AudioGhost v1.2: local open-source tool removes any sound by typing its name, built on Meta SAM-Audio
GGwithRabbit · reddit · 2026-11-18
AudioGhost v1.2 is a local, open-source web app built on Meta's SAM-Audio: you type what you want removed ("background music", "dog barking") and it separates or strips that sound from audio/video entirely on your own GPU — handy when YouTube flags a vlog for nearby shop music.
What's new in v1.2:
- Multiple sounds per run, each on its own track with mute/solo/volume like a tiny DAW
- The "everything else" track is now generated by the model instead of subtracted, so removals work much better
- Timeline markers to point the model at the right sound; WAV/FLAC/MP3 or video export; sessions survive restarts; 4-6x faster than v1
The small model needs just over 4 GB of VRAM, with install scripts for Windows/Linux (macOS runs on CPU for now). The author stripped out SAM-Audio's 25 GB of helper models (ImageBind, CLAP, judge model) to fit normal GPUs, trading some quality and run-to-run consistency. GitHub: 0x0funky/audioghost-ai.
More from Infra
- Google donates Agent Substrate to CNCF: 500ms restores, 10x density — krishnan · 2026-11-18
- Report: Nvidia brokered Alibaba-Megaspeed deal, Alice Huang ties run deeper than known — GaryMarcus · 2026-11-18
- n8n's open-source agent Nod: secure remote access via private reverse proxy — victorialslocum · 2026-11-18
- pci_eerie: Open-Source PCIe Soft Stack for Xilinx 7-Series FPGAs as a Drop-in PCIE_2_1 Replacement — IgorCarron · 2026-11-18
- Hugging Face Nears 1 Exabyte of Monthly Data Distribution as Open Model Downloads Explode — andimarafioti · 2026-11-18
- NVIDIA, AMD, Intel, Google and Meta join SGLang Consortium as v1.0 ships — heyshrutimishra · 2026-11-18