calculate_credits
ChatGPTEstimate credits from either a legacy
creationParams payload or the same v3 payload used by render_video.Overview
Revid turns an idea into a finished short-form video without leaving the chat. Give it a prompt, a script, an article URL, an audio file, or a music track, and Revid writes the script, generates the visuals, adds a voiceover and captions, and renders a vertical video ready for TikTok, Reels, and Shorts. You can estimate the credit cost before rendering, check render progress, export the MP4, and schedule or publish the video to your connected social accounts. Revid also searches viral short-form videos across YouTube, TikTok, Instagram, X, Reddit, Threads, and LinkedIn so you can study what is working before you create. Requires a Revid account with an active plan; renders consume credits from your Revid account.
Tools
cancel_episode
ChatGPTStop a Series Maker run from
create_episode: the step in progress finishes (what it generates is charged), nothing after it starts. Requires an active Revid plan.clone_voice
ChatGPTClone a custom voice from a public sample URL. Requires an Elite plan or better on Revid.
create_character
ChatGPTCreate a new consistent character from a PNG or JPEG image URL. Requires an active Revid plan.
create_episode
ChatGPTMake a whole episode of a series with Revid's Series Maker, on Autopilot: the server plans the story and shot list, draws each character, place and prop once, generates every scene clip with them so they look the same in every scene, and assembles the episode with music, captions and the series outro. Without
seriesId this starts a new series; with a seriesId from list_series it makes that series' next episode with the same cast, places, style, ratio, model and language (do not send those then). Returns a runId: poll get_episode_status every 30 to 60 seconds (a whole episode usually takes 10 to 30 minutes). When it is done, call export_video with its pid, then get_project_status until videoUrl. Credits are charged per reference image and scene clip as they come back, never past maxCredits (default 2500): the run stops with status needs_attention first. Call get_account to check the balance before a long Ultra episode. Requires an active Revid plan.create_upload_url
ChatGPTGet a public cdn.revid.ai URL for a local image, video or audio file, to use wherever Revid takes a media URL (render_video media and references, generate_media references and frames, create_character, clone_voice). This server cannot read your files: the call returns a one-time
uploadUrl; PUT the file bytes to it with the returned headers within expiresIn seconds, e.g. curl -X PUT -H "Content-Type: <contentType>" --data-binary @<path> "<uploadUrl>". The file is then served at url. With a shell and the Revid CLI, revid upload <file> does both steps. Without a way to send the bytes, ask the user for a public link instead. Uses no credits.delete_character
ChatGPTDelete a saved consistent character or trained character by id. Requires an active Revid plan.
delete_project
ChatGPTDelete a Revid project: it leaves the project list and its public page goes offline. A project still generating cannot be deleted. Confirm with the user first.
duplicate_project
ChatGPTDuplicate an existing Revid project: same content and folder, a new project id. Uses no credits.
export_video
ChatGPTExport an mp4 for an existing Revid project, with the export modal's resolution, frame rate and compression settings. The call waits up to 40 seconds. An export done by then returns status "ready" with its videoUrl. A longer render (most videos over a minute, 4k) returns status "processing" with an exportId: call
get_export_status with it until ready or failed. Do not call export_video again for the same project while it is processing: every call charges a new export. Requires an active Revid plan.generate_media
ChatGPTGenerate standalone image or video media using the same prompt, quality, ratio, preset, duration, frame, reference, audio, and character options as the Revid media generator. The media is saved to the user's Revid media library. The call waits up to 40 seconds. A generation done by then returns status "ready" with each media's url. A longer one (ultra images often, videos usually: 1 to 10 minutes) returns status "processing" with mediaIds: call
get_media_status with them until the status is ready or failed. Do not call generate_media again for the same request while it is processing: every call starts and charges a new generation. Requires an active Revid plan.get_account
ChatGPTGet the Revid account behind the API key: credits left, plan and subscription status, and the workspace.
get_edit_job
ChatGPTFollow a background edit edit_project started (a motion graphic rebuild, a generated clip). Waits up to 40 seconds per call; returns status running, applied or failed. A rebuild comes back with sourceChanges, the lines it changed: check they all belong to the request. Do not start the edit again while it runs: that charges again.
get_episode_status
ChatGPTFollow a Series Maker run from
create_episode: status (queued, running, done, needs_attention, failed, cancelled), step (plan, approve_plan, budget, references, scenes, assemble, done), credits, and once planned the episode's title and scenes. When done, pid is the editor project: export it with export_video. needs_attention means it stopped before spending more; message says why. Requires an active Revid plan.get_export_status
ChatGPTFollow an
export_video call that returned status "processing": pass its exportId. Waits up to waitSeconds (default 40) for the render, then returns status "ready" with the videoUrl, "failed" with the reason, or still "processing" with the render's progress (0 to 1): then call it again. Renders usually take 1 to 10 minutes; one still rendering after an hour fails with Timeout. Requires an active Revid plan.get_media_status
ChatGPTFollow a
generate_media call that returned status "processing": pass its mediaIds. Waits up to waitSeconds (default 40) for the generation to finish, then returns status "ready" with each media's url, "failed" with the reason, or still "processing": then call it again with the same mediaIds. Images usually take under a minute, videos 1 to 10 minutes; a media still generating after an hour fails with Timeout. Requires an active Revid plan.get_project_status
ChatGPTFetch the current status for an existing Revid project id (
pid). Requires an active Revid plan.get_publish_status
ChatGPTCheck whether a post created by
publish_now or schedule_publish went out, per platform (queued, processing, published, failed, with post links and error reasons). TikTok reports processing for a few minutes until it confirms the post.get_timeline
ChatGPTRead what is on a project's timeline: every item's id, type, track and timing, text and caption styling, a motion graphic's props, plus editRevision and editorUrl (the link to give your user to open it in the editor). Call this first when changing an existing video. Changing it beats a new render_video whenever the video is mostly right: nothing already made is generated again.
list_characters
ChatGPTList saved consistent characters available to the authenticated user. Requires an active Revid plan.
list_music_tracks
ChatGPTList the background music library (
music.trackName for render_video) with moods, descriptions and lengths. Use it only when the user asks for a library track: you cannot listen to them, so music.generateMusic: true is the better default.list_presets
ChatGPTList the style presets of Revid's create form: visual presets (
media.mediaPreset for moving-image and ai-video), motion graphics categories (media.mediaPreset for motion-graphics) and caption presets (captions.preset).list_projects
ChatGPTList projects owned by the authenticated API key user.
list_series
ChatGPTList the account's Series Maker series (newest first) with their style, ratio, video model and episode count. Pass a
seriesId to create_episode to make its next episode with the same characters and places. Requires an active Revid plan.list_social_accounts
ChatGPTList the social accounts (TikTok, YouTube, Instagram, X, Facebook Pages, webhooks) and channels connected to the Revid account or active workspace, without tokens. Call this before
publish_now or schedule_publish to pick a destination: pass accounts[].id as <platform>.id (or accounts[].username as <platform>.username, facebook.name for Pages), or channels[].id as channelId. Accounts with tokenInvalid: true must be reconnected by the user in Revid settings before publishing. Facebook Pages publishing is restricted to admins and the designated Meta review account while Meta review is pending. It is not generally available, and changing plans does not unlock Facebook Connect. Use only connected Pages returned by list_social_accounts.list_voices
ChatGPTList the narration voices of the Revid voice picker (gender, accent, age, language, preview URL) to pick
voice.voiceId for render_video. The account's cloned voices come first.publish_now
ChatGPTPublish a rendered video immediately using the selected channel or connected accounts (TikTok, YouTube, Instagram, X, Facebook Pages, webhook). Use
list_social_accounts first to find ids and handles. Facebook Pages publishing is restricted to admins and the designated Meta review account while Meta review is pending. It is not generally available, and changing plans does not unlock Facebook Connect. Use only connected Pages returned by list_social_accounts.rename_project
ChatGPTRename an existing Revid project.
render_video
ChatGPTStart a new Revid render using the v3 public API. Returns a
pid you can poll with get_project_status. Requires an active Revid plan.schedule_publish
ChatGPTSchedule a rendered video for the selected channel or connected accounts (TikTok, YouTube, Instagram, X, Facebook Pages, webhook). Pass
scheduledTime to publish at an exact date and time; omit it to use the next open queue slot. Use list_social_accounts first to find ids and handles. Facebook Pages publishing is restricted to admins and the designated Meta review account while Meta review is pending. It is not generally available, and changing plans does not unlock Facebook Connect. Use only connected Pages returned by list_social_accounts.search_viral_videos
ChatGPTSearch viral short-form videos and posts across YouTube, Instagram, TikTok, X, Reddit, Threads, and LinkedIn (same engine as revid.ai/search). Semantic query plus filters on engagement (min likes/views/shares/comments), platform, language, hashtags, posting date, duration, and AI-analyzed style (mood, content type, perspective, shooting style, visual origin, brand promotion). Each result includes the post URL, a playable video URL when available, the full transcript, a content summary, engagement counts and rate, author info, and hashtags — useful for studying what works before creating content with render_video. Rate limited per plan (free keys get a small daily allowance); a RATE_LIMIT_EXCEEDED error says when the allowance resets.
set_project_visibility
ChatGPTSet who can watch a project's public page (revid.ai/view/<pid>): public (on and listed), unlisted (anyone with the link) or private (page off). Returns the public URL.
upload_files
ChatGPTCopy files the user attached in the chat to Revid and get a public cdn.revid.ai URL for each, to use wherever Revid takes a media URL (render_video media and references, generate_media references and frames, create_character, clone_voice). In ChatGPT, use this for attached images, videos and audio rather than
create_upload_url. Up to 10 files per call, 200 MB each; call it again for more. Use only the url values it returns. If nothing is attached, ask the user to attach the files or give a public link. Uses no credits.view_frames
ChatGPTLook at a project's current timeline: composed frames at the seconds you ask for, returned as images, with no export and no credits. A motion graphic shows its real frame at that second; a video clip shows its poster. Read every word you changed before you tell your user it is done.
Capabilities
WritesInteractive
Example Prompts
Click any prompt to copy it.
