Make a karaoke track

Turn a song into a sing-along session: synced lyrics on screen, the vocals stripped out of the mix, your mic monitoring on top — and the key shifted to where your voice lives. All inside Weeps, all local.

The recipe

1. Bring the song in. The Explorer toolbar's Add a storage or import files button is the entry point — Add storage indexes a folder where it lives, or Import files copies loose files into the library (File → Storages manages them later).

   

2. Fetch the lyrics. Right-click the track → Enrich metadata… and let it pull cover art, tags and lyrics. Karaoke needs synced lyrics (the LRC kind, timed per line) — plain lyrics just sit there as static text. If the song comes back without them, add or sync them yourself in Now Playing → Edit Lyrics.

   

3. Open the track in the Editor. The track's ⋯ menu or right-click → Call Editor (also View → Editor). The song opens as a session with the track on the master lane.

   

4. Split the vocals out. Toolbar Stems opens the Separate stems panel — pick the model Vocals + Instrumental and hit Separate stems. It's the fast 2-stem split, made for karaoke — much quicker than the full 4-stem model because only two sources are computed. When the job finishes, the stems land as channels in the session: vocals and instrumental.

   

5. Keep only the backing. In the floating mixer, press S (Solo) on the instrumental channel — or M (Mute) on the vocals one. Now only the backing track plays.

   

6. Bring up the words — and your voice. The toolbar's Lyrics button opens a floating karaoke view that follows the playhead. To sing into the mix: Add channel → Microphone, pick your device, run the calibration, and enable Monitor input — use headphones to avoid feedback. Arm the mic lane and hit Record to capture the take.

   

7. Fit the key to your voice. If the song sits too high or low: on the instrumental channel's mixer strip, open More and set Key — up to ±6 semitones, duration preserved. When the lane's key is detected, the field even shows the resulting key (like 8A → 10A).

   

Keeping the result

• Render mix… (toolbar) bakes the session — muted vocals, your overdub, the key shift — into a file under Edits/ in your library.

  

• Just want the backing? After a 2-stem split, the instrumental's row in the track's Audio Lab tab offers Add Instrumental Version — a library copy that inherits the song's metadata and synced lyrics, ready to play like any track. Export on the same row saves the file elsewhere and writes a synced-lyrics .lrc sidecar next to it, so other players can show the words too.

Good to know

• The stems picker also carries minus-one models — Drums + Drumless, Bass + Bassless, Guitar + Guitarless, Piano + Pianoless — the same trick for practicing an instrument instead of singing.
• Separation runs locally via Demucs; the first run may need the audio tools pack — the panel tells you and offers the install.
• Sessions are resumable — close the Editor and pick it back up later; the session browser lists, reopens and deletes them.

Reference details live in the docs — this page is the guided version.