Voice cloning and RVC voice training for Windows
Clone a voice from 10 to 20 seconds of speech, or train an RVC voice from 10 to 30 minutes of clean recordings for the closest likeness. Train on your own NVIDIA GPU or for free in the cloud.
Two ways to get a voice
- 1
Instant cloning: record 10 to 20 seconds of speech and click Create voice. It's ready in under a minute and works live, in Speak and in Dub.
- 2
Trained voice on your PC: with an NVIDIA GPU (GTX 16 / RTX 20 series or newer, 4 GB or more), add your recordings, choose a quality and click Start training.
- 3
Trained voice in the cloud: no NVIDIA card needed. Connect Google Drive, upload your recordings and train on a free Google Colab GPU in your own Google account.
- 4
Already have RVC v2 voices? Import them (a folder with .pth or .onnx and .index) and use them live.
What you get
Instant cloning
No training and nothing uploaded. Clones work in 23 languages and power live conversion, text to speech and dubbing.
Closest likeness with RVC
Training makes the most faithful copy from 10 to 30 minutes of clean recordings of one person. Trained voices have the lowest live delay.
Free cloud training
Runs on a free Google Colab GPU in your own account. Mimicore can see only the files it creates, and the training notebook is open source.
Calibrate to your voice
Read one sentence and Mimicore measures your pitch and room, then sets the pitch and noise gate for the closest match.
Mix voices
Blend voices to make a character that belongs to no real person.
Private by default
Cloning, training on your PC and live conversion all run locally. Cloud training uses your own Google account, never our servers.
Cloning, mixing, importing and training are part of the Studio plan (7-day free trial). Only clone voices you have permission to use.
Questions
How long does RVC voice training take?
On your own PC, a quick test takes about 10 to 20 minutes, the recommended Good quality about 25 to 50 minutes, and Best about 1 to 2 hours. Cloud training on a free Colab GPU takes a little longer: roughly 25 to 40 minutes for a quick test, 50 to 90 minutes for Good and 1.5 to 3 hours for Best.
Can I train an RVC voice without an NVIDIA GPU?
Yes. Use cloud training. It runs on a free Google Colab GPU in your own Google account, with the same trainer and the same quality, and works on any PC.
How much audio do I need?
For instant cloning, 10 to 20 seconds. For a trained voice, 10 to 30 minutes of clean recordings of one person.
Can I import RVC voices I already have?
Yes. Import RVC v2 voices (a folder with .pth or .onnx and an .index file) that you have the rights to use.
Are my recordings uploaded to your servers?
No. For cloud training your browser uploads recordings straight to your own Google Drive. Ribanix Labs never receives them, and they are moved to your Drive trash once the voice is installed unless you choose to keep them.
More: AI voice changer for Discord · Free cloud training · AI video dubbing · Full user guide · Help and FAQ