Hysen Labswind-comic
Multi-agent AI pipeline that turns one line of text into a finished short-form drama: script, cinematic storyboards, character-consistent video. Provider-agnostic (OpenAI/Claude, MJ, Minimax, Veo/Sora, fal, ComfyUI). MIT.
Wind Comic turns one line of text into a finished short drama
Wind Comic is a multi agent AI studio that expands a single sentence into a script, storyboards, voiceover, and a final mp4 short form drama, with bring your own LLM support.
The agent pipeline
Wind Comic does not rely on a single model. It is a multi agent pipeline where each role is a specialist that hands off work with strict consistency contracts. The roles include Writer, Director, Producer, Character Designer, Storyboard Artist, Cameo Locker, Lipsync, and Editor. From one line such as a reborn CEO confronting an ex fiancee, the pipeline produces a script, character bible, multi shot storyboards, voice acted lines, background music, and a final mp4.
Consistency and collaboration
Character consistency is kept through a style bible and reusable characters, and a lip sync pipeline aligns a viseme track, scores alignment, auto corrects drift, and renders back into the timeline. A real time multiplayer timeline built on Yjs CRDT lets a team edit like Figma for film. The project supports bring your own LLM with three environment variables and no code change, and plugs in more than twelve image and video providers.
Providers and testing
Wind Comic is provider agnostic and works with OpenAI, Claude, Midjourney, Minimax, Veo or Sora, fal, and ComfyUI. Voice and lip sync can use Kling, Sync.so, or Hailuo, while photorealistic face level lip sync is available through bring your own wav2lip, SadTalker, or MuseTalk. The repository reports 4333 passing tests and targets Node 20 or later with Next.js 16. It is licensed under MIT.
Editorial conclusion
Wind Comic is written in TypeScript, licensed under MIT, requires Node 20 or later, and reports 4333 passing tests in its repository.
Community notes