read the current project version first; on conflict reread before retrying. Use the returned version for the next mutation. Persists changes in the connected account. Retrying a successful mutation can duplicate or repeat the change.
jsoncut_cancel_bulk_render
ChatGPTCancel a running template batch.
jsoncut_cancel_render
ChatGPTCancel a queued or running render.
jsoncut_check_project
ChatGPTOptionally persist a version-bound quality review; works without cut.md and never gates rendering. full_playback_confirmed is an explicit caller attestation: leave false for grids, onion skin, metadata or sampled frames. Only attest after actual complete playback in a capable client or explicit human confirmation, never on render success. Evidence accepts exact-frame renders/crops; reference comparisons require a cut.md-linked image. Persists a quality review and may queue evidence renders; this is not a read-only check.
jsoncut_clean_speech
ChatGPTCut pauses, filler sounds and words said twice out of spoken footage in one edit. apply=false answers what would go and the new length; apply=true saves it as one version. Pauses come from the file's analysis; filler sounds and repeated words need a stored transcript (untranscribed lists clips without one). Never transcribes. apply=false only reports what would change; apply=true saves one project version. Uses stored transcripts and FFmpeg or analysis only: no AI model, no AI credits (copies of files count as storage).
jsoncut_connect_preview
ChatGPTPrepare private playback of one owned export. UI-only; no public link, render or project changes.
jsoncut_connect_studio
ChatGPTEstablish a short-lived editor session for the connected account. UI-only; does not edit, render or share content.
jsoncut_create_batch
ChatGPTCreate one central production ledger for projects, style variation, asset uniqueness and approvals. Persists a new record in the connected account. Retrying can create a duplicate.
jsoncut_create_project
ChatGPTCreate one durable HTML-first image or video project. Set source.editabilityTarget to layered, hybrid or atomic when Studio editing granularity matters. Empty variables/customBlocks may be omitted. Persists a new record in the connected account. Retrying can create a duplicate.
jsoncut_create_upload_ticket
ChatGPTCreate a 60-900s one-use URL. Requires a client that can read and HTTP-PUT the real file bytes; OAuth alone does not grant that capability. PUT application/octet-stream with exact Content-Length. If the response is lost, get_upload_status before retrying. If client egress is blocked, use get_upload_handoff, never a proxy workaround. Persists an upload, upload session or font registration. Retrying can create a duplicate.
jsoncut_cut_ranges
ChatGPTCut timeline spans out in one ripple edit (pauses, filler words, a passage of speech): footage on all tracks splits and later clips move up; locked layers stay and music plays through. Seconds are timeline time. Read the current project version first; on conflict reread before retrying. Use the returned version for the next mutation. Persists changes in the connected account. Retrying a successful mutation can duplicate or repeat the change.
jsoncut_delete_clip
ChatGPTRemove one existing layer or group from the saved project. Removing a group may remove its children. Read the current project version first; on conflict reread before retrying. Use the returned version for the next mutation. Optional scene_id restricts this same operation to that scene: timing is scene-relative, targets must belong to it, and its boundary cannot be removed. Persists changes in the connected account. Retrying a successful mutation can duplicate or repeat the change.
jsoncut_delete_media
ChatGPTDelete an owned upload that is no longer referenced.
jsoncut_delete_project
ChatGPTMove an owned project to the recoverable Studio trash.
jsoncut_distribute_layers
ChatGPTSpace three or more layers that share a parent evenly; the outermost layers keep their position. Read the current project version first; on conflict reread before retrying. Use the returned version for the next mutation. Optional scene_id restricts this same operation to that scene: timing is scene-relative, targets must belong to it, and its boundary cannot be removed. Persists changes in the connected account. Retrying a successful mutation can duplicate or repeat the change.
jsoncut_edit_cut
ChatGPTEditor cuts that keep the track in step. slip: other footage in the same place and length (delta source seconds). ripple_trim: move the start or end (edge) by delta seconds; later layers on the track follow, so no gap or overlap opens. roll: move the cut between clip_id and next_clip_id by delta. freeze_frame: hold the frame shown at project second at for seconds (a still of that frame is made from the file), the clip continues after it and later layers move back. apply=false checks the cut and saves nothing (for freeze_frame the still is already made and lands in the media library); apply=true saves one version.
jsoncut_edit_pages
ChatGPTEdit the pages of an image carousel (1-based): set the page count, add, duplicate, move or remove a page, or repeat one page per item of a text-list field. Read the current project version first; on conflict reread before retrying. Use the returned version for the next mutation. Persists changes in the connected account. Retrying a successful mutation can duplicate or repeat the change.
jsoncut_emphasize
ChatGPTAdd one creator overlay at a spoken word or time (keyword pop, icon sticker, circle, underline or arrow on a layer, progress bar, follow pill) as editable layers with entrance and exit. Read the current project version first; on conflict reread before retrying. Use the returned version for the next mutation. Persists changes in the connected account. Retrying a successful mutation can duplicate or repeat the change.
jsoncut_get_authoring_guide
ChatGPTRead concise project-structure rules before authoring or restructuring.
jsoncut_get_batch
ChatGPTRead batch checklist, ordered projects, latest reviews, asset provenance and uniqueness conflicts.
jsoncut_get_bulk_render
ChatGPTRead batch progress and per-row outcomes.
jsoncut_get_component
ChatGPTRead one reusable component and its typed slots.
jsoncut_get_download_link
ChatGPTCreate a temporary download/playback link for an owned completed export when the user requests delivery or sharing. Anyone with the link can access only this export without Studio login. Defaults to 1 hour, never beyond artifact expiry. This creates a revocable share, not a local file; do not create links during status polling or private review. Return the URL and expiry to the user; host download tools may still have network restrictions. Creates a bearer-access share outside the private account boundary. Do this only for requested delivery; never for polling. Retrying creates another link.
jsoncut_get_layer
ChatGPTRead one exact editable V2 layer. Request markup only when a source-level replacement is necessary.
jsoncut_get_library_item
ChatGPTRead one native Library item including typed inputs and its declarative HTML example.
jsoncut_get_media
ChatGPTGet one upload including its stable jsoncut-media reference.
jsoncut_get_primitive_catalog
ChatGPTDiscover exact Studio primitives. Controls are editable grouped-layer recipes, not hidden presets. Search icons by at least two characters; returned SVG comes from the Studio Lucide library and must not be redrawn.
jsoncut_get_project
ChatGPTRead a project including its current source and optimistic version.
jsoncut_get_project_context
ChatGPTRead one compact project payload with active cut.md guidance and available media, components, templates, brand kits and fonts. Prefer this over several catalogue calls before authoring.
jsoncut_get_project_cut
ChatGPTRead the latest immutable cut.md version and resolved-reference state.
jsoncut_get_project_outline
ChatGPTRead a compact paged root timeline and canonical scene ids without returning the large HTML source. Use these ids for later scene calls; optional for long projects.
jsoncut_get_project_scope
ChatGPTRead a bounded layer window from a long V2 project. Filters are server-side and full HTML/CSS is not returned.
jsoncut_get_render
ChatGPTRead status; optionally wait and return the actual PNG/JPEG/WebP review image. For final file delivery use get_download_link; artifactUrl requires authentication outside MCP.
jsoncut_get_review
ChatGPTGet persisted findings, evidence, staleness and approval state for one review or the latest review.
jsoncut_get_scene
ChatGPTRead one scene and a compact paged list of descendants. relativeStart matches author_scene; canonical start remains visible for orientation.
jsoncut_get_template
ChatGPTRead a published template's metadata and source.
jsoncut_get_template_inputs
ChatGPTGet required typed inputs and examples before filling a template.
jsoncut_get_upload_handoff
ChatGPTGet the authenticated Studio project for a human-assisted upload when the client cannot transfer bytes. A link is not upload completion: list project media afterwards.
jsoncut_get_upload_status
ChatGPTRead an owned upload session, including its durable media reference on completion. Use after a lost PUT response; never resend a consumed ticket blindly.
jsoncut_group
ChatGPTCreate one named group from selected existing layers, using an explicit action id. Read the current project version first; on conflict reread before retrying. Use the returned version for the next mutation. Optional scene_id restricts this same operation to that scene: timing is scene-relative, targets must belong to it, and its boundary cannot be removed. Persists changes in the connected account. Retrying a successful mutation can duplicate or repeat the change.
jsoncut_group_layers
ChatGPTGroup selected layers into one editable group. Read the current project version first; on conflict reread before retrying. Changes existing saved account data; may replace or remove existing content.
jsoncut_import_host_file
ChatGPTImport a host-authorized file reference into a project, without model-generated base64. Requires the host to supply a real HTTPS download_url; sandbox paths are not downloadable. If MIME is omitted by the host, provide mime_type. Cloud-generated files are supported only if the host makes them available as file inputs. Fetches the supplied authorized URL and persists media in the connected account.
jsoncut_import_media_url
ChatGPTFetch, validate and pin an allowed public image, video or audio URL. Project-scoped imports record asset usage and apply active cut.md uniqueness rules when present. Fetches the supplied authorized URL and persists media in the connected account.
jsoncut_import_stock_media
ChatGPTImport one selected stock result into an owned project, preserving provenance and project uniqueness. Use an unchanged result from search_stock_media. Fetches and persists the selected asset in the connected account.
jsoncut_insert_library_item
ChatGPTInsert one versioned library item, or apply one library effect to a target layer. Use the exact item id, version and inputs returned by library discovery. Read the current project version first; on conflict reread before retrying. Use the returned version for the next mutation. Optional scene_id restricts this same operation to that scene: timing is scene-relative, targets must belong to it, and its boundary cannot be removed. Persists changes in the connected account. Retrying a successful mutation can duplicate or repeat the change.
jsoncut_inspect_audio
ChatGPTListen to a project before and after audio edits: loudness per track and of the mix (LUFS, true peak) against the -14 LUFS web target, where each track speaks, the caption words, the music's beats, cuts and findings (mix off target, peaks above -1 dBTP, music or sound competing with speech, silence, muted speech). include_image adds the sound sheet: one row per track with its loudness curve, speech and words, the mix with the target line, cuts and numbered findings. At most 180 s per call. Measures the project sound (loudness per track and mix, speech, findings) and returns a sound sheet image; changes nothing.
jsoncut_inspect_media
ChatGPTRead safe technical metadata for exactly one owned upload or public URL. May fetch a caller-supplied public media URL; no asset is persisted.
jsoncut_inspect_motion
ChatGPTMeasure the saved project's motion at every frame of a window of at most 12 s (default start..start+12): one-frame position jumps, size jumps and opacity pops, whatever animates the layer. Returns the measured report (authoritative) and one annotated image: dimmed end frame, faint earlier copies, one path dot per frame, red jumps, timing strip. Use it before claiming or fixing motion problems and again after a fix; leave layers it reports smooth unchanged. clip_ids picks up to 4 layers to draw. It does not decode video media or listen to audio; review_motion inspects a finished export. Queues a preview render and uses existing render allowance. Retrying can queue another render.
jsoncut_list_brand_kits
ChatGPTList workspace brand colors and typography defaults.
jsoncut_list_components
ChatGPTList reusable motion components available to the workspace.
jsoncut_list_fonts
ChatGPTList uploaded workspace fonts. Built-in render-safe fonts are in the authoring guide.
jsoncut_list_media
ChatGPTList reusable uploads, optionally scoped to a project or template.
jsoncut_list_projects
ChatGPTList durable editable image and video projects.
jsoncut_list_templates
ChatGPTSearch published image and video templates.
jsoncut_list_voices
ChatGPTList the voiceover voices (gender, accent, style), delivery tones and the languages for writing and dubbing. Every voice speaks every supported language; write the text in the target language.
jsoncut_move_sound_hit
ChatGPTMove one sound of a Sound effects clip to a new start (project seconds) and/or set its volume (0 to 1). clipId and hitId come from jsoncut_add_sound_effect or the soundHits of the project outline. Read the current project version first; on conflict reread before retrying. Use the returned version for the next mutation. Persists changes in the connected account. Retrying a successful mutation can duplicate or repeat the change.
jsoncut_open_studio
ChatGPTOpen the existing JsonCut Studio editor beside the chat. Pass an owned project_id to edit the original layered project. With no ID, show a project picker. Does not render, share or change a project. The editor uses the connected OAuth account; no second login is needed when the host supports embedded cookies.
jsoncut_patch_clip
ChatGPTPatch the supplied properties of one existing layer. Explicit null values clear supported animation or transition settings; omitted fields remain unchanged. Read the current project version first; on conflict reread before retrying. Use the returned version for the next mutation. Optional scene_id restricts this same operation to that scene: timing is scene-relative, targets must belong to it, and its boundary cannot be removed. Persists changes in the connected account. Retrying a successful mutation can duplicate or repeat the change.
jsoncut_patch_custom
ChatGPTReplace the JavaScript and supplied HTML, CSS or inputs of one existing custom layer. Read the current project version first; on conflict reread before retrying. Use the returned version for the next mutation. Optional scene_id restricts this same operation to that scene: timing is scene-relative, targets must belong to it, and its boundary cannot be removed. Persists changes in the connected account. Retrying a successful mutation can duplicate or repeat the change.
jsoncut_patch_layers
ChatGPTOne change on every layer matching kind, role (captions: voiceover-captions), part of the name or parent, e.g. the style of all captions. patch takes the fields of patch_clip except what differs per layer (times, text, position). apply=false answers how many layers match; apply=true saves one version. apply=false only reports what would change; apply=true saves one project version. Uses stored transcripts and FFmpeg or analysis only: no AI model, no AI credits (copies of files count as storage).
jsoncut_patch_svg
ChatGPTReplace the SVG markup of one existing SVG layer. Read the current project version first; on conflict reread before retrying. Use the returned version for the next mutation. Optional scene_id restricts this same operation to that scene: timing is scene-relative, targets must belong to it, and its boundary cannot be removed. Persists changes in the connected account. Retrying a successful mutation can duplicate or repeat the change.
jsoncut_process_media
ChatGPTAn edited copy of a video or sound layer's file, made with FFmpeg (no model, no credits beyond storage; files up to 20 minutes). reverse plays the same part backwards; slow-motion makes it 2 or 4 times slower with new frames in between (the layer gets longer, later layers on its track move back); lut applies a .cube file (text in cube, up to 2 MB); mirror repeats one half or a quarter (kaleidoscope); denoise (strength), voice (speech forward: rumble, hiss and level) and voice-effect change the sound. The last five switch every layer that plays the file, keeping their times. apply=false still makes the copy (it lands in the media library) but switches no layer; apply=true also changes the layers and saves one version. A long file can outlast the call: call again with the same arguments; it joins the running work or uses the finished copy.
jsoncut_project_versions
ChatGPTVersion history of a project (the newest 20 saved versions, shared with Studio). list: saved versions with name, origin (studio, agent, api) and time. save: keep the current version (or version) under name; do this before larger changes. restore: copy version back as a new current version; the state before it is saved as "Before restore". The first change over an API key after work in Studio is saved as "Before API changes" by itself. save keeps a named version; restore replaces the saved project with an earlier version (itself kept in the history).
jsoncut_read_transcript
ChatGPTWhat is said in the footage of a project, in seconds of the timeline as it is now, for placing cuts and overlays on spoken moments. Lines per call are limited; more.from continues. words=true adds word times for up to 60 s. Uses the transcript stored on the file; a file without one has no lines. Reads the stored speech of the project in timeline seconds; changes nothing and transcribes nothing.
jsoncut_register_font
ChatGPTRegister an owned, uploaded font with its license and exact weight/style. Use the returned reference in project CSS @font-face and apply the family to native text layers; do not invent a fontFaces source field or silently replace a brand font. Persists an upload, upload session or font registration. Retrying can create a duplicate.
jsoncut_remove_css_block
ChatGPTRemove one named project style block. Read the current project version first; on conflict reread before retrying. Changes existing saved account data; may replace or remove existing content.
jsoncut_remove_sound_hit
ChatGPTRemove one sound of a Sound effects clip by its clipId and hitId; the other sounds stay where they are. Read the current project version first; on conflict reread before retrying. Use the returned version for the next mutation. Persists changes in the connected account. Retrying a successful mutation can duplicate or repeat the change.
jsoncut_render_inline
ChatGPTRender disposable HTML-first input once without creating a project. Queues rendering using existing allowance; optional webhook delivery cannot be recalled. Reuse an idempotency key for retries.
jsoncut_render_project
ChatGPTRender one version-pinned output, preserving project audio unless output.muted=true. Verify the requested soundtrack exists before delivery; a silent draft is not a soundtracked final. cut.md and reviews do not gate ordinary renders. reviewId requests review-attested output. Grids accept frames, times, sampleCount or intervalSeconds. Queues rendering using existing allowance; optional webhook delivery cannot be recalled. Reuse an idempotency key for retries.
jsoncut_render_template
ChatGPTValidate inputs and render one pinned template version. Queues a render using existing allowance; optional webhook delivery cannot be recalled. Reuse an idempotency key for retries.
jsoncut_reorder_layer
ChatGPTChange the stacking position of one layer. Read the current project version first; on conflict reread before retrying. Changes existing saved account data; may replace or remove existing content.
jsoncut_reparent_layer
ChatGPTMove one layer into a different parent group. Read the current project version first; on conflict reread before retrying. Changes existing saved account data; may replace or remove existing content.
jsoncut_replace_clip_html
ChatGPTReplace the complete HTML fragment of one existing layer. Read the current project version first; on conflict reread before retrying. Changes existing saved account data; may replace or remove existing content.
jsoncut_resize_canvas
ChatGPTChange the canvas format, for example 1080×1920 for 9:16 (the page size in carousels); full-canvas layers fill it and other layers scale and keep their anchors. Read the current project version first; on conflict reread before retrying. Use the returned version for the next mutation. Persists changes in the connected account. Retrying a successful mutation can duplicate or repeat the change.
jsoncut_review_motion
ChatGPTInspect a completed MP4/WebM/MOV export as timestamped images, returned directly to vision. First use grid for overview, then narrow export-local start/end around transitions, text and motion. Onion skin adds colored pixel differences alongside originals; use <=2 seconds and preferably a crop. Downloads <=128 MiB, <=12 frames, <=90s; temporary files are removed. Audio stream presence is not listening/sync proof; never confirms full playback.
jsoncut_review_playback
ChatGPTQueue one draft MP4 with project audio by default. Poll get_render; inspect the completed export with review_motion (overview, then short windows). The MP4 link is for a capable player/human, not an image tool result. Frames do not confirm full playback or audible sync. For final quality inspect the final high-quality export as well. Queues a preview render and uses existing render allowance. Retrying can queue another render.
jsoncut_review_scene
ChatGPTQueue a contact-sheet review sampled only across one scene; inspect it later with get_render(include_artifact=true). Queues a preview render and uses existing render allowance. Retrying can queue another render.
jsoncut_revoke_download_link
ChatGPTRevoke an owned download link by the shareId returned by get_download_link. The project and export are not deleted; other links are unaffected.
jsoncut_search_inspiration
ChatGPTSearch workspace templates and components as style/structure inspiration. Returns previews, compact motion maps and cut.md reference drafts, but never imports or reuses their concrete media.
jsoncut_search_library
ChatGPTSearch the shared native V2 Library. Returns IDs, exact versions, editable properties, defaults, examples and compatibility. Insert via author_project/author_scene with type=insert_library_item.
jsoncut_search_sounds
ChatGPTSearch the JsonCut sound-effect library (nearly 500 sounds, CC0/MIT, hosted by JsonCut): whooshes, hits, risers, UI clicks, pops, success and error tones, notifications, typing, camera, money, glitch, sci-fi, magic, crowd, foley, ambience and jingles. English or German words work, with synonyms and small typos ("woosh", "Kasse", "applause"). Each result has its length and peakAt, the moment it hits; place one with jsoncut_add_sound_effect.
jsoncut_search_stock_media
ChatGPTSearch approved stock providers for media. Returns candidates and license information; does not import or change account data.
jsoncut_set_css_block
ChatGPTSet the CSS text of one named project style block. Read the current project version first; on conflict reread before retrying. Changes existing saved account data; may replace or remove existing content.
jsoncut_set_keyframes
ChatGPTSet editable Studio keyframes on one layer. Times are clip-local seconds. Use scene_id only for a layer inside an optional scene; position values are {x,y}, never separate x/y tracks. Keep primary content visible at frame zero and span a meaningful share of the clip instead of ending all motion in its first second. Changes existing layers; uses account-scoped reads and version checks.
jsoncut_set_layer_timing
ChatGPTSet the timing and timeline track of one layer. Read the current project version first; on conflict reread before retrying. Changes existing saved account data; may replace or remove existing content.
jsoncut_set_project_cut
ChatGPTSave one validated immutable cut.md version. Use expected_cut_version from get_project_cut; omit it only for the first version. Changes existing saved account data; may replace or remove existing content.
jsoncut_set_review_status
ChatGPTRecord an optional Agent review decision for the exact project and cut versions. Human approval remains a Studio/API-only decision and ordinary rendering needs neither. Changes existing saved account data; may replace or remove existing content.
jsoncut_stabilize_video
ChatGPTTake the shake out of a video layer: makes a stabilized copy of its file (about 1x its length; up to 20 minutes) and, with apply=true, switches every layer playing that file to it as one version. The picture is 5 to 15 percent larger, times stay the same. No model involved, no credits beyond storage. A long video can outlast the call: Core keeps working, so call again with the same arguments; it joins the running work or uses the finished copy. apply=false only reports what would change; apply=true saves one project version. Uses stored transcripts and FFmpeg or analysis only: no AI model, no AI credits (copies of files count as storage).
jsoncut_track_motion
ChatGPTMake a layer (text, sticker, shape, mask) follow an object in a video layer. region {x, y, width, height} in canvas pixels around the object at project second at (default: the video layer's start); it is tracked until until (default: the end of the shot or the follower) and the follower gets position keys along the path, keeping its offset. Up to 60 seconds of the file per call (less project time on a sped-up clip); the result says trackedUntil, and lostAt where the object was lost (covered, out of the frame): track again from there. No model, no credits. apply=false only reports; apply=true saves one version. apply=false only reports what would change; apply=true saves one project version. Uses stored transcripts and FFmpeg or analysis only: no AI model, no AI credits (copies of files count as storage).
jsoncut_translate_speech
ChatGPTDub one audio or video clip into another language: its speech is transcribed, translated line by line and spoken with the chosen voice where each line was said, with captions from the translation when the source had captions (or captions=true). The original clip is muted unless mute_original is false. Returns the new version, the clips and the spoken ranges. Uses AI credits. request_key: only to follow up a call that did not answer (its error names the key). Transcribes, translates and speaks one clip with AI credits, adds the dubbed clips and captions and mutes the original. Retrying adds another dub.
jsoncut_ungroup
ChatGPTUngroup one existing group while retaining its children, using an explicit action id. Read the current project version first; on conflict reread before retrying. Use the returned version for the next mutation. Optional scene_id restricts this same operation to that scene: timing is scene-relative, targets must belong to it, and its boundary cannot be removed. Persists changes in the connected account. Retrying a successful mutation can duplicate or repeat the change.
jsoncut_ungroup_layer
ChatGPTUngroup one layer while retaining its children. Read the current project version first; on conflict reread before retrying. Changes existing saved account data; may replace or remove existing content.
jsoncut_update_batch
ChatGPTUpdate the central batch state without writing local manifests or QA files. Changes existing saved account data; may replace or remove existing content.
jsoncut_update_canvas
ChatGPTChange the project canvas background. Read the current project version first; on conflict reread before retrying. Use the returned version for the next mutation. Persists changes in the connected account. Retrying a successful mutation can duplicate or repeat the change.
jsoncut_update_clip
ChatGPTUpdate the supplied text, layout or style fields of one existing layer; omitted fields remain unchanged. Read the current project version first; on conflict reread before retrying. Use the returned version for the next mutation. Persists changes in the connected account. Retrying a successful mutation can duplicate or repeat the change.
jsoncut_update_project
ChatGPTReplace a project source after local validation; prefer bounded edits for small changes. Changes existing saved account data; may replace or remove existing content.
jsoncut_upload_media
ChatGPTUpload a small, programmatically encoded data URL (max 3 MiB decoded). Never type/generate binary base64. For large files use host-file import, a raw-byte upload ticket, or Studio handoff. Persists an upload, upload session or font registration. Retrying can create a duplicate.
jsoncut_validate_project
ChatGPTStatic preflight by default, including custom JavaScript syntax. runtime_check=true additionally queues a bounded image/grid check using the real renderer (normal render permissions/credits apply). Poll its runtimeCheck.renderId with get_render. Samples are not full playback validation. runtime_frames optionally targets 1–24 exact source frames, e.g. a previous errorDiagnostic.sourceFrame. Fix CUSTOM_CODE_ERROR before retrying unchanged. With runtime_check=true, queues a render and uses existing render allowance. Static-only calls do not queue work.
jsoncut_validate_project_cut
ChatGPTValidate canonical cut.md Markdown or a typed document without saving it. Provide exactly one representation.
jsoncut_view_frames
ChatGPTClose look: 1–6 full-size frames (JPEG, up to 1568 px), one image per time in the order given, of the saved video project_id or of one owned media file (media_id or handle m3; a picture has one frame). region {x, y, width, height} in canvas/file pixels cuts out and enlarges a part. Use after an overview (review grid, inspect_media filmstrip) where a lot happens fast: more times there, fewer in calm parts. Extracts still frames with existing render allowance; changes no project.