title: Neural Prediction Settings
description: Configure AI-powered swipe predictions
category: Settings
difficulty: advanced
Neural Prediction Settings
Fine-tune the neural network-based swipe typing predictions for accuracy and performance.
Quick Summary
| What | Description |
|---|
| Purpose | Optimize neural predictions |
| Access | Scroll to Neural Prediction section in Settings |
| Key Setting | Beam Width controls accuracy vs speed |
Settings Location
In Settings, scroll to the Neural Prediction section (collapsible). All neural prediction settings are here.
Understanding Neural Predictions
CleverKeys uses an ONNX neural network to predict words from swipe gestures. The model processes your swipe trajectory and outputs probability distributions for each letter position.
Key Settings
Beam Width
The most important setting for prediction quality. Controls how many parallel word candidates the decoder tracks:
| Width | Effect |
| 3-4 | Faster, may miss less common words |
| 6 | Balanced (default) |
| 8-10 | More thorough search, finds rare words |
[!NOTE]
Higher beam width = more word candidates explored = better accuracy but slightly slower.
Show Suggestions
Toggle the prediction/suggestion bar above the keyboard:
| Setting | Effect |
| On | Shows word predictions above keyboard |
| Off | Hides prediction bar (more screen space) |
Autocorrect from Beam
When enabled, uses neural predictions for autocorrection:
| Setting | Effect |
| On | Neural model helps with tap-typing corrections |
| Off | Only dictionary-based autocorrect |
Advanced Settings
Confidence Threshold
Minimum score for a prediction to be shown (default: 0.01):
- • Lower = more suggestions, some may be weak
- • Higher = only confident predictions shown
Frequency Weight
How much word frequency influences scoring (default: 0.57):
| Weight | Effect |
| 0.0 | Neural model only |
| 0.5 | Balanced |
| 1.0+ | Favor common words heavily |
Length Penalty (Beam Alpha)
Normalizes scores by word length (default: 1.0):
- • Lower = favors shorter words
- • Higher = allows longer words to compete
Neural Profiles
CleverKeys includes preset profiles optimizing for different use cases:
| Profile | Beam Width | Use Case |
| Speed | 4 | Fast typing, acceptable accuracy |
| Balanced | 6 | Default experience |
| Accuracy | 8+ | Best predictions, slower |
Access via the Neural Prediction section settings.
Tips and Tricks
- • Accuracy issues: Increase beam width from 6 to 8 or 10
- • Slow predictions: Reduce beam width to 4-5
- • Missing words: Check that multi-language is configured correctly
- • Long words wrong: Increase beam width or adjust beam alpha
[!TIP]
Start with the default beam width of 6. Only adjust if you notice specific issues.
All Neural Settings
| Setting | Default | Range/Options |
| Beam Width | 6 | 3-10 |
| Show Suggestions | On | On/Off |
| Confidence Threshold | 0.01 | 0.01-0.5 |
| Frequency Weight | 0.57 | 0.0-2.0 |
| Beam Alpha | 1.0 | 0.5-2.0 |
| Beam Autocorrect | On | On/Off |
Common Questions
Q: Why are predictions slow?
A: Reduce beam width from 6 to 4-5. Also check if multiple languages are enabled, which increases processing.
Q: Why does it suggest wrong words?
A: Try increasing beam width for more thorough search. Also ensure your primary language is set correctly in Multi-Language section.
Q: How do I add words to the dictionary?
A: Type the word and tap it in predictions, or long-press a word and select "Add to dictionary" if supported.
Q: Can I reset neural settings?
A: Go to Settings and look for reset options, or use Backup & Restore to reset to defaults.
Technical Details
The neural model is:
- • Format: ONNX (Open Neural Network Exchange)
- • Architecture: Encoder-only transformer
- • Input: Normalized swipe coordinates (x, y, time)
- • Output: Per-position letter probabilities
- • Decoding: Beam search with vocabulary constraint
Related Features
- • Swipe Typing - How to swipe type
- • Autocorrect - Text correction
- • Multi-Language - Multi-language predictions