FFmpeg Compose
Compose builds one video from a timeline you write. You give it tracks, each track holds keyframes, and each keyframe places one piece of media at a start time for a length. It joins and layers; it does not generate.
assembly · model compose · catalog model-catalog@2026-10-06.v5 · pricing customer-pricing@2026-11-05.v15.Endpoint
POST https://api.roasos.com/v1/assembly/composeSend your API key as a bearer token. The key needs the scope api:write. Every paid request also needs an Idempotency-Key header.
The envelope fields quote_only, quote_id, webhook_url and wait_timeout_seconds behave the same on each model endpoint. Refer to Jobs and Quotes and credits.
Good at
Cut a finished ad from parts you already own: a video track for the shots, an audio track for the voice-over, an image track for a logo or an end card. Every time is in milliseconds. Keyframes inside one track must not overlap, because a track plays one source at a time; tracks overlap each other freely, which is what makes a timeline.
A timeline carries at most 8 tracks, 64 keyframes per track, 200 keyframes in total, and 12 distinct media URLs. The keyframe durations and the timeline end must each stay within 180,000 ms — three minutes. A longer timeline is refused with assembly_timeline_too_long. The whole body must be 64 KiB or smaller.
The media of this operation travels inside the timeline, in tracks[].keyframes[].url, and not in a top-level media field. EVERY one of those URLs is checked, not the first one. Each must be an immutable object on https://cdn.roasos.com that your workspace owns, in the role its track type implies: a video track takes video, an audio track takes audio, an image track takes an image. Any other URL is refused, and the refusal names the track id and the keyframe index.
The price is one credit, flat, whatever the timeline holds.
Pricing
The unit is one operation. Pricing key: compose.
| Request | Credits |
|---|---|
per_operation | 1 |
The credits come from the roasOS pricing table. No other field changes the price. To get the exact price of your request, send it with quote_only: true first.
Modes
compose
Provider endpoints: fal-ai/ffmpeg-api/compose.
Request fields
| Field | Type | Required | Default | Rule |
|---|---|---|---|---|
quote_id | string | no | — | UUID |
quote_only | boolean | no | false | — |
tracks | object[] | yes | — | 1–8 items; each { id, type, keyframes } |
wait_timeout_seconds | integer | no | 0 | 0–60 |
webhook_url | string | no | — | https URL; ≤ 4,096 characters |
The schema is strict. A field that is not in this table gets 400 unsupported_option. The field names are the roasOS role names. One role has one name in every family.
Media
The media of this endpoint travels inside the timeline, at tracks[].keyframes[].url, and not in a top-level media field. Every URL there is checked, in the role its track type implies: a video track takes video, an audio track takes audio, an image track takes an image.
Every other role is refused: avatar_id, elements, product_id, product_images, references, reference_images, image_url, end_image_url, scene_id, scene_image, video_urls, video_url, audio_urls, audio_url, voice_id. A refused role gets 400 unsupported_option.
https://cdn.roasos.com that your workspace owns. Any other URL gets 400 reference_media_not_workspace_owned. Refer to Reference media and uploads to learn how to put a file there.Prompt
None. This operation takes no prompt.
Write the timeline, not the intent. Give every track a short stable id so an error can name it. Start every clip at an exact millisecond; rounding a float here is the difference between a cut and a frame of black.
Send quote_only: true first. The quote runs the whole schema and every media check, so you learn about a refusal before you spend an idempotency key.
Not exposed
roasOS pins no provider field for this model.
Result
The result kind is video. Always present: url.
{
"data": {
"id": "gen_…",
"kind": "video",
"model": "compose",
"status": "completed",
"result": {
"url": "https://cdn.roasos.com/workspaces/ws_01/result"
}
}
}The files are immutable objects on https://cdn.roasos.com. A result URL is a valid media input for your next request.
Example
The three steps are always the same: quote, run, poll.
# 1. Quote. Nothing is spent.
curl -X POST https://api.roasos.com/v1/assembly/compose \
-H "Authorization: Bearer roasos_sk_live_..." \
-H "Idempotency-Key: 0f1c9e2a-compose-0001" \
-H "Content-Type: application/json" \
-d '{"quote_only":true,"tracks":[{"id":"example","type":"video","keyframes":[{"timestamp":0,"duration":1,"url":"https://cdn.roasos.com/workspaces/ws_01/uploads/video-1.mp4"}]}]}'
# 200 { "data": { "kind": "quote", "quote_id": "q_7Hk...", "credits": 1, "expires_at": "...",
# "confirmation": { "model": "compose", "example": "per_operation" } } }
# 2. Run. Same body, same Idempotency-Key, quote_id in place of quote_only.
curl -X POST https://api.roasos.com/v1/assembly/compose \
-H "Authorization: Bearer roasos_sk_live_..." \
-H "Idempotency-Key: 0f1c9e2a-compose-0001" \
-H "Content-Type: application/json" \
-d '{"quote_id":"q_7Hk...","tracks":[{"id":"example","type":"video","keyframes":[{"timestamp":0,"duration":1,"url":"https://cdn.roasos.com/workspaces/ws_01/uploads/video-1.mp4"}]}]}'
# 201 { "data": { "id": "gen_...", "status": "queued", "credits_reserved": 1 } }
# 3. Poll, then read the result.
curl https://api.roasos.com/v1/jobs/gen_.../status -H "Authorization: Bearer roasos_sk_live_..."
curl https://api.roasos.com/v1/jobs/gen_.../result -H "Authorization: Bearer roasos_sk_live_..."Today
The endpoint is POST /v1/assembly/compose.
It is live.
See also
llms.txtfor this model — the same facts as one plain-text file for an agent.GET https://api.roasos.com/v1/catalog/compose— the machine-readable form: enums, defaults, limits and prices. Refer to the API reference.- Jobs — how to poll a job and read its result.
- Provider endpoints:
fal-ai/ffmpeg-api/compose.
HeyGen Translate (speed)
HeyGen Translate re-voices a finished video and moves the speaker's mouth to match the new language. The speed tier is the faster and cheaper of the two HeyGen lanes.
FFmpeg Extract Frame
Extract Frame takes one still out of a finished clip: the first frame, the middle frame, or the last one.