Upload your song
Choose an audio file that you own or have permission to process.
AI Vocal Remover
Separate vocals from instrumentals online.
Turn an authorized song into a clean vocal/acapella stem and an instrumental/karaoke stem. TM Converter's HQ separator runs the AI processing in your browser and gives you separate WAV downloads for each result.
Run HQ separation
The browser loads the AI model and separates the track into instrumental and vocal content.
Preview and download
Listen to each stem, then download the vocals or instrumental as a WAV file.
AI music tool · HQ mode
Separate vocals from the instrumental.
TM Converter now uses a heavier HQ MDX model with a two-pass denoise process for cleaner vocals and less vocal bleed in the instrumental.
HQ 2-Stem AI Separator
Vocal Separator
Audio stays on your device
HQ mode downloads a larger AI model (about 67 MB) the first time and will take longer than the earlier version. The HQ model predicts the instrumental directly, then TM Converter derives the vocal stem from the original mix. A two-pass MDX denoise process is enabled to reduce separation artifacts. Desktop Chrome or Edge is recommended for the fastest test.
HQ separation uses UVR-MDX-NET Inst HQ 3 with the UVR/MDX processing approach. Credit: Ultimate Vocal Remover developers, KUIELab / Woosung Choi, and model contributors.
Private workflow
Your selected song is designed to stay on your device while the separation runs in-browser.
HQ two-stem separation
Built specifically for the common creator workflow of extracting vocals and making an instrumental.
Separate downloads
Preview and download the vocal/acapella or instrumental/karaoke output independently.
Vocal Separator FAQ
Common questions
What does the TM Converter Vocal Separator create?
It creates two audio stems: a Vocals / Acapella stem and an Instrumental / Karaoke stem.
Does TM Converter upload my song to its server?
The current Vocal Separator is designed to decode and process the selected audio in your browser. The AI model and runtime are downloaded to the browser, while the selected song is not intentionally uploaded to TM Converter for separation.
Why can vocal separation quality vary?
Results depend on the original mix, mastering, vocal effects, stereo placement, harmonies, background vocals and how much the vocal overlaps other instruments in the frequency spectrum.
What audio can I process?
Use only audio that you own or have permission to process. The current tool accepts common browser-decodable audio formats and produces 44.1 kHz stereo WAV stems.