Seedance 2.5

Returns a task_id immediately. Poll /api/v1/status for the result. Credits are deducted upfront and auto-refunded if the generation fails. Requires a paid plan.

Parameters

curl -X POST https://www.lunostudio.ai/api/v1/generate \
  -H "Authorization: Bearer $LUNO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "seedance-2.5",
    "prompt": "A lighthouse in a storm, waves breaking over the rocks",
    "duration": 10,
    "resolution": "720p",
    "aspect_ratio": "16:9"
  }'

{
  "task_id": "task_lb3mygsolx1zh14ylx4yve7r",
  "credit_cost": 340,
  "model": "bytedance/seedance-2.5",
  "status": "pending"
}

Billed per second, by resolution. Attaching a reference video changes both the rate and the number of seconds counted.

Billable seconds

A billable second is a second the model processes, which is not always a second of the video you get back.

With no reference video, the two are the same. A 10-second request is 10 billable seconds.

With one or more reference videos, the model reads those clips in addition to producing yours, and every second it reads is billable:

Each length is rounded up to the next whole second and counted separately. The whole reference clip counts, not the portion that appears in the result — a 60-second reference contributes 60 seconds whether the model draws on all of it or one frame. Reference images and reference audio contribute nothing.

The lower per-second rate offsets part of the extra seconds. A reference shorter than roughly half the output length costs less than the same clip would at the no-reference rate; beyond that it costs more.

Without a reference video

Billed on output duration only. This covers text-to-video, image-to-video, keyframe pairs, and audio references.

With one or more reference videos

A lower per-second rate, applied to billable seconds = output duration + the full duration of every reference video, each rounded up to the next whole second.

Every response includes credit_cost, so you always know the exact charge before you poll. Failed generations are refunded in full. Reference-video durations are measured at submit time; a clip whose duration cannot be read is counted as the output duration.

Attach up to 20 images, 6 videos, and 6 audio clips. Reference them positionally in the prompt with @image1, @video1, @audio1, numbered in the order you supply them.

Keyframes vs references

With reference images and no @ tags, the first image is pinned as the start frame. Add last_frame to pin the end frame as well, and any remaining images are treated as references. Set reference_mode: true to skip pinning entirely and treat every image as a style or subject reference.

curl -X POST https://www.lunostudio.ai/api/v1/generate \
  -H "Authorization: Bearer $LUNO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "seedance-2.5",
    "prompt": "@image1 walks toward the camera, matching the motion in @video1",
    "reference_images": ["https://example.com/character.png"],
    "reference_videos": ["https://example.com/walk-cycle.mp4"],
    "duration": 8,
    "resolution": "720p"
  }'

Poll with the task_id returned by the generate call. Generations typically complete in 1-5 minutes depending on duration and resolution.

curl "https://www.lunostudio.ai/api/v1/status?task_id=task_lb3mygsolx1zh14ylx4yve7r" \
  -H "Authorization: Bearer $LUNO_API_KEY"

{
  "task_id": "task_lb3mygsolx1zh14ylx4yve7r",
  "status": "success",
  "output": ["https://lunostudio.ai/storage/v1/object/public/generations/..."]
}

A separate preview build of Seedance 2.5 remains available as seedance-2.5-preview. It takes the same request schema and is billed at a flat per-second rate, with no reference-video surcharge. See Seedance 2.5 Preview.