Videos

Create video generation tasks with OpenAI-compatible authentication and a task-style response shape. Video generation is async: create a task, store the id, and poll for completion.

Video generation is key-enabled. Only keys with video access list video models from /models; keys without video access cannot create video tasks.

When to use this page

Use this page when you need application-level video generation with production-friendly status polling. For model-specific guidance, see Veo 3.1 Fast and Seedance 2.0 Fast.

Endpoints

POST https://api.quotaflow.ai/openai/v1/videos
GET https://api.quotaflow.ai/openai/v1/videos/{task_id}

Supported models

• veo-3.1-fast-generate-001
• veo-3.0-generate-001
• seedance-2.0-fast

Call /models with your production key for the exact list enabled for your account.

Key features

• Async task lifecycle: create, poll, render pending UI, then display or save the result.
• Text-to-video prompts: describe subject, motion, camera, lighting, duration, and style.
• Separate video families: use veo-* ids for Veo-compatible workflows and seedance-2.0-fast for Seedance 2.0 workflows; Quotaflow keeps the endpoint unified while preserving the requested public model.
• Optional image inputs: use image_url, image_urls, images, or reference_images for first-frame or reference-image workflows when supported by your enabled model.
• Reference-image readiness: for person, product, or style consistency, send stable HTTPS image URLs; Quotaflow prepares reference media automatically when required.
• Stable public response shape: Quotaflow preserves the public task id, status, model, data URL, and usage fields even when capacity is served by compatible video backends.
• Retry-safe product design: store task ids and make polling idempotent in your application.

Create a video task

curl https://api.quotaflow.ai/openai/v1/videos \
  -H "Authorization: Bearer $QUOTAFLOW_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "veo-3.1-fast-generate-001",
    "prompt": "A cinematic drone shot over a coastal city at sunrise.",
    "durationSeconds": 4,
    "resolution": "1080p",
    "aspectRatio": "16:9"
  }'

Reference-image video

For image-to-video or identity/reference workflows, prefer image_urls or reference_images arrays. A single legacy image_url is still accepted and normalized, but arrays make the intent explicit and avoid client-side ambiguity.

curl https://api.quotaflow.ai/openai/v1/videos \
  -H "Authorization: Bearer $QUOTAFLOW_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "veo-3.1-fast-generate-001",
    "prompt": "Animate this portrait into a natural 4 second speaking-style social clip, preserve the face identity and lighting.",
    "image_urls": ["https://example.com/reference-face.jpg"],
    "durationSeconds": 4,
    "resolution": "720p",
    "aspectRatio": "9:16",
    "watermark": false,
    "return_last_frame": true
  }'

reference_images may contain strings or objects such as { "url": "https://example.com/reference-face.jpg" }. Keep URLs reachable while the task is running. For heavy reference-image, video-reference, audio-reference, 1080p, 4K, or longer-duration jobs, use an async UI and longer polling windows.

Seedance 2.0 unified call

Seedance 2.0 uses the same public /openai/v1/videos create-and-poll contract. Use seedance-2.0-fast when you specifically want Seedance behavior instead of a Veo-compatible model. Quotaflow adapts reference media internally, so your application should not call provider asset endpoints directly.

For 4K Seedance output, keep using the public seedance-2.0-fast id and set "resolution": "4k". Quotaflow maps that request to the appropriate 4K-capable Seedance 2.0 call path internally; there is no separate public seedance-2.0-4k id.

curl https://api.quotaflow.ai/openai/v1/videos \
  -H "Authorization: Bearer $QUOTAFLOW_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "seedance-2.0-fast",
    "prompt": "Animate this portrait into a natural 4 second social clip. Preserve identity, lighting, and framing.",
    "reference_images": [{ "url": "https://example.com/reference-face.jpg" }],
    "durationSeconds": 4,
    "resolution": "4k",
    "aspectRatio": "9:16",
    "generate_audio": true,
    "watermark": false,
    "return_last_frame": true
  }'

Portrait, person-reference, or audio-reference jobs can be slower and may fail model safety review. Keep reference URLs reachable until the task completes, then poll the returned task id.

Fetch task status or result

curl https://api.quotaflow.ai/openai/v1/videos/$TASK_ID \
  -H "Authorization: Bearer $QUOTAFLOW_API_KEY"

Parameters

Response shape

The create call returns a task object or task id. Poll GET /videos/{task_id} until the task is completed or failed. Use a pending state in your UI; do not block an end-user request waiting for the render to finish.