Editors, podcasters and musicians who need to pull one sound out of a mix get a free, open model that takes text or clicks.
Kenning's take
What it does
SAM Audio separates a chosen sound, such as a voice, an instrument or background noise, from an audio or video mix and returns both the isolated sound and the rest. It takes text descriptions, clicks on the part of a video making the sound, or a selected time span, alone or combined. Meta released it on 16 December 2025 in its Segment Anything Playground and as a download on GitHub.
- Text prompts such as 'dog barking'
- Visual prompts by clicking an object in a video
- Span prompts that mark a time range
- Returns target and residual audio stems
- Evaluation dataset on Hugging Face
Who it suits. Video editors, podcasters, musicians and researchers who need to isolate or remove sounds.
Key facts
- Category
- AI models
- Made by
- Meta
- Based in
- United States
- Released
- 16 December 2025
- How to get it
- free to try in the Segment Anything Playground; model on GitHub (facebookresearch/sam-audio)
- Works on
- Web, GitHub
- Free plan
- free in the Segment Anything Playground and as an open model
Reviews and write-ups
- The Register 2025-12-16
FAQ
When was SAM Audio released?
Meta released SAM Audio on 16 December 2025.
How do I use SAM Audio?
Try it in Meta's Segment Anything Playground or download it from GitHub.
What prompts does SAM Audio accept?
Text descriptions, clicks on a video, and time-span selections, alone or together.
Facts from the maker's own pages and the sources listed. Checked 2026-10-11.
Similar tools
ACE-Step 1.5
Open-source music generation model that writes full songs with lyrics and runs on ordinary consumer GPUs.
AlphaGenome Atlas
Google DeepMind database of AI-predicted effects for every possible single-letter change in human DNA.
Amazon Nova 2
Amazon's second family of Nova AI models, covering reasoning, speech conversation and combined text-and-image generation.
Cisco Antares
Cisco's small open-weight models that find where known security vulnerabilities sit in a codebase, running locally.