what is khamaileon fr?

Khamaileon is a full stack pipeline that turns a short video into an anime style animation using AI image generation. It takes an ordinary clip and restyles it frame by frame into looks inspired by JoJo's Bizarre Adventure, Bleach, Dragon Ball Z, or Studio Ghibli, while trying to keep the original motion, facial structure, and scene layout intact. Under the hood, the pipeline extracts every frame from the input video with FFmpeg, resizes and prepares each one, then runs it through SDXL, an image generation model, guided by ControlNet, a technique that locks in the original pose, camera angle, and composition while letting the model redraw details like eyes, hair, and clothing into the chosen art style. Because each frame is generated somewhat independently, small differences between frames can cause visible flickering. The project addresses this with a fixed random seed, a limited number of diffusion steps, and carefully written negative prompts that suppress unstable effects like glare or glowing eyes. After generation, FFmpeg reassembles the stylized frames into a video, applies a filter to smooth out remaining flicker, and merges the original audio back in. The project has a React and TypeScript frontend built with Vite, and a Python backend built with FastAPI, with the actual image generation calls going through the Replicate API. Setting it up locally means cloning the repository, creating a Python virtual environment and installing backend dependencies, adding a Replicate API token to an environment file, starting the backend, and separately installing and running the frontend with npm. The README lists a roadmap of planned additions, including custom prompt support, more anime styles, Docker deployment, batch processing of multiple videos, character specific style tuning, and better tools for keeping frames consistent over time. The project is licensed under the MIT license, and the author welcomes contributions through forks and pull requests. As of this writing, some of the demo videos referenced in the README have not yet been embedded.