What is Sam Audio?

SAM Audio applies Meta’s Segment Anything Audio Model for AI audio separation and sound isolation.

It isolates vocals, instruments, speech, and sound effects from complex mixes using multimodal prompts—natural-language text, visual selection, or time-span annotations.

Span prompting enables precise temporal selection for targeted extraction; outputs include target and residual stems while preserving original sample rates.

Workflows supported include stem separation for music production, dialogue enhancement for podcast editing, extraction of effects for film/video post-production, and signal isolation for research and accessibility.

Inputs include common audio and video formats; deployment uses Apache-2.0 licensed models on dedicated infrastructure for offline inference and integration into editing pipelines.

Designed for music producers, audio engineers, podcasters, filmmakers, researchers, and accessibility developers seeking accurate, prompt-driven audio component extraction.

Sam Audio user reviews

Would you recommend Sam Audio?

Recommend this tool?

Sam Audio's key features

Multimodal prompting (text, visual, and time-based)
Span prompting for precise temporal (time-range) isolation
Unified audio separation model for speech, music, instruments, and sound effects
Exports target and residual audio for editing workflows
Preserves original sample rates for professional-quality outputs

Sam Audio use cases

Create isolated vocal, instrument, and effect stems from mixed tracks using SAM Audio's multimodal (text, visual, time-span) prompts to produce target and residual stems at original sample rates for high-quality remixing, mastering, and collaborative production
Develop clean, intelligible dialogue for podcasts and film post-production by using SAM Audio's time-span and visual audio selection to precisely extract speech, remove background ambience, and prepare stems for ADR, noise reduction, and loudness-compliant delivery
Create reusable instrument samples and sound-design assets or run large-scale audio research locally with SAM Audio's prompt-driven, offline inference to extract specific sound events, build sample libraries, and analyze mixes while preserving privacy