🎙️

Voice Twin

Clone your voice, then talk to yourself.

🎙️ Voice Twin

The voice model is starting up (about 1–2 minutes after the VM boots). You can already record; trying out your clone works once it's ready.

✅ Your voice is ready

Reference sample

Record your voice

  1. Quiet room, no music. Background noise gets cloned too.
  2. Read the text below naturally and expressively, like you're talking to a friend. The clone copies your tone, pace and microphone.
  3. 15–20 seconds is ideal. Speak English (the voice model speaks English and Chinese).
Read this out loud

0:00

…or of just you talking (10–60 s; WAV, MP3, M4A, …).

Check your sample

I trimmed the silence and kept the best s. Listen to it, then make sure the transcript matches exactly what you said: the clone depends on it.

    Your twin

    Who am I talking to?

    More settings

    connecting…

    You haven't cloned your voice yet, so replies are text only. Clone it first →

    Tap to turn on the mic
    Self-hosted on your own Azure GPU VM · Voice: Breeze TTS 2 (open weights; research & non-commercial license) · Speech recognition: Whisper large-v3-turbo · Replies: Azure OpenAI … · The VM switches itself off after min without use