what is vulture-ai fr?

Vulture AI is a desktop app that bundles a collection of AI creative tools into one window, and it runs entirely on your own computer with no cloud connection, no API keys, and no subscription. The idea is to spare non-technical people from having to stitch together separate AI tools themselves, which normally means dealing with complicated node-based interfaces or the command line. Inside the app you get image generation using models like FLUX and Stable Diffusion, with a built in prompt helper that turns casual descriptions into proper prompts, plus 4K upscaling, image-to-image editing, face swapping, and lip syncing that brings a still photo to life using a driving video. There is also a local chat feature powered by tools like Ollama, and a local coding assistant built on Aider that can automatically test and repair its own code, along with a splitter that breaks big requests into smaller steps so weaker local models can keep up. Two local search systems, one for your own code projects and one for chat memory, run through the same embedded local database so nothing ever leaves your machine. The app itself is just a launcher, interface, and set of setup scripts. The actual AI models, which can total close to 100 gigabytes, are downloaded to your machine during setup rather than bundled with the repository. Most of these models are fine for commercial use, but a handful, like the face-swap and lip-sync models, are limited to personal or research use and are never downloaded automatically, only offered as manual links in the setup screen. Setup is aimed at Windows users with an NVIDIA graphics card with at least 6 gigabytes of video memory, and the project notes it currently runs reliably on the author's own machine while cross-machine compatibility is still being worked on. Installation can be done with a single script that downloads the required components, or manually through Python and Git. The project is early and actively evolving, with local video generation, text-to-music, voice cloning, and training your own custom models planned for future releases. The launcher and interface are released under the MIT license, while the bundled third-party tools and models each keep their own separate licenses.