Replay

Verified

Free desktop app from Weights for AI vocal conversion, stem splitting and voice training

Visit Replay →No rating yetFree

Product preview

Replay

What it is

Replay is a desktop application from Weights, the AI voice platform, built for musicians and producers. It runs offline on your own computer, with downloads for Mac, Windows and Linux, and is free to use. Its main workflows include separating a song into stems such as vocals and instruments, converting a vocal performance into a different voice using RVC voice models, and training your own voice model from recordings. Models can come from the Weights library or other sources. Because processing happens on-device, audio does not need to be uploaded to a server. Using Replay requires a Weights account.

Key features

Stem separation for vocals and instruments
Vocal conversion with RVC voice models
Train your own voice model
Offline, on-device processing

What we liked

+ Free to use
+ Runs locally on Mac, Windows and Linux
+ No need to upload audio to the cloud

Worth noting

· Performance depends on your computer's hardware
· Requires a Weights account

Editor's notes

More context from AI BirdView

Replay brings several AI audio tasks that usually live on separate websites into one free desktop app. Producers can split a track into stems, swap the voice on a vocal and train a custom voice model without paying per minute or uploading files to a server. Running locally also means you can keep working with large projects without waiting on an online queue.

Separates songs into stems such as vocals and instrumentals for remixing or practice
Converts vocals using RVC voice models, either from Weights or from other sources
Lets you train your own voice model from your recordings
Processes audio on-device, with builds for Mac, Windows and Linux
Free to use, with ongoing version updates from Weights

Replay is best for home producers, cover artists and hobbyists experimenting with AI vocals who have a reasonably capable computer. A practical tip: train or use voice models only from voices you own or have permission to use, and start with short clean vocal stems, since well-separated, dry vocals give voice conversion far better results.