Skip to main content

​
Common Issues and Solutions

This guide covers solutions to frequently encountered issues with SlasshyWispr. If you don’t find your issue here, please visit the GitHub issues page.
SlasshyWispr is in strong beta. Expect bugs, rough edges, and breaking changes while features are actively evolving. Always check for updates in Settings > Update and Security.

​
Audio and Microphone Problems

​
Microphone Not Detected

Symptoms: Empty microphone dropdown or missing devicesSolutions:
1

Check System Permissions

Ensure SlasshyWispr has microphone permissions in your operating system:
  • Windows: Settings > Privacy > Microphone > Allow apps to access microphone
  • Verify SlasshyWispr is in the allowed apps list
2

Verify Hardware Connection

  • Check that your microphone is physically connected
  • Try unplugging and reconnecting USB microphones
  • Test microphone in Windows Sound settings
3

Restart Application

Close and reopen SlasshyWispr after connecting microphone or granting permissions.
4

Check Default Device

Set your preferred microphone as the default recording device in Windows Sound settings.

​
No Audio Input Detected

Symptoms: Recording starts but no voice detected, transcription is emptySolutions:
1

Test Microphone Levels

  • Open Windows Sound Settings
  • Go to Input > Device properties
  • Speak and watch the input level meter
  • Adjust input volume if too low
2

Select Correct Device

In Settings > General, verify the correct microphone is selected in the Microphone Device dropdown.
3

Check Capture Mode

  • Verify your capture mode setting
  • Try switching between “push-to-talk” and “single-tap”
  • For push-to-talk, ensure you’re holding the hotkey while speaking
4

Disable Exclusive Mode

In Windows microphone properties:
  • Right-click microphone > Properties
  • Advanced tab
  • Uncheck “Allow applications to take exclusive control”

​
Poor Transcription Quality

Symptoms: Incorrect words, gibberish, missing phrasesSolutions:
1

Improve Audio Quality

  • Reduce background noise (close windows, turn off fans, mute music)
  • Use a higher-quality microphone (headset mic or USB mic preferred)
  • Maintain consistent distance from microphone (6-12 inches for desktop mics)
  • Speak clearly at normal pace and volume
2

Try Better Models

  • For online mode: Use higher-quality STT models if available from your provider
  • For local mode: Try larger models:
    • Switch from Moonshine to Parakeet v3
    • Switch from Parakeet to Whisper Medium/Large
    • Check Settings > Offline STT for recommendations
3

Configure Language Settings

In Settings > General:
  • Set correct Dictation Language for your spoken language
  • For multi-language use, configure Dictation Language Mode
4

Enable Processing Features

Toggle these settings to improve output:
  • Auto Punctuation: Automatically add punctuation
  • Remove Fillers: Filter “um”, “uh”, etc.
  • Check Settings > General for these options

​
Recording Cuts Off

Symptoms: Recording ends before you finish speakingSolutions:
  • Maximum Duration: Recordings are limited to 45 seconds (MAX_RECORDING_MS: 45_000)
  • For longer dictation, break into multiple sessions
  • In push-to-talk mode, ensure you’re continuously holding the hotkey
  • Check if “Mute Music While Dictating” setting is interfering with detection
If you consistently need recordings longer than 45 seconds, consider breaking your dictation into logical segments or phrases.

​
Model Loading Issues

​
Local STT Model Won’t Download

Symptoms: Download hangs, errors during download, incomplete modelsSolutions:
1

Check Internet Connection

Ensure stable internet connection. Large models (Whisper Large: 1.6GB) require reliable downloads.
2

Verify Disk Space

Check available disk space:
  • Parakeet models: ~500 MB
  • Whisper models: 500 MB - 1.6 GB
  • Ensure 2-3x the model size for extraction and processing
3

Check Download Status

Navigate to Settings > Offline STT to view download progress and any error messages.
4

Retry Download

  • Cancel stuck download if option available
  • Close and reopen SlasshyWispr
  • Attempt download again
5

Try Smaller Model First

Test with smallest model (Moonshine Base: 58 MB) to verify download functionality works.

​
Local STT Model Won’t Load

Symptoms: “Model failed to load”, errors during warmup, crashesSolutions:
1

Check Hardware Requirements

Navigate to Settings > Offline STT and check hardware recommendations:
  • Review your detected CPU, RAM, and GPU
  • Check “Performance Tier” assessment
  • Verify selected model matches recommendations
  • Look for “Selected Model Warning” if your hardware is underpowered
2

Use Recommended Model

The hardware advisor suggests models based on your system:
  • High-end GPU: Whisper Large, Parakeet v3
  • Mid-range system: Parakeet v3, Whisper Medium
  • Lower-spec system: Moonshine Base, SenseVoice
Check the “Slasshy Suggestion” in hardware advisor.
3

Verify Model Files

  • Use “Open Path” option in model settings to view model directory
  • Ensure all model files are present and not corrupted
  • Try deleting and re-downloading if files seem incomplete
4

Check Runtime State

Monitor the runtime state response:
  • loaded: Should be true when model is ready
  • daemonCount: Background processes running
  • details: Error messages if loading failed
Some models require significant RAM and VRAM. Whisper Large models may need 4-8GB RAM and benefit from GPU acceleration. Check hardware advisor before downloading large models.

​
Ollama Connection Failed

Symptoms: “Ollama not running”, connection errors, model pull failsSolutions:
1

Verify Ollama Installation

Check Settings > Offline AI for Ollama status:
  • installed: Should be true
  • running: Should be true
  • version: Should show Ollama version
2

Start Ollama Service

  • Ensure Ollama is running (check system tray or task manager)
  • Try restarting Ollama service
  • On Windows, run Ollama from Start menu
3

Check Base URL

In Settings > Offline AI, verify Local Ollama Base URL:
  • Default: http://127.0.0.1:11434
  • Ensure no typos
  • If Ollama runs on different port, update accordingly
4

Test Connection

Try pulling a small model to test connection:
  • Use model pull feature in settings
  • Start with a small model like “llama3.2:1b”
  • Check for error messages in pull response
5

Firewall and Network

  • Check Windows Firewall isn’t blocking localhost connections
  • Ensure no VPN or proxy interfering with local connections
  • Try disabling antivirus temporarily to test

​
API Connection Errors

​
Online API Authentication Failed

Symptoms: “Invalid API key”, “Unauthorized”, 401 errorsSolutions:
1

Verify API Key

In Settings > Online:
  • Double-check API key for typos
  • Ensure no extra spaces before/after key
  • Verify key hasn’t expired with your provider
  • Test key validity on provider’s website
2

Check API Base URL

  • Ensure correct base URL for your provider
  • Common formats:
    • OpenAI: https://api.openai.com/v1
    • Custom providers: Verify correct endpoint
  • Remove trailing slashes if present
3

Re-enter Credentials

  • Clear existing API key
  • Re-enter fresh copy from provider dashboard
  • Ensure “Remember API Key” is checked if you want persistence
4

Check Account Status

  • Verify API account is active and paid (if required)
  • Check for billing issues with provider
  • Ensure usage limits haven’t been exceeded

​
Online API Timeout or Network Errors

Symptoms: Long waits, timeout errors, “Failed to connect” messagesSolutions:
1

Check Internet Connection

  • Verify stable internet connection
  • Test accessing provider’s website in browser
  • Run speed test to check bandwidth
2

Try Different Network

  • Switch between WiFi and wired connection
  • Disable VPN if active (may interfere)
  • Try mobile hotspot to rule out network restrictions
3

Check Provider Status

  • Visit provider’s status page
  • Check for ongoing outages or maintenance
  • Review provider’s rate limits
4

Switch to Local Models

If online connectivity is unreliable:
  • Use local STT models (offline mode)
  • Use local Ollama for AI (offline mode)
  • See Hybrid Mode guide for configuration

​
Model Not Found

Symptoms: “Model not found”, invalid model name errorsSolutions:
1

Verify Model Name

  • Check exact model name with your provider
  • Ensure correct spelling and capitalization
  • Some providers use specific naming formats
2

Check Available Models

Use provider models feature if available:
  • Settings > Online may show available models
  • Consult provider’s documentation for supported models
3

Update Model Names

  • STT Model Name: e.g., “whisper-1”
  • AI Model Name: e.g., “gpt-4”, “gpt-3.5-turbo”
  • For local Ollama: Verify model is pulled (e.g., “llama3.2”)

​
Performance Optimization

​
Slow Response Times

Symptoms: High latency, waiting several seconds for transcription or AI responseSolutions:
1

Check Pipeline Latency

Navigate to Settings > Pipeline to view timing breakdown:
  • sttLatencyMs: Speech-to-text processing time
  • aiLatencyMs: AI model response time
  • ttsLatencyMs: Text-to-speech generation time
  • Identify which component is slowest
2

Optimize STT Performance

For local STT:
  • Use smaller/faster models (Moonshine Base, Parakeet v3)
  • Enable GPU acceleration if available
  • Warm up models before intensive use
  • Close other applications to free RAM
For online STT:
  • Check internet speed
  • Try different API provider if consistently slow
3

Optimize AI Performance

For local AI (Ollama):
  • Use smaller models (llama3.2:1b vs llama3.2:70b)
  • Reduce maxTokens setting to limit response length
  • Ensure Ollama has sufficient system resources
  • Check Ollama logs for performance issues
For online AI:
  • Use faster models from provider
  • Reduce maxTokens (default: 320)
  • Simplify systemPrompt if very long
4

Optimize TTS Performance

In Settings > TTS:
  • Use Piper engine (faster than Coqui)
  • Set Piper Quality to “fast” instead of “high”
  • Disable TTS if you don’t need voice responses
5

System Optimization

  • Close unnecessary background applications
  • Ensure adequate free RAM (4GB+ recommended)
  • Check CPU usage in Task Manager
  • Consider upgrading hardware for local models

​
High CPU or Memory Usage

Symptoms: High CPU%, high RAM usage, system slowdown, fan noiseSolutions:
1

Identify Resource-Intensive Components

  • Local STT models use significant RAM (1-4GB) when loaded
  • Local AI (Ollama) can use 4-16GB depending on model size
  • Check if multiple models are loaded simultaneously
2

Use Lighter Models

  • STT: Switch from Whisper Large to Moonshine Base or SenseVoice
  • AI: Use smaller Ollama models (1-3B parameters instead of 7-70B)
  • Check hardware advisor recommendations
3

Deactivate Unused Models

  • Use model deactivation feature in settings
  • Unload STT models when not in use
  • Stop Ollama when not needed for local AI
4

Switch to Online Mode

For lower-spec systems:
  • Use online STT and AI to offload processing
  • This moves resource usage to cloud providers
  • See Hybrid Mode guide for configuration
Local models trade system resources for privacy and offline capability. If your system struggles, consider hybrid mode with online components.

​
Application Crashes or Freezes

Symptoms: Application hangs, “Not Responding”, sudden crashesSolutions:
1

Check System Requirements

Verify your system meets minimum requirements:
  • RAM: 8GB+ recommended for local models
  • CPU: Multi-core processor recommended
  • GPU: NVIDIA GPU helpful for large models
  • Storage: Sufficient space for models and temp files
2

Update Application

  • Check Settings > Update and Security
  • Install latest version
  • Bug fixes and stability improvements in newer releases
3

Reduce Concurrent Operations

  • Don’t trigger multiple voice commands simultaneously
  • Wait for previous operation to complete
  • Disable wake word if causing issues
4

Clear and Reset

  • Deactivate all local models
  • Restart application
  • Reload models one at a time
  • Test with minimal configuration first
5

Check for Conflicts

  • Disable antivirus temporarily to test
  • Close other voice/audio applications
  • Check Windows Event Viewer for crash details

​
Feature-Specific Issues

​
Wake Word Not Working

See: Wake Word Setup Guide for detailed troubleshootingQuick Solutions:
  • Verify wakeWordEnabled: true in Settings > General
  • Check microphone is selected and working
  • Speak wake phrase clearly (“Hey [AssistantName]”)
  • Test with different wake phrase patterns
  • Ensure command mode is enabled

​
Context Awareness Not Detecting Selection

See: Context Awareness Guide for detailed troubleshootingQuick Solutions:
  • Enable contextAwareness: true in Settings > General
  • Ensure commandMode: true
  • Select text before activating voice command
  • Try copying selection manually (Ctrl+C) before command
  • Verify application allows clipboard access

​
Hotkeys Not Responding

Symptoms: Pressing hotkey has no effectSolutions:
1

Check Hotkey Configuration

In Settings > General:
  • Verify Push to Talk Hotkey (default: Ctrl+Space)
  • Verify Command Hotkey (default: Ctrl+Shift+Space)
  • Ensure hotkeys don’t conflict with other applications
2

Test Different Hotkeys

  • Try changing to different key combination
  • Avoid combinations used by Windows or other apps
  • Use modifier keys (Ctrl, Shift, Alt) for uniqueness
3

Check Application State

  • Ensure SlasshyWispr is running (check system tray)
  • Verify app is not in error state
  • Try restarting application
4

Administrator Privileges

Some applications block hotkeys from apps without admin rights:
  • Try running SlasshyWispr as administrator
  • Check if issue only occurs in specific applications

​
TTS Not Playing Audio

Symptoms: Assistant processes command but doesn’t speak responseSolutions:
1

Check TTS Engine

In Settings > TTS:
  • Verify TTS engine is selected (Piper or Coqui)
  • Default engine is Piper
  • Coqui may be disabled in zero-Python mode
2

Verify Piper Installation

  • Check Piper path in settings
  • Ensure Piper executable is accessible
  • Try reinstalling or re-setting up TTS runtime
3

Check Audio Output

  • Ensure system volume is not muted
  • Check default audio output device in Windows
  • Test audio with other applications
4

Review TTS Settings

  • Piper Speed, Quality, and Emotion settings
  • Try changing to different voice or speed
  • Reset to defaults if customized
If using Coqui TTS, ensure Python path is correctly configured and TTS dependencies are installed. Piper is recommended for most users.

​
Getting Support

​
Reporting Issues

If you can’t resolve your issue with this guide:
1

Gather Information

Before reporting, collect:
  • SlasshyWispr version (check Settings > Update and Security)
  • Windows version
  • Hardware specs (CPU, RAM, GPU)
  • Relevant settings (runtime mode, models, etc.)
  • Steps to reproduce the issue
  • Error messages or screenshots
2

Search Existing Issues

Check if your issue is already reported:
  • Visit GitHub Issues
  • Search for keywords related to your problem
  • Review open and closed issues
3

Create New Issue

If no existing issue matches:
  • Click “New Issue” on GitHub
  • Use descriptive title
  • Provide detailed description with:
    • What you expected to happen
    • What actually happened
    • Steps to reproduce
    • System information
    • Screenshots or logs if applicable

​
Community Support

Contributions are accepted and welcome! If you fix an issue or improve documentation, please open a pull request on GitHub.

​
Update and Maintenance

1

Check for Updates Regularly

Navigate to Settings > Update and Security to check for new versions.
2

Review Release Notes

When updates are available, review release notes for:
  • Bug fixes that may resolve your issue
  • New features or improvements
  • Breaking changes that may affect your configuration
3

Install Updates

  • Use in-app update feature
  • Or download manually from Releases page
Always backup your settings and data before major updates. While settings are typically preserved, it’s safer to have a backup.

​
Additional Resources

This troubleshooting guide is based on SlasshyWispr’s beta version. Some issues may be resolved in future updates. Always use the latest version for the best experience.

Build docs developers (and LLMs) love

Get started for freeTalk to us