Rivya AI Docs

Wan 2.7 Video API Integration

Use Wan 2.7 Video through Rivya Public API v1 with model id wan-2-7-video, supported inputs, parameters, Files API rules, credits, and response examples.

Available via APIText or URL ready; reference modes use Files APIVideo
API model id

wan-2-7-video

Inputs

text, file

Files API

Required for reference modes

Base credits

80

Capability

Video

Billing

FIXED

Prompt limit

5000 characters

Request contract

Send the model id at the top level. Model-specific controls belong in params.

KeyTypeRequiredDefaultDescription
modelstringYeswan-2-7-videoUse the wan-2-7-video API model id.
promptstringYes-Maximum 5000 characters for this model.
paramsobjectNo-Model-specific parameter object. Use the rows below for allowed keys.
client_request_idstringNo-Optional client-side id for tracing requests in your system.

Model parameters

KeyTypeRequiredDefaultRangeOptionsDescription
resolutionselectNo720p-720p, 1080pResolution
durationselectNo5-5, 10, 15Text and image runs can use 5, 10, or 15 seconds. Video editing currently tops out at 10 seconds, and Rivya normalizes longer selections automatically in that mode.
aspect_ratioselectNo16:9-16:9, 9:16, 1:1, 4:3, 3:4Text-to-video and video-editing runs use this directly. Image-to-video currently follows the uploaded frame geometry instead of exposing a separate aspect-ratio switch.
negative_prompttextNo---Negative prompt
prompt_extendselectNotrue-true, falseControls whether the model rewrites your prompt before generation. Keep it on for convenience, or turn it off when you already have a very precise prompt.
watermarkselectNofalse-false, trueWatermark
audio_settingselectNoauto-auto, originControls how audio is handled in the video-edit path. Auto lets the model decide, while Keep original tries to preserve the source video's audio.
seednumberNo-min 0 / max 2147483647-Leave it empty for a new random result each time. Reusing the same integer makes it easier to reproduce or fine-tune a previous result. It stays closer only when the prompt and other key settings also remain similar, and it is not a guaranteed 1:1 match.

Upload policy

Upload reference files first, then place the returned URL and duration token inside params.referenceMediaItems.

Reference rule

Reference media: Upload up to 2 assets. Use up to 2 images for image-to-video, or 1 video plus 1 optional image for video editing.

Max files

2

Accepted file kinds

image, video

Duration token

Video and audio references should carry durationToken from /api/v1/files when duration verification is required.

kindMax sizeMIME types
image20 MBimage/jpeg, image/png, image/webp
video95 MBvideo/mp4, video/quicktime

Minimal request

{
  "model": "wan-2-7-video",
  "prompt": "A cinematic product reveal with smooth camera movement",
  "params": {
    "resolution": "720p",
    "duration": "5",
    "aspect_ratio": "16:9",
    "negative_prompt": "Optional exclusions or failure modes to avoid.",
    "prompt_extend": "true",
    "watermark": "false",
    "audio_setting": "auto",
    "seed": 0
  }
}

Reference-file request

{
  "model": "wan-2-7-video",
  "prompt": "A cinematic product reveal with smooth camera movement",
  "params": {
    "resolution": "720p",
    "duration": "5",
    "aspect_ratio": "16:9",
    "negative_prompt": "Optional exclusions or failure modes to avoid.",
    "prompt_extend": "true",
    "watermark": "false",
    "audio_setting": "auto",
    "seed": 0,
    "referenceMediaItems": [
      {
        "url": "https://cdn.example.com/reference-image.png",
        "kind": "image",
        "name": "reference-image.png",
        "mimeType": "image/png"
      }
    ]
  }
}

Create response

The create endpoint returns a public task id. Poll the status endpoint until the task succeeds or fails.

{
  "id": "task_wan_2_7_video_example",
  "status": "queued",
  "model": "wan-2-7-video",
  "reserved_credits": 80,
  "final_credits": 0,
  "created_at": "2026-05-11T00:00:00.000Z",
  "updated_at": "2026-05-11T00:00:00.000Z",
  "result": null,
  "error": null
}

Common errors

validation_failed, insufficient_credits, idempotency_conflict, rate_limited, not_found

Some modes require Files API reference uploads.