Devotd Github
NewMultimodal

lucataco/open-source-higgsfield — local video character-replacement pipeline

A repository that bundles open-weight video components to swap the person in filmed scenes for a different character in local ComfyUI.

Repositorylucataco/open-source-higgsfield ↗

A repository that bundles open-weight video components to swap the person in filmed scenes for a different character in local ComfyUI.

GitHub description: Same take, new cast: open-weights video character recasting on local ComfyUI with SAM 3.1, FLUX 3 Image Klein and SCAIL-2. Built for #OpenSourceHiggsfield.

stars
★ 23
Forks
4
Language
Python
License
Apache-2.0
Read original source ↗View on GitHub ↗

lucataco/open-source-higgsfield is a local ComfyUI pipeline repository that changes the person in a filmed video into a different character while keeping the scene intact. It uses SAM 3.1, FLUX 3 Image Klein, and SCAIL-2 together. No separate newly released version or release of this repository was given in FACTS. Therefore, the change discussed here is not an update to a specific version but the workflow the repository presents.

According to the README and the code, the work is divided into stages: person tracking and re-identification, camera-motion tracking, per-segment generation, and quality checking and resumption. Re-identification is the process of linking the locations where the same person reappears in the video. The key point is that it structures editing that changes only the person while preserving the scene into multiple stages. The README explains that it does not depend on separate node bundles or a closed application interface, and the tool aims to be an alternative to Higgsfield Genjutsu. The repository's language is Python and its license is Apache-2.0.

However, the provided information contains no official release and no performance measurements. The fact that the work stages are broken down concretely is not enough to judge result quality or stability across diverse videos. Therefore, what can be confirmed right now is the composition and direction of the local character-replacement pipeline, while actual performance remains open until measurement data is presented.