Voice Mirror™Voice Studio
Private session

VOICE STUDIO NEW SESSION

Your voice, new words.

Record a short sample, then give your voice a sentence it has never said.

Runs in your browserYour audio stays in this tab.
Model: ChatterboxOpen voice-cloning model, runs in this tab.

Capture your voice

Record a short sample or use an audio file.

Read this aloud after pressing Record10–15 sec
“The quietest ideas often become the loudest changes. Today I am trying something new, one sentence at a time.”
00:00Microphone off until you press Record.
Click to start recordingThe passage above is your recording script.

Write what it should say

Type a new sentence for your cloned voice.

Longer text takes longer to render.37 / 120
Try an example

Hear the result

Generate speech and play it back.

Speech is generated locally in your browser.

Your generated audio will appear here.

Ready when you are.

Capture a voice sample to begin. Your microphone is off.

  1. CaptureWaiting…
  2. Load modelWaiting…
  3. Prepare voiceWaiting…
  4. Generate speechWaiting…

Ready to record.

Checking browser compute…
How this demo works

This demo records only after you approve microphone access. The sample stays in this tab. It is passed to an open voice model running locally in your browser; the first use downloads about 1.5 GB of model files to the browser cache. Generated audio stays in this tab unless you download it. Reloading clears the recording and generated audio. Use only your own voice or a voice you have permission to clone.

NVIDIA Parakeet transcribes speech; it cannot generate a cloned speaking voice. This demo uses Chatterbox for voice cloning and speech generation. Download progress is shown for the current model file. Voice preparation and speech generation have no reliable percentage or time estimate, so their bars show activity and elapsed time instead.