Local Models
Whisper and Parakeet models run entirely on your Mac. No internet required, complete privacy.
Transcription Overview
FloWords uses state-of-the-art AI models to convert your voice into text. All processing happens locally on your Mac, ensuring complete privacy.
How It Works
Section titled “How It Works”When you speak, FloWords:
- Captures audio from your selected microphone
- Processes it locally using AI models on your Mac
- Transcribes to text with high accuracy
- Pastes automatically where your cursor is
Transcription Methods
Section titled “Transcription Methods”FloWords offers multiple transcription approaches:
Cloud Providers
Optional cloud services for enhanced accuracy or speed. Requires API keys and internet.
Local vs Cloud
Section titled “Local vs Cloud”| Aspect | Local Models | Cloud Providers |
|---|---|---|
| Privacy | 100% private | Data sent to servers |
| Internet | Not required | Required |
| Cost | Free (included) | Pay per usage |
| Speed | Depends on Mac | Generally faster |
| Accuracy | Very good to excellent | Excellent |
Supported Local Models
Section titled “Supported Local Models”Whisper Models
Section titled “Whisper Models”OpenAI’s Whisper models, optimized for Apple Silicon:
| Model | RAM | Speed | Accuracy |
|---|---|---|---|
| Tiny | ~1 GB | Fastest | Good |
| Base | ~1.5 GB | Fast | Better |
| Small | ~2 GB | Medium | Great |
| Medium | ~5 GB | Slower | Excellent |
| Large | ~10 GB | Slowest | Best |
Parakeet Models
Section titled “Parakeet Models”NVIDIA’s Parakeet models via FluidAudio:
- Optimized for real-time transcription
- Excellent accuracy for English
- Lower memory requirements
Supported Cloud Providers
Section titled “Supported Cloud Providers”For users who prefer cloud transcription:
- OpenAI Whisper API - High accuracy, reliable
- Groq - Ultra-fast transcription
- Deepgram - Real-time streaming
- Google Gemini - Multimodal capabilities
- ElevenLabs - Speech recognition
- Mistral - European AI provider
Fallback System
Section titled “Fallback System”FloWords includes an intelligent fallback system:
- Primary: Your selected model (local or cloud)
- Secondary: Alternative model if primary fails
- Tertiary: Apple’s native Speech Recognition
This ensures you always get a transcription, even if your preferred method encounters issues.
Audio Sources
Section titled “Audio Sources”FloWords can transcribe from:
- Live microphone - Real-time dictation
- Audio files - WAV, MP3, M4A, AAC, FLAC, AIFF, CAF
- Video files - MP4, MOV (extracts audio)
Language Support
Section titled “Language Support”Whisper models support 99+ languages including:
- English, Spanish, French, German
- Chinese, Japanese, Korean
- Arabic, Hindi, Portuguese
- And many more…
Set your language in Settings > Model > Language or enable auto-detection.
Next Steps
Section titled “Next Steps”- Learn about Local Models for privacy-first transcription
- Explore Cloud Providers for enhanced options
- Transcribe Audio Files for existing recordings