Anil-matcha/open-genjutsu

Local-first AI video recasting with draft review, explicit HD approval, and source-audio restoration.

What it solves

Open Genjutsu allows you to replace a person in a video with a different character while preserving the original movement, background, and location. It simplifies the process of creating character animations by handling the motion-control workflow in a single request.

How it works

You provide a source video (4–30 seconds) and a reference image of your desired character. The application runs a local server that manages the upload, sends the data to a configured video generation API, and monitors the progress. Once the generation is complete, the app automatically downloads the result and merges it with the original video's soundtrack using FFmpeg.

Who it’s for

This tool is for creators and developers who want a local-first interface to manage video recasting workflows. It is designed for users who prefer to keep their media files and project previews on their own machine while using an external API for the heavy lifting of video generation.

Highlights

  • Local-first workspace: keeps source media and exports on your local device.
  • Automatic audio restoration: merges the original source soundtrack into the generated video.
  • Simplified workflow: uses a single motion-control request rather than complex multi-step jobs.
  • Privacy-conscious: API keys are held only in browser memory and are not saved to disk.
  • Built-in preview: allows you to review source clips and character references before submitting a job.

Related

  • Project
    marcelo-earth/generative-manim

    Generative Manim is an open‑source tool that uses LLMs (GPT‑4o, Claude, Gemini, etc.) to turn plain‑English prompts into runnable Manim animation code and rendered videos. It provides a web demo, a REST API, and a desktop app (Animo) for local rendering, plus a training pipeline and benchmark suite for fine‑tuning open‑weight models to generate Manim code.

  • Project
    lemenkov/libyuv

    libyuv is an open-source library for scaling, converting, and rotating YUV video data, optimized for multiple CPU architectures. It handles YUV scaling (with three filter options), webcam-to-YUV conversion, YUV-to-RGB conversion, and 90/180/270-degree rotation for mobile portrait mode. Performance optimizations target SSSE3/AVX2 on x86/x64, Neon/SVE2/SME on Arm, MSA on Mips, and RVV on RISC-V. Development resources include a getting started guide and a docs directory.

  • Project
    Thmen/EGVSR

    A PyTorch implementation of EGVSR, an efficient video super-resolution framework that uses subpixel convolution to speed up inference for high-quality video upscaling.

  • Project
    UCSC-VLAA/story-iter

    A training-free iterative framework for long story visualization that maintains semantic consistency across up to 100 frames using a global reference cross-attention module.