Show HN: AudioGhost AI – Run Meta's Sam-Audio on Consumer GPUs (4GB-6GB VRAM) (github.com)

AudioGhost AI has launched a user-friendly tool that leverages Meta's SAM-Audio model, enabling audio separation tasks on consumer-grade GPUs with as little as 4GB of VRAM. Users can simply describe the audio elements they wish to extract or remove—such as vocals or specific background sounds—using natural language. Significant optimizations in the software reduce memory usage dramatically from the original ~11GB required to just 4GB in a "Lite Mode." This means more users can experiment with sophisticated audio processing without needing high-end hardware. The platform boasts a sleek modern UI with features for real-time progress tracking, a stem mixer for audio comparison, and support for video uploads. AudioGhost's backend utilizes FastAPI and a task queue with Celery to streamline processing, allowing for efficient audio manipulation. Notably, this innovation democratizes access to advanced audio separation technology, opening new possibilities for musicians, sound designers, and creators who may not have previously had the necessary computational resources. The tool's modular setup also encourages community contributions and enhancements, making it a notable development within the AI/ML ecosystem.