Seedance 2.0 on Replicate

Replicate's official guide + API for ByteDance Seedance 2.0 text/image/video-to-video

Tier A · StrongArticleAI ToolServicereplicate.com

Replicate's engineering blog post 'How to make remarkable videos with Seedance 2.0' by Shridharathi, published 15 April 2026. It documents the model's multimodal input contract — up to 9 images, 3 video clips, 3 audio files and a text prompt in one generation, each addressable inside the prompt as [Image1], [Video1], [Audio1] — plus native synchronised stereo audio from the same engine, and time-coded multi-shot prompting in the form [0-4s]: ... [4-8s]: .... It ships copy-ready Python and JavaScript snippets against the model slug bytedance/seedance-2.0 and closes with a prompting-tips section. Use it as the canonical API-side reference when you need programmatic video generation rather than a web UI.

Install

pip install replicate

For an agent

Use this over the Higgsfield UI whenever the video generation has to happen in code — it gives you the exact slug bytedance/seedance-2.0 and working Python/JS calls. Hard limits to design around: 15 seconds maximum per generation, 480p or 720p only, aspect ratios 16:9 / 4:3 / 1:1 / 3:4 / 9:16 / 21:9 / adaptive, and duration -1 to let the model choose. Structure long prompts as time-coded blocks and describe the audio explicitly, because audio is generated by the same pass — if you leave sound undescribed you get a soundtrack you did not ask for rather than silence.

Why it is here. It is the shortest path from 'we want AI video in the product' to a running API call, and the time-coded prompt grammar it documents is the piece most third-party tutorials get wrong.

Install command, framework, licence and component list.

Facts

License: Proprietary / all rights reserved
Pricing: Usage-based / credits, Free but account required
Framework: Python, Vanilla JS
Styling: Not applicable
Distribution: Public HTTP API, Hosted SaaS product, Article, thread, or post
Platform: Web
Content type: Video, Prose / writing, Code
Agent readiness: Public API without auth, Structured documentation, Ships a ready LLM prompt
Accessibility: Not applicable
Maturity: Actively maintained
Pricing detail: Not stated in the post; Replicate bills per run and the model page did not expose a per-second rate at fetch time. Treat cost as unverified usage-based.
Package: replicate

Notable for

exact model slug bytedance/seedance-2.0
9 images + 3 videos + 3 audio in one call
[Image1]/[Video1]/[Audio1] prompt addressing
[0-4s]: time-coded multi-shot syntax
native millisecond-synced stereo audio

Provenance

Verified 2026-08-15. HTTP 200. https://replicate.com/blog/seedance-2 (fetched twice — once for the summary, once for section headings, code identifiers and the prompting-tips list), https://replicate.com/bytedance/seedance-2.0 (fetched for input parameters and limits; pricing and run count were not exposed).

Also in Generative Media Tools

replicate.comTier A