SAM Audio supports text, visual, and temporal prompts so you can target specific sounds in dense mixes.
Stem separation for production audio.
SplitStudio brings Meta SAM Audio into a pro desktop workflow with waveform selection, model-size control, queue-aware runs, and export-ready isolated/remainder stems. Powered by Meta's SAM Audio.
macOS desktop appSmall / Base / Large modelsLocal queue + batch processingExport: WAV, FLAC, MP3, MP4
Launching soon - Join the waitlist for first access
Interactive demo
Prompt-driven audio editor
Real waveform editor extracted from the app. Choose one of 3 prompt dropdown options, run fake processing for 2 seconds, then preview the matching output.
Drag the waveform playhead, change settings, then run separation.
City Night Mix.wavReady
Duration 0:00
Output Preview
Run separation to preview output.
Why SplitStudio
Fast, private, local by default
SplitStudio keeps everything on-device, so you can tweak prompts, swap model sizes, and export stems without waiting on uploads, cloud queues, or render credits.
Local queue + progressRun multiple jobs without waiting on a cloud pipeline.
Target + residual outputAlways get the isolated stem and the leftover audio.
Model size controlSmall, Base, and Large modes for speed or maximum detail.
Power feature
Prompt-Driven Isolation
This visual is an example workflow using a podcast mix. In the app, you can prompt for voices, instruments, or background elements in your own audio.
Example only: "Male vocals" and "Female vocals" are demo prompts, not fixed modes.
- Input: Any recording or full mix
- Prompt: Plain-language target request
- Result: Isolated stem plus remainder output
Input audio fileSource
Podcast (original)episode_mix.wav
Original blend
Prompt
Male vocals"Isolate male speech"
Male extract
Prompt
Female vocals"Isolate female speech"
Female extract
System requirements
Runs entirely on your machine.
No cloud. No uploads. No tracking.
macOS 12+Apple Silicon only
8 GB RAM minimum16 GB recommended for larger jobs
Platform credibility
SAM Audio power, private by design.
SplitStudio runs Meta's SAM Audio workflow locally so your project audio stays on your machine from import to export.
Each run returns the requested stem plus the remainder, so edits stay precise and reversible.
Inference stays local on Apple Silicon with no cloud queues and no tracking.
Ready to run local separation
Be first to try SplitStudio.
Join the launch waitlist and share what you need. We are collecting early user feedback and feature priorities before release.
InputAudio + video source files
OutputIsolated and remainder stems
FormatsWAV, FLAC, MP3, MP4
ProcessingFully local on your machine
ModelSmall / Base / Large