Voice

TTS, speech-to-text, hands-free continuous conversation, Whisper, Mac system speech, and Kokoro on-device or on LM Mini Home / Connect.

Answer first

Voice mode lets you speak to LM Mini and hear replies. Use native TTS, Kokoro (on-device or via Home/Connect), or Pro voices from ElevenLabs and Grok — plus hands-free continuous conversation.

Open voice settings

Settings → Voice (and Transcription where separated).

Text-to-speech

ProviderRuns onNotes
NativeThis deviceOffline, system voices
Kokoro on-devicePhone / MacNeural TTS. Download languages separately (en/es/fr/zh share one pack; de/ru are smaller Piper voices)
Kokoro on Mac / PCHome or ConnectPro (uses Remote Access pairing). Highest quality; phone plays what the computer synthesizes. Connect 1.5.2+ has the same language packs
ElevenLabsElevenLabs APIPro. Add your API key under Settings → Voice, then pick a voice
GrokxAI APIPro. Add your key under Settings → Voice, then pick a Grok voice

Adjust rate and pitch (where the provider supports them), then Test voice. Enable Auto-read responses to speak every assistant message.

A Persona can use the app-wide voice, or lock to Kokoro / ElevenLabs / Grok and a specific speaker. Group participants can do the same.

Speech-to-text

Toggles:

Keep LM Mini open while you talk: voice mode runs in the foreground.

Using voice in chat

  1. Tap the microphone in the composer
  2. Speak
  3. Review text or let auto-send fire
  4. Hear the reply if auto-read / voice surface is on

In group chat, replies can be spoken in order with each participant’s voice.

Kokoro on PC

Needs LM Mini Pro: the phone reaches Home or Connect through Remote Access pairing, the same path as remote chat. Kokoro on-device is free.

Mac (current)

  1. Install LM Mini Home and turn on Share with phone
  2. Pair the phone (Settings → Remote Access)
  3. In Voice settings on the phone, choose Kokoro (PC) / the Home path
  4. Test. If it fails, confirm Share with phone is still running (menu bar is fine)

Windows / Linux

  1. Install LM Mini Connect and enable Kokoro in its panel
  2. Download the language packs you need (Connect 1.5.2+). English / Spanish / French / Chinese share one Kokoro pack; German and Russian are smaller Piper voices. If Download hangs on Windows, get the .tar.bz2 from GitHub, extract it with 7-Zip, then Use folder (or pick the archive with Use file)
  3. On the phone, choose Kokoro (PC). Pairing must be healthy — remote voice uses the same path as chat

Permissions

Grant microphone (and speech recognition) when asked. On Mac, also check System Settings → Privacy & Security for Speech Recognition and Microphone.