Docs

Running a model

MiniMaxText to video

MiniMax H3

Multimodal video generation with native synced audio, multi-reference consistency, and continuation workflows

POSTauray-serverless/minimax/h3-0
openapi.jsonllms.txt

One of Auray's catalogue models: the same key, the same credits, the same job lifecycle as every other address.

Its request shape is published below. Whether it runs right now is GET /v1/catalogue's word, with no key needed: its addresses say connected for an address that runs, and a submit to any other is refused before anything is charged — described answers 501 model_not_connected, paused 503 model_disabled.

Category text-to-video · Scope video:write (or external:write on an older key) · Price Credits per megapixel-second (counted as width*height*floor:1.0322*duration): 9.1392. width*height counts as at least 1.0322 megapixels. From 6.1932 to 113.0496 megapixel-seconds a request. Priced up to height 2560, width 2944; past that a request is refused before any charge. Plus 4.716667 credits per input image (counted as items:inputs.referenceImages+inputs.frameImages, the first 5 free), up to 11 input images. Plus 229.9375 credits when the request sends inputs.referenceVideos (counted as any:inputs.referenceVideos). Plus 5.895834 credits per second of output when resolution=1440p (counted as any:resolution*duration), up to 15 seconds of output. Plus 25.267891 credits per second of output when height=1920, width=1440 (counted as any:width*duration), up to 15 seconds of output. Plus 33.690561 credits per second of output when height=2560, width=1440 (counted as any:width*duration), up to 15 seconds of output. Plus 33.578304 credits per second of output when height=1248, width=2944 (counted as any:width*duration), up to 15 seconds of output. Quoted when you submit, at the most the request can cost; the charge settles to what the job cost and never above the quote. The submit's answer carries the quote as credits_quoted, and credits_charged says what has been taken.

#What it can do

  • No operation is listed for this address yet; its fields say what a request may send.
  • Length: 4 to 15 seconds
  • Aspect ratios: 1:1, 16:9, 21:9, 3:4, 4:3, 9:16
  • Resolutions: 768p, 1440p
  • Prompt: up to 7,000 characters
  • Sound: makes its own sound
  • Priced by: megapixel_second (the whole rule is under Price)

The same model also answers at auray-ai/minimax-h3/text-to-video, with its own capabilities and its own price — its page says them.

Find the addresses that do any of these with GET /v1/models?operation=<id>; the openapi.json carries them as data under x-auray-capability.

#Input

FieldTypeRequiredDefaultDescription
inputsobjectno—
inputs.referenceImagesstring[]no—List of reference images (an http(s) URL). Documented range: 1 to 9. When the request also sends inputs.referenceAudios, this model makes the person in a picture here speak or move: a picture you upload, or a link from anywhere else, is taken once your account has accepted, one time, that you have the right to the faces and voices you use — send acceptance: {"real_people": "2026-11-03"} with the request, or accept in the Auray app — and is refused 403 acceptance_required until then, before anything is charged. A picture made here — a job's result — asks nothing. While uploads are paused here, a picture you upload, or a link from anywhere else, is refused 403 likeness_not_made_here instead, before anything is charged. The files the examples here send are ours, and are taken.
inputs.frameImagesstring[]no—Images that pin frames of the video, as a list of image URLs: one image is the first frame, two are the first and the last, and three or more anchor the first and the last with the rest spread evenly between. Each item is a plain URL; the object form that names a frame or a timestamp is not accepted here. Documented range: 1 to 2. When the request also sends inputs.referenceAudios, this model makes the person in a picture here speak or move: a picture you upload, or a link from anywhere else, is taken once your account has accepted, one time, that you have the right to the faces and voices you use — send acceptance: {"real_people": "2026-11-03"} with the request, or accept in the Auray app — and is refused 403 acceptance_required until then, before anything is charged. A picture made here — a job's result — asks nothing. While uploads are paused here, a picture you upload, or a link from anywhere else, is refused 403 likeness_not_made_here instead, before anything is charged. The files the examples here send are ours, and are taken.
inputs.referenceVideosstring[]no—List of reference videos (an http(s) URL). Documented range: 1 to 3. This model may keep the sound of a clip here as a voice: a clip you upload with a sound track, or a link from anywhere else, is taken once your account has accepted, one time, that you have the right to the faces and voices you use — send acceptance: {"real_people": "2026-11-03"} with the request, or accept in the Auray app — and is refused 403 acceptance_required until then, before anything is charged. A clip you upload with no sound track asks nothing, nor does one made here. When the request also sends inputs.referenceAudios, this model makes the person in a clip here speak or move: a clip you upload, or a link from anywhere else, is taken once your account has accepted, one time, that you have the right to the faces and voices you use — send acceptance: {"real_people": "2026-11-03"} with the request, or accept in the Auray app — and is refused 403 acceptance_required until then, before anything is charged. A clip made here — a job's result — asks nothing. While uploads are paused here, a clip you upload with a sound track, or a link from anywhere else, is refused 403 voice_sample_not_made_here instead, before anything is charged. While uploads are paused here, a clip you upload, or a link from anywhere else, is refused 403 likeness_not_made_here instead, before anything is charged. The files the examples here send are ours, and are taken.
inputs.referenceAudiosstring[]no—List of reference audios (an http(s) URL). Documented range: 1 to 3. This model may keep the voice in a recording here: a recording you upload, or a link from anywhere else, is taken once your account has accepted, one time, that you have the right to the faces and voices you use — send acceptance: {"real_people": "2026-11-03"} with the request, or accept in the Auray app — and is refused 403 acceptance_required until then, before anything is charged. Or send the request without it. While uploads are paused here, a recording you upload, or a link from anywhere else, is refused 403 voice_sample_not_made_here instead, before anything is charged; send the request without it. The files the examples here send are ours, and are taken.
positivePromptstringyes—Text prompt describing elements to include in the generated output. Documented range: 2 to 7,000 characters.
widthintegerno2560Width of the generated media in pixels.
heightintegerno1440Height of the generated media in pixels.
resolutionstringno—Resolution preset for the output. When used with input media, automatically matches the aspect ratio from the input. One of "768p", "1440p".
durationintegerno5Length of the generated video in seconds. The total number of frames produced is determined by duration multiplied by the model's frame rate (fps). Documented range: 4 to 15.
seedintegerno—Random seed for reproducible generation. When not provided, a random seed is generated in the unsigned 32-bit range. Documented range: 0 to 4,294,967,295.
settingsobjectno—
settings.promptExpansionstringno"balanced"Level of automatic prompt rewriting applied before generation. Affects reproducibility. One of "disabled", "balanced", "quality".
outputTypestringno"URL"Video output type. One of "URL".
outputFormatstringno"MP4"Specifies the file format of the generated output. The available values depend on the task type and the specific model's capabilities. - `JPG`: Best for photorealistic images with smaller file sizes (no transparency). - `PNG`: Lossless compression, supports high quality and transparency (alpha channel). - `WEBP`: Modern format providing superior compression and transparency support. - `MP4`: Widely supported video container (H.264), recommended for general use. - `WEBM`: Optimized for web delivery. - `MOV`: QuickTime format, common in professional workflows (Apple ecosystem). - `GIF`: Animated image format (no audio), suitable for short loops or previews. - `MP3`: Compressed audio, smaller file size. - `WAV`: Uncompressed, high-quality audio. - `FLAC`: Lossless compression. - `OGG`: Open-source compressed audio format (Vorbis codec). - `SVG`: Scalable Vector Graphics. - `TIFF`: High-quality output supporting layers. - `GLB`: Binary glTF format for 3D models, including geometry, textures, and PBR materials. > [!NOTE] > **Transparency**: If you are using features like background removal or LayerDiffuse that require transparency, you must select a format that supports an alpha channel (e.g., `PNG`, `WEBP`, `TIFF`). `JPG` does not support transparency. One of "MP4", "WEBM", "MOV".
outputQualityintegerno95Compression quality of the output. Higher values preserve quality but increase file size. Documented range: 20 to 99.
safetyobjectno—
safety.checkContentbooleanno—Enable or disable content safety checking. Increases total generation time. Always on: a false is sent as true.
safety.modestringno"fast"Safety checking mode for video generation. Only fast or full is sent here: any other value is sent as fast. One of "fast", "full".
idempotency_keystringno—Your own name for this request. Send the same one twice and the second call is answered with the first request rather than charged again. Idempotency-Key: as a header is accepted as an alias — that is the spelling a retrying proxy replays, and the one to use when running a model on auray.run, where a request that outlives its connection is answered 504 and recovered by replay. Omit it and the platform mints one. 1 to 64 characters.
acceptanceobjectno—Your account's one acceptance, sent with the request that first needs it: a file of a real person's face or voice you upload, in a field this model makes speak or move, copies the voice of or puts into a clip, is taken once your account has accepted, one time, that you have the right to the faces and voices you use. Kept for your account before anything is charged, so no later request asks again — in any app or through the API — and asked of no file made here. You can accept in the Auray app (Account → Privacy) instead; an assistant cannot accept for you, so through MCP it is accepted in the app.
acceptance.real_peoplestringno—The version of the words you accept: that you have the right to every face and voice you upload, and to any voice or style you pick that sounds like a real person — it is yours, or you have permission — and that you are responsible for what you make with them. One of "2026-11-03".

#Rules

Across fields, as the model's documentation states them. Each is judged before anything is charged — warned with a Deprecation header until enforcement starts — and listed as JSON Schema under x-auray-rules in the openapi.json.

  • Send height whenever width is sent. Send width whenever height is sent.
  • Do not send a request where inputs.frameImages is sent and inputs.referenceImages is sent, inputs.referenceVideos is sent or inputs.referenceAudios is sent.
  • Do not send resolution together with width or height.
  • When inputs.frameImages is sent, do not send a request where width is sent, or where height is sent.
  • When resolution is sent, send at least one of inputs.frameImages, inputs.referenceImages or inputs.referenceVideos.
  • When width is sent or height is sent, width and height must be one of these: 768p (~16:9): width 1344, height 768; 768p (4:3): width 1024, height 768; 768p (1:1): width 768, height 768; 768p (3:4): width 768, height 1024; 768p (~9:16): width 768, height 1344; 768p (~21:9): width 1536, height 672; 1440p (16:9): width 2560, height 1440; 1440p (4:3): width 1920, height 1440; 1440p (1:1): width 1440, height 1440; 1440p (3:4): width 1440, height 1920; 1440p (9:16): width 1440, height 2560; 1440p (~21:9): width 2944, height 1248.
  • When inputs.referenceAudios is sent, send at least one of inputs.referenceImages or inputs.referenceVideos.

#Example

curl -X POST https://queue.auray.run/auray-serverless/minimax/h3-0 \
  -H "Authorization: Bearer $AURAY_API_KEY" \
  -H "content-type: application/json" \
  -d '{
  "positivePrompt": "a brass orrery on a walnut desk, morning light from a window to the left"
}'

The same body on https://auray.run runs synchronously and answers the result — unless it usually takes longer than a run may hold its connection, which is refused 400 model_not_synchronous before anything is charged. Options — ?webhook=, ?wait=, priority and timeout headers — are on Submit a request.

#Machine contracts

  • openapi.json — this page as an OpenAPI 3.1 document, public.
  • llms.txt — the same, written for an agent.

Start typing to search every page of the documentation.

Up arrow Down arrow to moveEnter to openEscape to close