Drop a file
A full track or a dry vocal. The app works out which it is and isolates the vocal when needed.
← All tools
NEURAL AI · STANDALONE APP · WINDOWS, MACOS
AI Voice
Sing in another voice without recording again
- Drop in a track — the app isolates the vocal and replaces the timbre
- Runs on your computer: no cloud and no internet after activation
- Import your own RVC models, fine-tune with plain-language controls
- One installer: no Python, no CUDA to set up
4 990 ₽one-time purchase · 1 computer · updates included
how it sounds
Listening beats reading
Hold the button to hear the original part. Let go — the same part in the chosen voice. The sound never stops and the position never resets.
How the mechanism workssynthesised live · real demos will replace this
four steps
From file to result
Pick a voice
Key is matched automatically — an octave up, down, or as is.
Listen
Compare against the original on a shared playhead, without restarting.
Take the files
The new vocal, the instrumental and a finished mix. Drag them straight into your DAW.
what's inside
Features
Vocal isolation
Lead vocal without backing vocals or reverb — a dry part ready for conversion. The instrumental keeps its original balance.
- Input: WAV, MP3, FLAC, AIFF, M4A — up to 10 minutes
- Automatic full-track / dry-vocal detection with a manual switch
Fine-tuning
Collapsed by default and labelled in plain language: you never see the network's technical parameters.
- Voice similarity
- Consonant clarity
- Breath like the original
- Pitch shift in semitones
Your own voices
Import an RVC model the way people download them. Conversion happens locally.
- .pth + .index or .zip
- A “My voices” library
- Need a voice you don't have? We train it and send you the file
Any hardware
NVIDIA, AMD, Intel or no GPU at all. The app picks where to compute and estimates the time before it starts.
- Auto / GPU / CPU switch
- Runs out of video memory — processes in smaller chunks instead of crashing
system requirements
What your computer needs
| Item | Minimum | Recommended |
|---|---|---|
| OS | Windows 10 x64 (1903+) | Windows 11 |
| macOS | 13+, Apple Silicon | M1 Pro or newer |
| GPU | Any, or none — then on the CPU, slower | NVIDIA RTX / AMD RX 6000+ / Intel Arc, 6 GB VRAM or more |
| Memory | 8 GB | 16 GB |
| Disk | 6 GB — any drive, not just the system one | 10 GB |
What the app doesn't do — so you don't find out after buying
- This is a standalone app, not a VST/AU plugin. There is no real-time mode.
- There is no voice training inside the app: we train voices ourselves and send you the model.
- No voices are bundled — you work with your own models.
- No cloud processing: everything runs on your computer.
buying
LICENCE
One licence, one computer, forever. Internet is needed only to activate. New computer? Move the licence from the app.
OFFLINE
After activation the app works offline. Every couple of weeks it quietly re-checks the licence; no network — it keeps working.
VOICE TRAINING
Need a timbre you don't have? We train a model on your voice and send you the file. Request it right from the app.
buying