Model Issues
Troubleshoot speech recognition model download and loading problems
Yapper uses AI models for speech recognition that need to be downloaded and loaded before use. Yapper supports two engines — Parakeet (NVIDIA) and Whisper (OpenAI) — each with their own models. This guide helps you troubleshoot model-related issues.
Model Download Issues
Download Won’t Start
If clicking a model doesn’t start the download:
- Check your internet connection — Models are downloaded from Hugging Face
- Check available storage — Models need 75 MB to 3 GB depending on size
- Try a smaller model first — If Large models won’t download, try Small or Base
Canceling a Download
If a download is taking too long or you changed your mind, click the cancel button next to the progress bar to stop the download. You can restart it later by selecting the model again.
Download Is Stuck
If the progress bar isn’t moving:
- Cancel the download and try again
- Check your internet connection stability
- Close and reopen Yapper Settings
- Select a different model, then select your original choice again
- Try during off-peak hours (download servers can be busy)
Download Failed Error
If you see a download error:
- Verify internet connection — Open a website to confirm connectivity
- Check firewall/VPN — Some networks block Hugging Face; try disabling VPN
- Free up disk space — Ensure you have enough space (check About This Mac → Storage)
- Try a different model — If one fails, another may work
Partial Download
If a model seems partially downloaded or corrupted:
- Go to Settings → Transcription and click Clear Downloaded Models to remove all models cleanly
- Restart Yapper
- Download the model again
Alternatively, you can manually delete model folders:
- Whisper models:
~/Library/Application Support/Yapper/models/ - Parakeet models:
~/Library/Application Support/Yapper/parakeet-models/
Model Loading Issues
”Model not loaded” Error
This means the model hasn’t finished loading into memory.
Solution:
- Open Settings → Transcription
- Check the model status
- Wait for loading to complete (progress indicator shown)
- If stuck, try selecting a different model and then back
Model Takes Forever to Load
First loads are slower; subsequent loads should be faster.
If loading is consistently slow:
- Try a smaller model — Large v3 takes longer than Small or Base
- Close other apps — Free up RAM for model loading
- Restart Yapper — Sometimes helps clear cached state
- Restart your Mac — Clears memory fragmentation
Loading Fails Repeatedly
If the model fails to load:
- Check Activity Monitor for available memory
- Close memory-intensive applications
- Try the Tiny or Base model first
- If even small models fail, restart your Mac
Model Performance Issues
Transcription Is Very Slow
If transcription takes much longer than expected:
| Symptom | Likely Cause | Solution |
|---|---|---|
| Processing takes minutes | Wrong model size for your Mac | Try a smaller model |
| First transcription slow | Model warming up | Subsequent transcriptions faster |
| Consistently slow | Insufficient resources | Close other apps, try smaller model |
Model Using Too Much Memory
If Yapper is using excessive RAM:
- Check Activity Monitor for Yapper’s memory usage
- Larger models use more memory:
- Parakeet TDT v3/v2: ~800 MB
- Tiny: ~150 MB
- Base: ~250 MB
- Small: ~500 MB
- Large v3 Turbo: ~1.5 GB
- Large v3: ~3 GB
- Switch to a smaller model if needed
Model Using Too Much CPU
High CPU usage during transcription is normal. If CPU stays high when idle:
- Quit and reopen Yapper
- Check if the model is stuck loading
- Try a different model
Storage Issues
Not Enough Disk Space
Models require significant storage:
| Model | Space Needed |
|---|---|
| Parakeet TDT v3/v2 | ~800 MB |
| Tiny | ~150 MB |
| Base | ~250 MB |
| Small | ~500 MB |
| Large v3 Turbo | ~1.5 GB |
| Large v3 | ~3 GB |
To free space:
- Use Clear Downloaded Models in Settings → Transcription to remove all downloaded models at once
- Or delete unused models manually:
- Whisper models:
~/Library/Application Support/Yapper/models/ - Parakeet models:
~/Library/Application Support/Yapper/parakeet-models/
- Whisper models:
- Clear other files (Downloads, Trash, etc.)
- Consider using a smaller model
Model Storage Location
Models are stored locally at:
- Whisper models:
~/Library/Application Support/Yapper/models/ - Parakeet models:
~/Library/Application Support/Yapper/parakeet-models/
You can see how much disk space your models are using in Settings → Transcription.
To access these folders in Finder:
- Open Finder
- Press ⌘ Shift G
- Paste the path and press Enter
Switching Models
Best Practices for Changing Models
- Wait for current recording/processing to complete
- Go to Settings → Transcription
- Select the new model
- Wait for download (if needed) and loading
- Test with a short recording
Model Recommendations
| Your Priority | Recommended Model |
|---|---|
| Best balance (recommended) | Parakeet TDT v3 |
| English-only, top accuracy | Parakeet TDT v2 |
| 100+ languages | Large v3 Turbo |
| Maximum accuracy | Large v3 |
| Streaming transcription | Large v3 Turbo |
| Good accuracy, faster | Small |
| Fastest results | Base |
| Minimal resources | Tiny |
Clearing Model Cache
If models are behaving unexpectedly:
- Open Settings → Transcription and click Clear Downloaded Models
- Or manually: Quit Yapper, delete the contents of the model folders (see storage locations above)
- Reopen Yapper
- Re-download your preferred model
Still Having Issues?
If model problems persist:
- Check macOS version — Yapper requires macOS 14+
- Update macOS — Some issues are fixed in updates
- Contact support — Email hello@yapper.to with:
- Your Mac model (Apple menu → About This Mac)
- Which model you’re trying to use
- The specific error or behavior
Related Topics
- Transcription Settings — Model selection
- System Requirements — Hardware requirements
- Recording Issues — Other recording problems