Genjustsu Open Source Workflow

Genjustsu Open Source Workflow is a local AI video replacement workflow published by TEIN. It helps creators prepare video clips, review segmentation and tracking assets, call hosted generation services, and export a final video with the original source audio restored. The GitHub repository describes a workflow that combines colored depth maps, SAM 3 video segmentation masks, face mesh tracking, audio isolation and pitch processing, and Seedance generation through Enhancor. The project is aimed at users who want more control than a one-click video generator gives them. Genjustsu breaks the job into reviewable steps. Users upload MP4 or MOV clips, prepare depth and mask assets, inspect the prepared media in a local web interface, generate drafts, and approve HD output only after the preview looks right. That review loop matters because subject replacement work is easy to ruin with bad masks, face drift, or audio mismatches. The repository explicitly describes a draft-first flow and says the final export restores the whole original source soundtrack rather than replacing it with generated audio. Genjustsu is not a hosted SaaS with bundled credits. It is an open-source workflow that needs external keys for generation and model calls. The README references Enhancor for Seedance access, Replicate for hosted model calls such as Demucs, Depth Anything Video, and SAM 3 Video, and an optional Hugging Face token for an alternative depth path. The project includes setup material for agents, local launch scripts, examples, and a default local UI on port 8770. The best audience is technical video creators, AI filmmakers, and agent-assisted media teams who are comfortable running local Python tooling and supplying API keys. It can help with character replacement, object replacement, scene experiments, and repeatable draft-review-export workflows. The code and documentation are MIT licensed, while TEIN artwork, names, logos, and example media are excluded. That distinction matters for teams planning to fork or adapt the workflow for their own production pipeline. For OpenTools readers, Genjustsu is strongest when treated as a workflow scaffold rather than a finished creative app. The repository gives builders a starting point for video replacement experiments that still need careful setup, key management, asset review, and human approval. It is a good fit for technical teams that want to inspect each stage of the generation chain instead of sending a clip to a black-box editor and hoping the result is usable. That transparency is the core reason to track it as a builder tool.