Configuration
Speech Recognition
Configure the Whisper model, language, and recording mode.
Customize how Dial8 listens and transcribes your voice.
Model Selection
Choose your Whisper model in Settings > Speech to Text:
| Model | Description |
|---|---|
| Small | Higher accuracy with moderate processing time. Best for English. |
| Large V3 Turbo | Optimized large model with faster processing. Best balance of speed and accuracy. |
Models are downloaded on demand and stored locally. You can switch between them at any time.
Language
Select your primary language or use auto-detect. Setting a specific language improves recognition accuracy and filters out hallucinations in unrelated scripts.
Supported languages include: English, Chinese, German, Spanish, Russian, Korean, French, Japanese, Portuguese, Turkish, Polish, Italian, Vietnamese, Dutch, Persian, Arabic, and more.
Recording Mode
- Push-to-Talk — Hold the hotkey while speaking, release to transcribe
- Toggle — Press once to start, press again to stop
- Streaming — Real-time transcription with automatic pause detection
AI Text Cleaning
Enable in Settings > AI Rewrite to have AI polish your transcriptions:
- Removes filler words and false starts
- Adds proper punctuation and formatting
- Adapts tone to the active app (professional for Slack/email, casual for iMessage)
- Uses your configured AI provider (Apple Intelligence, Claude, or Groq)
Text Replacements
Configure automatic corrections in Settings > Text Replacements:
- Multiple trigger words per replacement (e.g., "Sean", "Shawn" → "Shaun")
- Case-insensitive, whole-word matching
- Enable/disable individual rules or the entire system