Troubleshooting Guide

Common Issues

Chat Not Working

1. Check your internet connection
2. Verify API keys are correctly configured in Settings > LLM / OpenRouter
3. Check that at least one provider is enabled (Groq checkbox or OpenRouter key)
4. Try clicking "Go Online" in the chat header
5. Check Debug Panel for error messages

Groq Provider Issues

• API Key Error: Ensure your key starts with gsk_ and is valid at console.groq.com
• Model Not Found: Check the model name is correct (e.g., llama-3.3-70b-versatile)
• Rate Limit: Free tier has rate limits; wait or switch to OpenRouter
• No Response: Verify the "Use Groq" checkbox is checked in LLM settings

OpenRouter Issues

• API Key Error: Ensure your key starts with sk-or-v1- and has sufficient credits
• Model Not Found: Model names are case-sensitive; check openrouter.ai/models
• JSON Errors: Some models don't support JSON mode; the app falls back to text parsing with emotion detection

Live2D Model Not Loading

• Verify the model URL points to a valid .model3.json file
• Check browser console for CORS errors
• Ensure the model is compatible with Cubism SDK
• Try loading a built-in model from the gallery first

Voice Not Working

• Check microphone permissions for STT
• Verify TTS is enabled in Voice Settings (TikTok, Kokoro, or Browser)
• Kokoro issues: Requires WebGPU or WASM support. Check browser compatibility. Disabled on low RAM (<4GB) or slow networks
• TikTok issues: Depends on external API availability
• Try switching between TTS providers

Lip-Sync Not Working

• Ensure TTS is enabled and producing audio
• A Live2D model must be loaded with mouth parameters (ParamMouthOpenY)
• Check that audio is actually playing (TTS volume not muted)

Connection Errors

If the AI service is temporarily unavailable, the chat shows a brief error message. Just retry in a moment.

Background Not Loading

• Verify the image URL is publicly accessible
• Check for CORS restrictions on the image host
• Try a different image source
• Use the background library for pre-tested images

Ambient Mode Not Triggering

• Verify ambient mode is enabled in Settings > Automation
• Check the trigger delay setting (default: 10 seconds)
• Ensure an ambient prompt is configured
• Ambient mode only triggers after the AI finishes responding

Model Gallery Issues

• Thumbnails load from external CDNs; check internet connection
• If a model fails to load, try another or check the browser console for errors
• Some models may have CORS restrictions on their hosting CDN

Debug Panel

The Debug Panel provides detailed logs for troubleshooting:

• Live2D logs: Model loading and rendering issues
• AI logs: API calls, responses, and provider selection
• TTS logs: Voice synthesis status and queue state
• Network logs: HTTP requests, response times, and errors
• Chat Context: View the full prompt sent to the AI (enable in Debug Settings)

Debug Settings

• Enable Debug Panel: Show/hide the debug log viewer
• Show AI Logs: Filter AI-related debug messages
• Show TTS Logs: Filter TTS-related debug messages
• Show Verbose Logs: Include detailed/verbose messages
• Show Chat Context: Log the full system prompt and conversation sent to AI
• Copy Debug Log: Copy all debug history to clipboard
• Auto-scroll: Automatically scroll to newest log entries

Access via Settings > Debug Settings.

Categorized Log Filtering

Logs are automatically categorized:

• AI completions, translations, and summarization
• TTS synthesis, queue, and preloading
• Network requests and responses
• Live2D model operations
• General application events

Getting Help

If issues persist:

1. Check the Debug Panel for specific error messages (enable verbose logging)
2. Review Settings > Help & Tutorial for the interactive guide
3. Check the GitHub issues for known problems
4. Visit waifuai.com for the latest version