Small models, big voice.
Have a text read aloud or transcribe a recording, for free, right inside your browser. Nothing leaves your device: no account, no limits.
This sentence is read by the voice built into your browser: 0 MB, instant. MiniSpeech’s own models are just below.
Engine
1.00×
Transcription is coming soon. Speech synthesis is ready now, in the other tab.
Your transcript will appear here, with timestamps.
The model is downloaded once, then kept on your device.
It runs on your graphics card, or on your processor if need be.
Your texts and recordings are never sent anywhere.
Your device…Diagnostics →
The right model for your language
You pick a language, MiniSpeech suggests the model that speaks it best. The full catalogue stays backstage.
| Language | Text → speech | Speech → text |
|---|---|---|
| English | Kokoro 82M · 93 MB available | Moonshine tiny · 55 MB soon |
| Français | Supertonic 3 · ≈ 400 MB soon | Whisper tiny · 66 MB soon |
| Deutsch | Supertonic 3 · ≈ 400 MB soon | Whisper tiny · 66 MB soon |
| Italiano | Supertonic 3 · ≈ 400 MB soon | Whisper tiny · 66 MB soon |
| Español | Supertonic 3 · ≈ 400 MB soon | Whisper tiny · 66 MB soon |
| Português | Supertonic 3 · ≈ 400 MB soon | Whisper tiny · 66 MB soon |
| Nederlands | Supertonic 3 · ≈ 400 MB soon | Whisper tiny · 66 MB soon |
| Polski | Supertonic 3 · ≈ 400 MB soon | Whisper tiny · 66 MB soon |
| Türkçe | Supertonic 3 · ≈ 400 MB soon | Whisper tiny · 66 MB soon |
| Українська | Supertonic 3 · ≈ 400 MB soon | Moonshine (uk) · ≈ 110 MB planned |
| Русский | Supertonic 3 · ≈ 400 MB soon | Whisper tiny · 66 MB soon |
| العربية | Supertonic 3 · ≈ 400 MB soon | Moonshine (ar) · ≈ 110 MB planned |
| हिन्दी | Supertonic 3 · ≈ 400 MB soon | Whisper tiny · 66 MB soon |
| 日本語 | Supertonic 3 · ≈ 400 MB soon | Moonshine (ja) · ≈ 110 MB planned |
| 한국어 | Supertonic 3 · ≈ 400 MB soon | Moonshine (ko) · ≈ 110 MB planned |
| 中文 | MOSS-TTS-Nano planned | Moonshine (zh) · ≈ 110 MB planned |
| Tiếng Việt | Supertonic 3 · ≈ 400 MB soon | Moonshine (vi) · ≈ 110 MB planned |
| Shqip | Llafazani under discussion | Whisper tiny · 66 MB soon |
The models
All open source, all small enough for a browser. Sizes are what actually gets downloaded.
| Model | Type | Size | Languages | License | Status |
|---|---|---|---|---|---|
| Kokoro 82M | text → speech | 93 MB | en | Apache-2.0 | available |
| Kitten TTS nano | text → speech | 28 MB | en | Apache-2.0 | soon |
| Supertonic 3 | text → speech | ≈ 400 MB · to be quantized | 31 languages | OpenRAIL-M | soon |
| Pocket TTS | text → speech | ≈ 125 MB | en · fr de it es pt | CC-BY-4.0 | planned |
| Piper | text → speech | 20–60 MB per voice | 35 languages | MIT + per voice | planned |
| MOSS-TTS-Nano | text → speech | ≈ 670 MB · to be quantized | 20 languages | Apache-2.0 | planned |
| Whisper tiny | speech → text | 66 MB | 99 languages | MIT | soon |
| Whisper base | speech → text | 139 MB | 99 languages | MIT | soon |
| Whisper small | speech → text | 252 MB | 99 languages | MIT | soon |
| Moonshine tiny | speech → text | 55 MB | en | MIT | soon |
| Moonshine base | speech → text | 67 MB | en | MIT | soon |
| Moonshine ar · zh · ja · ko · uk · vi | speech → text | ≈ 110 MB | 6 | to check | planned |
| Silero VAD | segmentation | ≈ 2 MB | — | MIT | planned |