Video Generation API

Video generation is an asynchronous task: POST /videos/generations returns a task ID immediately, then poll GET /videos/generations/{id} until status becomes completed, then read video_url to download the result.

Supported models: seedance-1.5-pro, seedance-2.0, etc. Use GET /videos/models for the live list. Full schema and constraints in the interactive API explorer.

• GET https://api.alltoken.ai/v1/videos/models — list available video models
• POST https://api.alltoken.ai/v1/videos/generations — create a generation task
• GET https://api.alltoken.ai/v1/videos/generations/{id} — poll task status
• GET https://api.alltoken.ai/v1/videos/generations — list tasks
• POST https://api.alltoken.ai/v1/videos/generations/{id}/cancel — cancel a task

Returns all available video generation models in OpenAI-compatible format.

Text-to-video:

Image-to-video (first-frame guidance):

Response (task created, status queued):

• model (required) — Video model ID (e.g. "seedance-1.5-pro")
• prompt (required) — Text description of the desired video
• content — Multimodal input array (image-to-video or video reference). Each item has type (image_url / video_url / audio_url / draft_task) and role (first_frame / last_frame / reference_image / reference_video / reference_audio)
• ratio — Aspect ratio: 16:9 / 9:16 / 4:3 / 3:4 / 21:9 / 1:1 / adaptive
• duration — Duration in seconds; -1 for model-decided
• resolution — 480p / 720p / 1080p
• frames - Frame count (takes precedence over duration)
• generate_audio - Generate matching audio track
• input_has_video - Set when the input content includes a video reference
• seed - Random seed for reproducibility
• camera_fixed — Lock the camera
• watermark — Add watermark (default false)
• callback_url — URL called when the task completes
• return_last_frame — Return last-frame URL (for continuation)
• service_tier — default / flex
• execution_expires_after — Execution expiry window in seconds (3600-259200)
• draft — Create or use a draft task when supported
• tools — Optional tool configuration, such as web_search
• safety_identifier — Optional stable safety identifier, up to 64 characters

Returns task details. When status is completed, video_url is the download URL.

video_url is a presigned URL that expires after video_url_ttl seconds. Re-call this endpoint to refresh the URL.

• queued — Enqueued, waiting for GPU
• processing — Model is generating
• completed — Done, video_url available
• failed — Generation failed; details in error
• expired - Task or generated asset expired
• cancelled — Cancelled by the user

Recommended polling interval: 2-5 seconds. Most tasks complete in 30-120 s; high-resolution jobs may take 3-5 minutes.

Paginated list of the current user's video generation tasks. Query parameters:

• limit — Page size (default 20, max 100)
• status — Filter by status (queued / processing / completed / failed / expired / cancelled)

Cancels a task that has not yet completed (queued or processing only). Already completed / cancelled tasks return 409 Conflict.

• 400 — Bad request (e.g. empty prompt, invalid ratio)
• 401 — Invalid or missing API key
• 402 — Insufficient balance
• 404 — Task ID not found
• 409 — Task state does not allow this action (e.g. cancel a completed task)
• 429 — Rate limit exceeded