Meta’s SAM Audio Explained (And Why It Matters)

2026-01-06 · Source: Matthew Berman · Field: Technology & Digital — Artificial Intelligence & Machine Learning · Depth: Intermediate, short

Summary

Meta has released SAM Audio, an open-source and open-weights model designed for isolating specific sounds from video and audio files using simple text prompts. The model, part of the SAM 3 family, is available for free in Meta's Segment Anything playground. Demonstrations showcase its ability to accurately isolate a woman's voice from a Tomb Raider video, separate voice, footsteps, and utensils from a noisy restaurant scene, and isolate individual instruments like a guitar from a song. Users can generate three tracks: the original, the isolated sound, and everything but the isolated sound. The platform also offers various sound effects, such as studio sound, classic 8s, and robot voice, which can be applied and tuned. Isolated tracks can be downloaded individually for further use.

Key takeaway

For video editors and audio engineers seeking efficient sound manipulation, Meta's SAM Audio offers a powerful, free solution. You can use it to quickly remove unwanted background noise from recordings or isolate specific audio elements for creative mixing. Experiment with its text-prompt isolation and built-in sound effects to streamline your post-production workflow and achieve cleaner, more focused audio.

Key insights

Meta's SAM Audio model enables precise sound isolation from media using text prompts, offering significant utility for audio manipulation.

Principles

• Text prompts enable granular audio control.
• Open-source models foster broad utility.

Method

Upload audio/video, type a sound prompt (e.g., "woman," "footsteps," "guitar"), and SAM Audio generates isolated, original, and inverse tracks. Apply optional sound effects and download individual tracks.

In practice

• Clean up background noise in videos.
• Isolate specific instruments in music.
• Enhance voice clarity in recordings.

Topics

• SAM audio
• Sound Isolation
• Open-Source AI
• Audio Processing
• Audio Cleanup

Best for: Creative Technologist, Machine Learning Engineer, AI Product Manager

Related on AIssential

• Introducing SAM Audio: The First Unified Multimodal Model for Audio Separation | AI at Meta — SAM Audio unifies multimodal audio separation using text or visual prompts for diverse applications.
• SAM Audio - AI at Meta — SAM Audio offers state-of-the-art sound separation using multimodal prompts across diverse audio types.
• How to Remove Background Music from a Video | Free Tool for Clear Audio — An online tool effectively isolates spoken voice from background music in videos, improving clarity and usability.
• DirectAudioEdit: Inversion-Free Text-Guided Audio Editing via Diffusion Prediction Contrast — DirectAudioEdit enables efficient, inversion-free text-guided audio editing, outperforming inversion-based methods in speed and quality.
• Local Whisper Audio Transcription — Faster-Whisper enables fast, private, and local audio transcription using optimized Whisper models in Python.
• New open-source voice model listens nonstop and decides every 0.4 seconds whether to speak or stay silent — The "Audio Interaction" model unifies continuous audio processing, multi-tasking, and proactive response in a single system.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Matthew Berman.