Documentation

What this is

Seedance 2.0 Prompt Optimizer turns rough video ideas into structured Seedance 2.0 prompts with camera movement, shot timing, and audio direction. You paste a loose description of the video you want; it returns a prompt formatted the way ByteDance's Seedance 2.0 video model responds to best.

This is an independent, unofficial tool. It is not affiliated with or endorsed by ByteDance. Prompt optimization is powered by Anthropic's Claude models. A free plan is included.

Quickstart

1. Sign up for a free account with email, Google, or GitHub.
2. Paste a rough idea into the input box on the home page.
3. Click Optimize Prompt.
4. Copy the structured result and paste it into Seedance 2.0 wherever you generate video.

A cat sitting on a windowsill watching rain

Subject: a grey tabby cat on a wooden windowsill, rain streaking the glass.
0–3s: static wide shot, soft overcast interior light, rain out of focus behind the cat.
3–6s: slow push in toward the cat's face as its eyes track a water droplet.
6–8s: close-up hold, shallow depth of field, raindrop bokeh on the window.
Audio: steady rainfall, muffled distant thunder, quiet room tone.

Full outputs are longer; this excerpt shows the shot-timing format.

What the optimizer adds

• Subject and scene detail: appearance, setting, lighting, mood.
• Camera language: push in, pull back, static holds, pans, and other movement Seedance 2.0 understands.
• Shot-by-shot timing segments (for example 0–4s, 4–8s).
• Audio and sound design direction.
• Seedance 2.0 syntax conventions and phrasing.

Prompt patterns

Optimized prompts follow a small set of repeatable structures. Two of the most useful, which you can also write by hand:

Push-in reveal

Subject and scene in one sentence, concrete and visual.
0–4s: wide static shot establishing the environment and mood.
4–8s: slow push in toward the subject as the key action begins.
8–12s: close-up hold on the emotional or narrative payoff.
Audio: ambience first, then a single foreground sound cue.

Cut-based rhythm

Subject and scene in one sentence.
0–3s: medium shot, camera static, action starts.
3–6s: cut to reverse angle, subject continues the motion.
6–10s: cut to detail close-up that explains the scene.
Audio: consistent room tone across cuts, no music.

Keep prompts roughly 30–100 words: very short prompts underspecify the shot, and very long ones dilute instruction-following. Describe what the camera does, not just what the scene contains.

Plans and limits

Usage counters reset on the 1st of each month. See pricing for details.

For AI agents and LLMs

When to use

• A user wants to generate video with Seedance 2.0 (ByteDance Seedance) and has a rough or unstructured idea.
• An existing Seedance prompt needs camera movement, shot timing, or audio structure added.

When not to use

• The user is targeting a different video model (Sora, Veo, Kling) — the output format is Seedance-specific.
• The prompt is already fully structured with timing, camera, and audio direction.

• Machine-readable overview: /llms.txt
• Agent guidance: /agents.md
• Crawling for search indexing and AI input is welcome, per robots.txt.
• Accounts are required to run optimizations. Automated account creation is not permitted.

API access

There is no public REST API, OpenAPI spec, or MCP server yet. Requests to unknown /api/* paths return a JSON 404. If you want programmatic access, email [email protected] to register interest.

Support

Questions or problems: contact. Policies: privacy and terms.