Free · in your browser · nothing uploaded
Change your voice live, right in the browser
An AI voice changer for Discord, games and streams. Pick a voice and talk: they hear the new voice, live. Use a built-in voice or any RVC voice model, with nothing to install.
Guide
Use the voice in Discord, OBS and games
Apps can't hear a browser tab directly, so the new voice goes through a virtual audio cable: a pretend speaker that other apps can use as a microphone.
1. Install a virtual cable
On Windows and macOS, install the free VB-Cable and restart your browser. On Linux, create a virtual sink with PipeWire or PulseAudio (for example pactl load-module module-null-sink sink_name=voice).
2. Send the new voice to the cable
Above, set New voice goes to to CABLE Input. Turn on Hear yourself and pick your headphones to hear the new voice too.
3. Use the cable as your microphone
- Discord: User Settings → Voice & Video → Input Device → CABLE Output.
- OBS: add an Audio Input Capture source with CABLE Output, and mute your usual mic.
- Games and other apps: choose CABLE Output as the microphone in their audio settings.
Where to find voices
RVC voice models, from RVC itself or from Applio, are shared on sites like Hugging Face and in AI voice communities, usually as a .zip with a .pth file and an .index file. Drop the whole .zip in. Check each voice's terms, and don't use anyone's voice to deceive people.
Questions
Common questions
Is it really free? +
Yes. The AI runs on your own computer, so there's no server to pay for and nothing to sign up for.
Is my voice uploaded anywhere? +
No. Your microphone, your recordings and the voices you add stay in your browser. The page downloads the AI models once and works on your device from then on.
Why is there a delay? +
The AI needs a short stretch of your speech to turn it into another voice, so every AI voice changer adds a delay. Here it's about a third of a second. A shorter delay setting feels more natural; a longer one is smoother on slower computers.
What computer do I need? +
A computer with a reasonably recent graphics card, and Chrome or Edge. The app uses the graphics card through WebGPU. If it can't keep up, it tells you; choose a longer delay then.
Which voice files work? +
RVC voice models, v1 and v2, at 24 to 48 kHz, as a .pth file or a .zip holding one, including voices made with Applio's MRF HiFi-GAN and RefineGAN vocoders. Add the .index file too if the voice has one. Voices trained on a speech encoder other than the standard ContentVec (like SPIN or a Japanese HuBERT) aren't supported.
It sounds robotic or out of tune. What helps? +
Match the pitch to the voice: about +12 for a man using a woman's voice, about −12 the other way. Speak clearly, close to the microphone, and raise the silence level if background noise turns into strange sounds.
Does it work on phones? +
Live voice changing needs a computer. Converting a recording may work on a recent phone, slowly.