Engage Media Suite

Get the latest Reels V3 composition and show it in the chat. Returns lint findings, a 1-hour preview link, and a credit estimate. Returns caption options, suggested songs, and clip sound. Preview is free. Rendering a video (Render now, or when the reel is scheduled) uses credits. Requires content:read. Call this once when the reel is ready for the user to watch. Do not call it between drafts. The host plays preview_url. Do not fetch that URL.

motion_reels_get_composition

Get one Reels V3 project, its renders, and its active schedule. Returns caption options, suggested songs, and clip sound. Requires content:read.

motion_reels_get_project

Read one Reels V3 render. Returns status and, once ready, a 1-hour download link for our copy of the file. Requires content:read.

motion_reels_get_render

Get one workspace the authenticated user can access. brand_id must come from list_brands. Returns brand voice, values, audiences, claims to avoid, and each channel writer brief: role, voice notes, brief, do list, don't list, preferred formats, and the assigned audience voice adjustment. Each channel includes api_publish available or unavailable with a short reason and no tokens. brand_setup_status is not_started for a new account with no brand context yet; that is normal, write in a neutral voice. Call this before create_post or update_post. Requires brands:read. Does not modify data.

get_brand

Read up to 100 authorized media files in one call. brand_id must come from list_brands. media_ids must come from list_media or search_media. Send every file you might use in this one call. Do not call get_media once per file. Returns each file summary plus, for a video, recommended highlight windows, timeline segments, transcript cues, and an audio description of the native track. Each cue says whether it ends a sentence and how long the silence after it is. On a transcribed video, highlight and segment edges sit in pauses, not inside words. To cut to what is said, call get_transcript. Each segment and highlight includes framing: where the main subject sits (left, center, or right; top, center, or bottom) and how large it is (tight, medium, or wide), so a portrait clip can be cropped into landscape or the reverse. A photo returns that framing once, on the file. start_sec and end_sec are seconds on that source file. Prefer a highlight when choosing a Reel trim. Use a segment when no highlight fits. Photos return empty windows. A legacy video with no segments returns a capped shot log. Ids that are not on the brand are listed in missing_media_ids. Does not return file bytes, storage paths, or Mux ids. Requires media:read. Does not modify data and does not record an embedding charge.

get_media

Get the full creative document for one reel, carousel, or quick post. brand_id must come from list_brands. kind is reel, carousel, or quick_post. post_id must come from get_project. Returns caption, assigned media (media_id plus in/out or still duration), overlay text, placement, and when each overlay is on screen. Reel overlays include start_sec and end_sec. Carousel text stays up for the whole slide. Quick posts have media and a caption, and overlays is empty. Does not return storage paths, Mux ids, or download URLs. Requires content:read. Does not create, edit, or render the post.

get_post

Get one reel, carousel, or quick-post project the authenticated user can access. brand_id must come from list_brands, and project_id must come from list_projects. Returns project settings, selected album ids, ai_in_progress, and a summary of each post. ai_in_progress is the Projects chip, separate from render status. Does not return overlay text or media trims; call get_post for those. Requires content:read. Does not modify data. A project on another brand is denied.

get_project

Read the status of one render by export_id from render_post. brand_id must come from list_brands. Returns kind, status, and carousel mode when the export is a carousel. Status is none, pending, rendering, ready, failed, or cancelled for a reel, and the stored image or video status for a carousel. Does not return a download URL, storage path, or media id. Requires content:read. Does not modify data.

get_render_status

Read the word-level transcript of one authorized video, to cut a reel to what is said. brand_id must come from list_brands and media_id from list_media, search_media, or get_media. Optional start_sec and end_sec limit the range in seconds on the source file. words are [start_sec, end_sec, text], timed to the millisecond. cut_points are the clean places to cut: after_word indexes into words, in_sec is where a clip should start to open on the next word, out_sec is where a clip should end to close on that word, and sentence_end marks the end of a sentence. Use in_sec and out_sec as trims instead of word times, so word endings are not clipped. sounds lists laughter and other sound events. speaker_turns is listed only when more than one person speaks. At most 2500 words return per call; when next_start_sec is set, call again with that start_sec. status no_speech means the video has no voice; unavailable means it has not been transcribed. Requires media:read. Does not modify data or record a charge.

get_transcript

Mint a fresh Reels V3 preview link for the chat player. App-only: the host hides this tool from the model. Requires content:read. Does not render or spend credits.

motion_reels_refresh_preview

Renders a Reels V3 project now. A new render holds the estimated credits (estimated_render_credits from motion_reels_get_composition or motion_reels_lint), charges them only if it succeeds, and releases them if it fails. 60 fps costs twice the 30 fps price. Unchanged HTML, media, aspect, and fps reuse the in-flight or ready render and hold 0 credits. Requires motion_reels:render. Returns held_credits. Does not wait for the file.

motion_reels_render

Enqueue a render for one saved reel or carousel. Call this only when the user asks. A reel takes brand_id, kind reel, and post_id. A carousel also takes mode images or video. Returns export_id and status. Does not wait for the file and does not change ai_in_progress. Requires content:update.

render_post

Schedule a Reels V3 project to post at scheduled_for. Scheduling renders the video and uses credits: if the video changed or there is no ready video, it holds the estimated credits, starts the render now, and returns immediately; the hold is charged only if the render succeeds. A new render needs scheduled_for at least 30 minutes ahead. A ready or in-flight video with the same content is reused and holds 0 credits. Changing only the time, caption, hashtags, Post via API, or channel is free. With auto_post on an Instagram channel, trending_audio defaults to true and attaches the top trending licensed track at post time; send false to post without it. reel_feed defaults to main_feed; trial posts an Instagram Trial Reel. A schedule saved with auto_post replaces a track chosen in the app. Returns held_credits. Requires the motion_reels:render scope from OAuth consent.

motion_reels_schedule

Put one reel or carousel on Due today. channel_id comes from get_brand. scheduled_for is a future instant within a year, not on a minute another post already has. delivery remind shows it for the user to share from the app. delivery api posts through Instagram or Facebook when that channel api_publish is available. If api is unavailable, nothing is saved. Reel feed and trending audio apply only to an Instagram reel. Saving the schedule sets ai_in_progress false. A failed schedule leaves the chip as it was. The response says whether a ready export exists. Optional idempotency_key replays the first result. Requires publishing:write. Does not publish during this call. Quick posts stay out.

schedule_post

Search one authorized brand's media library. brand_id must come from list_brands. Uses the existing semantic media index and records a small embedding usage charge when that index runs. If semantic search is unavailable, falls back to filename, description, and tag text search, which does not add an embedding charge. Returns ranked metadata, not file bytes. limit defaults to 10 and cannot exceed 20. Requires media:read. Does not upload, delete, or publish media.

search_media

Offer up to 3 caption options (each one caption as posted, with any hashtags at the end of it); the user picks one on the reel page. Writing options never re-renders. Song suggestions are suggestions only and are never added to the video. clip_audio native plays the footage audio in the next render; muted keeps it silent. With native, a video tagged muted or data-volume="0" stays silent and data-volume between 0 and 1 sets its level, unless every video is tagged silent. Requires content:update.

motion_reels_set_caption_options

Set the Projects chip for one reel or carousel. ai_in_progress true shows In progress via AI. false shows Ready to share. schedule_post also clears the chip when the schedule is saved. Creative edits and render_post do not. This tool does not change caption, media, overlays, or render. project_id comes from get_post or list_projects. Requires content:update.

set_ai_progress

Remove the active schedule for one Reels V3 project. Releases an open credit hold, writes that release on the hold, leaves the reel as a draft, and does not charge. Repeating the call is safe. Requires motion_reels:render. Does not require publishing:write.

motion_reels_unschedule

Save a new composition version. Reels V3 is 1080p only: 9:16 (1080×1920), 16:9 (1920×1080), or 1:1 (1080×1080), at 30 or 60 fps. 4:5 and 4K are not supported. 60 fps costs twice as much. GSAP is provided as a global. A jsDelivr or cdnjs gsap.min.js script is rewritten to that copy. Do not load other scripts from a CDN. Use Google Fonts family names, or embed a font with @font-face. A Google Fonts <link> counts as an external URL. Desktop-only fonts such as Bodoni 72 or Didot are not available to the renderer; the video swaps them for a close Google font and motion_reels_lint warns. Optional title renames the reel. If the project is scheduled and the video changed, the reply says the next schedule makes a new render. Time, caption, hashtags, Post via API, and channel changes are free. Requires content:update. Does not spend credits or show the chat preview. Call motion_reels_get_composition when the reel is ready to watch. Give each top-level scene wrapper data-start and data-duration (seconds on the reel) matching when your timeline shows it.

motion_reels_update_composition