Desktop and phone
Listening to your notes
Have any note or book chapter read aloud, with a proper player and a queue.
The headphones button in a note header, or in the reader, starts reading aloud. A player bar appears with the usual transport controls, a queue, and a speed setting.
Choosing a voice
- System voices
- Where it runs: On your device, offline
- Notes: Always available. Quality depends on your OS. On Android, Settings ▸ Voice explains how to install better offline voices (Google's voice packs or RHVoice); the app ships no voice of its own on the phone.
- On-device neural (sherpa-onnx)
- Where it runs: On your machine, offline
- Notes: Desktop only. Download a voice in Settings.
- Pocket (Kyutai)
- Where it runs: On your machine, offline
- Notes: Desktop only, a 98 MB download under the on-device models. The most natural voice here: it speaks in a built-in voice or in one you record in the Voice Clone Studio, after you confirm the voice is yours or used with permission. About 0.9x real time on a modest laptop, so reading starts in a few seconds and may pause briefly near the start while the voice gets ahead.
- OpenAI-compatible speech API
- Where it runs: A server you configure
- Notes: Best quality if you already run one. On the desktop, a server that can stream its audio (the Qwen3-TTS server does) starts speaking in about a second, however long the note; streaming is used at 1x speed.
| Engine | Where it runs | Notes |
|---|---|---|
| System voices | On your device, offline | Always available. Quality depends on your OS. On Android, Settings ▸ Voice explains how to install better offline voices (Google's voice packs or RHVoice); the app ships no voice of its own on the phone. |
| On-device neural (sherpa-onnx) | On your machine, offline | Desktop only. Download a voice in Settings. |
| Pocket (Kyutai) | On your machine, offline | Desktop only, a 98 MB download under the on-device models. The most natural voice here: it speaks in a built-in voice or in one you record in the Voice Clone Studio, after you confirm the voice is yours or used with permission. About 0.9x real time on a modest laptop, so reading starts in a few seconds and may pause briefly near the start while the voice gets ahead. |
| OpenAI-compatible speech API | A server you configure | Best quality if you already run one. On the desktop, a server that can stream its audio (the Qwen3-TTS server does) starts speaking in about a second, however long the note; streaming is used at 1x speed. |
The player
- A persistent player bar so you can keep browsing while something reads.
- A queue, so a multi-chapter book plays straight through.
- Playback speed control.
- On Android, lock-screen and notification controls.