MiniMax H3 Recast
MiniMax H3 Recast replaces the main people in a source video with new people from reference photos. The output keeps the motion, the gestures, the camera, the shot timing and the soundtrack of the source. The output length is the source length.
video-edit · model h3-max-recast · catalog model-catalog@2026-10-06.v5 · pricing customer-pricing@2026-11-05.v15.Endpoint
POST https://api.roasos.com/v1/video-edit/h3-max-recastSend 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
Put your own cast into an ad that works. Send video_url (5 to 30 seconds, MP4, no single shot longer than 15 seconds) and 1 to 4 new people. Send one photo for each person in references. An avatar_id adds one person: the front portrait of the saved avatar. The new people replace the main people of the source from left to right, in this order: your references, then the avatar. Each new person is in every shot where the original person is.
Pricing
The unit is one source_second. Pricing key: h3-max-recast.
| Request | Credits |
|---|---|
5s_source_768p | 18 |
5s_source_1080p | 27 |
10s_source_768p | 36 |
10s_source_1080p | 54 |
30s_source_768p | 108 |
30s_source_1080p | 162 |
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
edit
Provider endpoints: minimax/h3-max/recast.
Request fields
| Field | Type | Required | Default | Rule |
|---|---|---|---|---|
avatar_id | string | no | — | UUID |
enhance_realism | boolean | no | false | — |
prompt | string | no | — | ≤ 10,000 characters |
quote_id | string | no | — | UUID |
quote_only | boolean | no | false | — |
references | string[] | no | — | ≤ 4 items |
resolution | enum | no | 1080p | 768p, 1080p |
seed | integer | no | — | 0–2147483647 |
video_url | string | yes | — | https URL; ≤ 4,096 characters |
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
This endpoint accepts these media roles:
avatar_id— identity; the saved character reference sheet, with a derived portrait only where the model requires a single face or source framereferences— style and appearance referencesvideo_url— the source video
| Role | Items | Maximum size | Content types | Duration |
|---|---|---|---|---|
references | ≤ 4 | 10 MB | image/jpeg, image/png, image/webp | — |
video_url | ≤ 1 | 100 MB | video/mp4 | 5–30 s |
Every other role is refused: elements, product_id, product_images, reference_images, image_url, end_image_url, scene_id, scene_image, video_urls, 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
Limit 10,000 characters. The prompt is optional.
- Use it to say who becomes whom: "The woman on the left becomes the person in the second photo."
- Use it to say what else must stay or change: "Keep her red jacket."
- Do not ask for a new shot, a new camera move or a different length.
Not exposed
These catalog entries describe omitted inputs, pinned values and source constraints. They do not imply that the provider accepts every listed field.
enable_safety_checker— pinned toprovider default (on); never sent.duration_seconds— pinned tothe output follows the source clip.aspect_ratio— pinned tothe output keeps the source framing.generate_audio— pinned tothe output keeps the source sound.negative_prompt— not accepted by this public endpoint.
Result
The result kind is video. Always present: url.
{
"data": {
"id": "gen_…",
"kind": "video",
"model": "h3-max-recast",
"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/video-edit/h3-max-recast \
-H "Authorization: Bearer roasos_sk_live_..." \
-H "Idempotency-Key: 0f1c9e2a-h3-max-recast-0001" \
-H "Content-Type: application/json" \
-d '{"quote_only":true,"video_url":"https://cdn.roasos.com/workspaces/ws_01/uploads/video_url.mp4"}'
# 200 { "data": { "kind": "quote", "quote_id": "q_7Hk...", "credits": 18, "expires_at": "...",
# "confirmation": { "model": "h3-max-recast", "example": "5s_source_768p" } } }
# 2. Run. Same body, same Idempotency-Key, quote_id in place of quote_only.
curl -X POST https://api.roasos.com/v1/video-edit/h3-max-recast \
-H "Authorization: Bearer roasos_sk_live_..." \
-H "Idempotency-Key: 0f1c9e2a-h3-max-recast-0001" \
-H "Content-Type: application/json" \
-d '{"quote_id":"q_7Hk...","video_url":"https://cdn.roasos.com/workspaces/ws_01/uploads/video_url.mp4"}'
# 201 { "data": { "id": "gen_...", "status": "queued", "credits_reserved": 18 } }
# 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
POST /v1/video-edit/h3-max-recast with video_url, references, avatar_id, prompt, resolution (768p|1080p, default 1080p), enhance_realism (more natural skin and better blending, off by default) and seed.
See also
llms.txtfor this model — the same facts as one plain-text file for an agent.GET https://api.roasos.com/v1/catalog/h3-max-recast— 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:
minimax/h3-max/recast.
Gemini Omni Flash Edit
Gemini Omni Flash Edit applies a requested visual change to an owned video. It accepts a prompt and one source clip; roasOS currently uses the provider's default 720p edit tier.
Kling 4.0 Edit
Kling 4.0 Edit changes one source video with a text instruction and optional image references and identity elements. It keeps the motion, the cuts and the timing of the source, and it renders at 720p or 1080p. The output is 3 to 30 seconds long.