Tired of Spleeter's fixed stems? SAM-Audio by Meta AI lets you isolate any sound with text prompts. Not just vocals, drums, bass, and other. Literally any sound you can describe.
See why SAM-Audio is the clear upgrade from Spleeter.
| Feature | SAM-Audio | Spleeter |
|---|---|---|
| Separation Categories | Unlimited (any sound) | 2 or 4 fixed stems only |
| Text Prompts | Yes | No |
| Visual Prompts | Yes | No |
| Temporal Prompts | Yes | No |
| Specific Instrument Isolation | Yes | No |
| Background Vocal Separation | Yes | No |
| Sound Effect Isolation | Yes | No |
| Video Support | Yes | No |
| Quality (SOTA) | Yes | No |
| Developer | Meta AI (2024) | Deezer (2019) |
Spleeter was groundbreaking in 2019. SAM-Audio is the next generation.
SAM-Audio solves Spleeter's biggest limitations.
Spleeter only does vocals, drums, bass, and "other". SAM-Audio isolates anything: "acoustic guitar", "crowd cheering", "hi-hat cymbal", "child laughing".
"violin in the background"
Spleeter is audio-only. SAM-Audio can use video frames to identify sound sources. Click on a speaker in a video to isolate their voice.
Spleeter requires knowing what category you want. SAM-Audio lets you mark a time range where your target occurs, then extracts similar sounds.
Real scenarios where SAM-Audio outperforms Spleeter.
With Spleeter, if you want just the piano, it's stuck in "other" with everything else. With SAM-Audio, type "piano" and get just the piano.
Spleeter's "vocals" stem includes all vocals together. SAM-Audio can separate "background harmonies" from "lead vocals".
Spleeter is trained on music. SAM-Audio handles speech, sound effects, ambience - any audio content.
Spleeter was state-of-the-art in 2019. SAM-Audio uses 2024's best techniques with Meta AI's massive training resources.
Common questions about switching from Spleeter to SAM-Audio.
SAM-Audio by Meta AI is the best Spleeter alternative. While Spleeter only separates into 2 or 4 fixed stems (vocals, drums, bass, other), SAM-Audio uses text prompts to isolate any sound you describe. It also produces higher quality separations using state-of-the-art AI research.
Yes! You can try SAM-Audio for free at TwoShot. Like Spleeter, SAM-Audio's code is also open-source on GitHub for developers who want to run it locally.
Yes, and much more. SAM-Audio can replicate Spleeter's 4-stem separation by running it multiple times with different prompts ("vocals", "drums", "bass", "other instruments"). But it can also isolate specific instruments, background sounds, or any audio you describe.
SAM-Audio achieves state-of-the-art quality that significantly exceeds Spleeter. It uses a flow-matching diffusion transformer architecture trained on Meta AI's massive datasets, representing 5 years of AI advancement since Spleeter's release.
SAM-Audio achieves 0.7x real-time processing on modern GPUs. For most use cases, you'll get your results within seconds. The cloud version at TwoShot handles the processing for you.