Create video

Generate with bytedance/seedance-2.5 in the Volcengine Ark native shape: prompt and material live in one content[] array. If you already have Ark code, point base_url at this site.

Three key request-body conventions#

1. The prompt is optional#

content may carry no text item at all — a first frame alone, first-and-last frames, even reference audio on its own, are all combinations the vendor supports. The only requirement is a non-empty array. This is the opposite of the MiniMax shape, where the prompt is always required.

2. type is not the role, and the URL is not flat#

{
  "content": [
    { "type": "text", "text": "the person walks toward the camera" },
    { "type": "image_url", "role": "first_frame", "image_url": { "url": "https://…/a.png" } }
  ]
}

role may be omitted in exactly three cases, matching the vendor: a single image as the only material is treated as the first frame (first_frame); a video_url as reference_video; an audio_url as reference_audio. First/last frames and reference images need an explicit role — an image without one alongside other material returns a 400 rather than being guessed.

3. The three modes are mutually exclusive#

First frame, first-and-last frame and omni reference are three separate modes, and mixing them is rejected upstream. Coming from MiniMax this is the easiest one to hit, since that shape lets you combine frames and reference material.

Value ranges#

Required: with a first frame, ratio accepts only adaptive — the output ratio always follows that image, and a concrete ratio is rejected at submit time.

Seedance 1.x parameters are rejected#

seed, camera_fixed, frames, draft and service_tier=flex are valid on 1.x and no Seedance 2.x model accepts them. Sending one is rejected at submit time rather than dropped silently — dropping it silently means you think seed applied and get a random result back.

Omni-reference subtasks (bytedance/seedance-2.5 only)#

A request qualifies as an omni-reference task when content includes any of the reference_image, reference_video, or reference_audio roles. The model further classifies it into one of three subtasks based on prompt intent, and each subtask enforces different constraints on ratio and duration:

For editing and extension, the output aspect ratio always follows the selected reference video; for editing, the output duration follows it as well. These constraints therefore cannot be overridden.

Validation timing depends on omni_reference_task_type#

Specifying the subtask explicitly moves validation to submission time and avoids failures that surface only after the task has started. Note that the model still determines the actual subtask from the prompt: if it disagrees with the value you specified, the task fails asynchronously with InvalidParameter.TaskTypeMismatch.

Recommended configuration: when the subtask is not known in advance, use omni_reference_task_type: "auto" together with ratio: "adaptive" and duration: -1. That combination satisfies the constraints of all three subtasks.

Billing is per token#

Every other video model here bills output seconds. Seedance does not:

tokens = width x height x frames / 1024      frames = seconds x 24 + 1

So a higher resolution or a longer clip usually costs more, driven by token count. Reference images and audio are free; reference video's cost is already inside that token count, and a request carrying reference video moves the whole task to a lower rate.

On retrieval, usage.output_video_tokens is what the charge is computed from (see the unified shape).