Spleeter
Free.ai (self-hosted)
·
audio_separate
·
~1000 tokens per track
Spleeter is a stem separation model built by Deezer. Strongest at DJ/production workflows, karaoke extraction.. Self-hosted on Free.ai GPUs — runs free against your daily token pool (1,000 tokens per track). Released under MIT — commercial use permitted on Free.ai.
Use via API
OpenAI-compatible REST API. Generate a key and call this model in seconds.
curl -X POST https://api.free.ai/v1/music/separate/ \
-H "Authorization: Bearer sk-free-..." \
-H "Content-Type: application/json" \
-d '{"model":"spleeter","prompt":"your prompt here"}'Frequently Asked Questions
Spleeter separates a mixed audio track into stems — vocals, drums, bass, and other instruments — each as its own audio file. Upload an MP3 or WAV and Spleeter returns 4 separate stem files you can remix or mute.
Demucs-family separators are state-of-the-art — vocals are nearly studio-quality on modern productions, drums and bass are very clean. Older or heavily-compressed source material yields lossier separation.
Standard 4-stem: vocals, drums, bass, other. Some models offer 6-stem (adding piano + guitar separately) — pick the variant in the form.
MP3, WAV, FLAC, M4A, OGG up to 50 MB. Stereo input gives the cleanest results; mono is upscaled to stereo before processing.
Spleeter runs on our GPUs against your daily free pool first; $5 → 200,000 paid tokens. About ~1,000 tokens per minute of audio.
Yes — extract the stems, then download "drums + bass + other" stitched together for an instrumental, or download "vocals" alone for an a-cappella. The form above has a one-click "instrumental" output.
Yes — separation does not alter timing or pitch. Stems align frame-for-frame with the source so they remix cleanly.
Yes — /batch/ accepts a folder of audio files. Each output (4 stems per source) lands in /account/?tab=history. For folder-tree preservation use the API.
Yes — POST your audio to /v1/music/separate/ with model="Spleeter". Returns 4 download URLs (one per stem). /api/ has the full reference.
You need rights to the source audio (your own recording, licensed material, or a song with explicit remix permission). Free.ai cannot grant rights to material you do not own; separating a chart pop song does not legalize publishing the stems.
WAV by default (lossless, ready to remix). MP3 is available in the format picker for smaller downloads.
Real-time factor is roughly 0.1–0.3× — a 4-minute song separates in 30–80 seconds. Use the queue button to close the tab.