Skip to content

Seedance 1.5 Pro

Overview

Seedance 1.5 Pro is BytePlus ModelArk’s video-generation line. BytePlus describes Seedance as a high-quality video generation family with strong semantic understanding and prompt guidance, and the 1.5 Pro prompt guide focuses on helping users drive coherent cinematic output from text and reference assets. The implemented billing logic treats it as a token-priced model, which is a key distinction from providers that charge more directly by wall-clock duration.

API base URL

https://ark.ap-southeast.bytepluses.com/api/v3/contents/generations/tasks

Example payload

{
  "model": "seedance-1-5-pro-251215",
  "content": [
    {
      "type": "text",
      "text": "A smooth orbit shot around a sports car"
    },
    {
      "type": "image_url",
      "image_url": {
        "url": "https://example.com/first.jpg"
      },
      "role": "first_frame"
    }
  ],
  "generate_audio": true,
  "ratio": "16:9",
  "duration": 5,
  "watermark": false
}

Our Current Pricing

Seedance 1.5 Pro uses the most nuanced video formula because the code can meter directly from completion tokens when they are available, and otherwise estimate token volume from video dimensions, frame rate, and duration.

Pricing key resolution is based on the first four model-id segments. For seedance-1-5-pro-251215, the key is seedance-1-5-pro.

Token price per 1M tokens:

seedance-1-5-pro with audio    -> 2.4 USD / 1M tokens
seedance-1-5-pro without audio -> 1.2 USD / 1M tokens
seedance-1-0-pro               -> 2.5 USD / 1M tokens
seedance-1-0-lite              -> 1.8 USD / 1M tokens
unknown                        -> 1.8 USD / 1M tokens

Preferred formula when the provider returns completion tokens:

cost_usd = (completion_tokens * token_price_per_million) / 1_000_000

Fallback token estimate when completion tokens are missing:

estimated_tokens =
  (width * height * fps * ceil(duration_seconds)) / 1024

cost_usd =
  (estimated_tokens * token_price_per_million) / 1_000_000

The implementation enforces a minimum charge of 0.01 USD.

The fields that influence cost are therefore generate_audio, resolution, fps, duration, and any provider-returned completionTokens. For billing, always prefer the real token count from the API response. Only fall back to the geometric estimate when usage is absent, and store a flag such as metering_source = provider_usage | estimated so finance teams can distinguish exact from inferred charges.

Sources