TuneLab
Vocal Remover
Remove vocals from any song using AI. Runs on your device — files never uploaded.
Drop audio file here or click to browse
MP3, WAV, FLAC, OGG, M4A · Up to 15 minutes
Private — processed on your device, never uploaded
Faster processing, smaller download. First use downloads a 14 MB AI model.
Preparing...
Processing locally
How to Remove Vocals
- 1 Upload a song (MP3, WAV, FLAC, OGG)
- 2 AI separates vocals, drums, bass, and instruments
- 3 Download individual stems or play them back
Frequently Asked Questions
What's the difference between Fast and Quality mode?
Fast mode uses a 14 MB model optimized for speed — typically 15-30 seconds for a 4-minute song. Quality mode uses a larger 32 MB model trained on more diverse material, better at preserving bass clarity and handling reverb tails. Choose Fast for quick previews, Quality for final production use.
How does vocal removal actually work?
A neural network trained on thousands of professional multitracks learns to separate the frequency patterns of vocals from instruments. It processes the audio spectrogram in overlapping chunks and reconstructs separate vocal and instrumental waveforms.
Does it use my GPU?
When WebGPU is available (most modern desktop devices), inference runs on your GPU — significantly faster than CPU-only processing. On older devices, it falls back to multi-threaded WASM automatically. No configuration needed.
Will it remove backing vocals too?
The model separates lead and backing vocals together from the instrumental. For tracks where you want to keep harmonies but remove the lead vocal only, results may vary — the model treats all vocal content similarly.
Does it work on podcasts or speech?
Yes, though it's optimized for music. For spoken word with background music, it effectively separates the voice from the accompaniment. For speech with only ambient noise, a dedicated noise removal tool may produce cleaner results.
What about processing on mobile?
On mobile devices without sufficient GPU memory, processing automatically routes to Cloud Assist — secure server-side processing where files are deleted immediately after completion. Desktop users process entirely on-device.
What output formats are available?
Both the isolated vocal track and instrumental track are exported as lossless WAV at the original sample rate. The files maintain the same duration and can be loaded directly into a DAW.
