Breaking

Curated Seedance 2.0 prompt collection shared on GitHub

July 6, 2026 at 19:57 EDT

Media Generation
Open Source
Foundation Models

Developer Tom Dörr has highlighted a GitHub repository, "awesome-seedance," that gathers high-quality prompts for ByteDance's AI video model Seedance 2.0. It focuses on cinematic transitions and character consistency, curating examples from across X, WeChat and seasoned prompt engineers.

February 2026 · Seedance 2.0

A Community Playbook for Cinematic AI Video

A curated prompt collection for ByteDance's flagship "thinking video model" is spreading among creators — teaching precise timestamp control and character consistency to tame a powerful but hard-to-steer generator.

images can be fused in a single generation

3 + 3

video clips plus audio tracks per prompt

~15s

multi-shot race scene example

Multimodal input, stacked into one generation

Number of inputs the model can fuse simultaneously (per unit block).

1×

Text prompt

9×

Images

3×

Video clips

3×

Audio tracks

The prompt recipe for "professional-grade" control

Subject · Action · Camera · Style · Constraints

→

Timestamp notation
[00-05s] Shot 1

→

@-mention reference assets to lock characters

→

Consistent multi-shot scene

Strengths

Multimodal input flexibility vs. Kling, Veo, Sora
Stronger character consistency
Native audio sync and frame-level control

Open challenges

"Drift" — characters becoming someone else
Coarse prompts break physics and consistency
Copyright concerns prompted an MPA request

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.