Seedance 2.5 Flash

Seedance 2.5 Flash supports text-to-video, first/last-frame generation, reference generation and video editing. Use dreamina-seedance-2-5-flash with the Default tier. Send media as public HTTP or HTTPS URLs without credentials. Save the returned task ID and query it for the finished video.

See the current rates, OpenAPI specification.

Quick example

This request creates a four-second video. Review the current price before sending it; submitting a successful generation incurs a charge.

curl https://api.toklab.ai/v1/videos/generations \
  -H "Authorization: Bearer $TOKLAB_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "dreamina-seedance-2-5-flash",
    "content": [{"type": "text", "text": "A ceramic cup slowly rotates under warm studio lighting."}],
    "resolution": "480p",
    "duration": 4,
    "ratio": "16:9",
    "generate_audio": false
  }'

Parameters

model and content are required. All other fields are optional. Explicit null values and unknown fields are rejected.

Prompts are documented with a maximum of 10,000 tokens. No public tokenizer is specified, so a character count must not be treated as an exact token count.

Content and modes

Text uses {"type":"text","text":"Describe the shot"}. Images, videos and audio use a nested URL and an explicit role:

{
  "type": "image_url",
  "image_url": {"url": "https://example.com/reference.jpg"},
  "role": "reference_image"
}

Video items use type: "video_url", video_url: {"url": "..."}, role: "reference_video"; audio items use audio_url and reference_audio in the same shape. First/last frames cannot be combined with reference media. Draft, Extend, 1080p, watermark, seed and last-frame output controls are not available for this integration. Callbacks are not offered.

Media constraints

Files must be readable before submission. Video duration is measured from the media to calculate the charge. Use stable public URLs.

Task response and polling

Creation returns an id, model, status and created_at. Query the same task until it completes:

curl "https://api.toklab.ai/v1/tasks/$TASK_ID" \
  -H "Authorization: Bearer $TOKLAB_API_KEY"

The status is processing, completed or failed. On success, read output.video_url for the stored video. On failure, read error.code and error.message. Public responses do not include costs. A polling timeout does not prove that generation failed: keep the task ID and query it again instead of submitting another generation.

Billing

The live rate depends on resolution and whether a video is supplied. Without video input, billed seconds equal output duration. With reference videos, billed seconds equal output duration plus the greater of output duration and the combined input duration rounded up. Editing bills twice the source duration rounded up. Audio is included. Failed requests are not charged.

For example, a five-second output with three seconds of reference video bills ten seconds. A ten-second source edit bills twenty seconds. Multiply billed seconds by the applicable live rate.

See task lifecycle and error codes for common responses and retry guidance.