open-genjutsu: open-source local-first AI video performer replacement hits 1k stars
matchaman11 · x · 2026-11-03
open-genjutsu, a new open-source project (1k GitHub stars, MIT license), offers a local-first AI video recasting workflow: swap the performer in a short video using Seedance 2.5 or Wan 3.0, review the draft, explicitly approve HD export, and restore the source soundtrack in a local MP4.
- Both modes share the same Muapi preparation stages (depth, environment, character, prompt references): Genjutsu uses Seedance Motion Control, Genjutsu Lite uses Wan 3.0
- Local-first: source media is previewed and retained on your device
- Built with Python, FastAPI, and FFmpeg; self-hostable
Related event: Open-Source Video Character Swap Tool open-genjutsu Gains Traction(2 posts)→
More from Multimodal
- HeyGen Voice Turbo, top of TTS leaderboard, goes free on OpenRouter until Nov 30 — HeyGen · 2026-11-04
- Caméra Studio: interactive camera-framing workshop built with Scenario MCP and GPT Astra — Loo_Atreides · 2026-11-04
- OneShot MV: Open-Source ComfyUI Pipeline Turns a Song and One Sentence Into a Lip-Synced Music Video — luckyyirish · 2026-11-04
- Dreamina Canvas demo: one-canvas 30s short film with locked character consistency — eyishazyer · 2026-11-04
- A MiniMax H3-generated shot costs just $0.38 on heterogeneous compute — DavidBennett__ · 2026-11-04
- Musician demos an impromptu 'ComfyUI anthem' generated live in Stem Studio — PurzBeats · 2026-11-04