Dictation that never leaves your Mac
Tidytalk always recognizes your voice on your Mac. Pair it with a local model and your text stays there too.
Updated 2026-11-05
What leaves your Mac
| Setup | Your voice | Your text |
|---|---|---|
| Cloud AI service (for example DeepSeek or OpenAI) | Stays on your Mac | Sent to the AI service you chose |
| Local model in Ollama or LM Studio | Stays on your Mac | Stays on your Mac |
| "Tidy up text" turned off | Stays on your Mac | Stays on your Mac (raw transcript) |
Speech recognition uses the SenseVoice model, which runs on your Mac after a one-time download of about 240 MB. Dictation history is kept on your Mac for 30 days; you can turn that off in Settings › Privacy & Data. Settings are stored in ~/.tidytalk.json, readable only by you, and your API key is kept in the macOS Keychain.
Apart from the AI service you set up, Tidytalk only goes online to download models, check for updates (at launch, at most once a day, which you can turn off in Settings › General), and transcribe links. Each case is listed in the privacy policy.
Set up Ollama
- Install Ollama.
- Download a model in Terminal, for example:
ollama pull qwen2.5:7b - In Tidytalk, open Settings › AI Service, and choose Ollama at the top. Check that the model name matches the one you pulled, then click the checkmark button to save.
- Click the purple network icon next to it to test the connection. When it says "Connected", you're done.
Set up LM Studio
- Install LM Studio, download a model, and load it.
- Start LM Studio's local server.
- In Tidytalk, choose LM Studio at the top of Settings › AI Service, enter the name of the loaded model, save, and click the purple network icon to test the connection.
Local models don't need an API key, and Tidytalk gives them a 60-second timeout instead of 12 seconds.
Choosing a model
- Tidying is a light task, but the model has to follow instructions: tidy what you said instead of answering it. Small models are less reliable at this and at following custom rules.
- A model around 7B parameters or larger is a good starting point.
qwen2.5:7bis about a 5 GB download and handles both Chinese and English well. - If the local model is slow or returns something unexpected, Tidytalk pastes the raw transcript, so you never lose what you said.
No AI at all
Turn off "Tidy up text" in Settings › Dictation and Tidytalk becomes a fast, fully offline speech-to-text tool. Replacements, snippets, and Chinese–English spacing still work, because they run on your Mac without AI.
Verify it yourself
Tidytalk is open source under the MIT License. You can read exactly what it sends and when on GitHub, or build it from source.