Native Swift macOS menu bar dictation app.
- Start and stop dictation with
Cmd+Shift+Space(customizable) - Transcribe with OpenAI or Groq speech-to-text
- Type directly into the current app when Accessibility is enabled
- Automatically restore focus to the original app if you switch away during recording
- Keep custom vocabulary for names, acronyms, and product terms
Switch providers from the menu bar under Provider.
- Models:
gpt-4o-mini-transcribe(default),gpt-4o-transcribe - ~30 min/day costs about $2.7/month (mini) or $5.4/month (premium)
- Requires an OpenAI API key
- Models:
whisper-large-v3(default),whisper-large-v3-turbo - Runs the same open-source Whisper model on Groq's LPU hardware — ~200x faster than GPU
- Free tier: 2,000 requests/day, 25 MB max file size, no credit card required
- Requires a Groq API key
Both providers use the same Whisper architecture, so transcription accuracy is equivalent.
Click Change Hotkey... in the menu bar to set your preferred key combination. System shortcut conflicts are detected automatically.
API keys are stored locally in an encrypted file under ~/.talktype/keys/.
The app uses a local random master key plus machine-bound encryption.
Each provider's key is stored separately. You can also set keys via environment variables: TALKTYPE_API_KEY (OpenAI) or TALKTYPE_GROQ_API_KEY (Groq).
Use Vocabulary in the menu bar app to add words or phrases you want the model to prefer.
Current behavior is intentionally simple:
- saved terms are added to the transcription prompt
- saved terms are also used for conservative post-processing
- there is no fuzzy matching or automatic learning
- macOS 13.0+
- API key (OpenAI or Groq)
- Microphone permission
- Accessibility permission for direct typing
MIT
