Troubleshooting

AI Character Card Troubleshooting

Find the closest symptom, check the simplest causes first, then choose a safe fallback before changing your card or app setup.

Browser APIs, model availability, and app behavior can change. Use these pages as a practical checklist and also check the current app status or official documentation when the result depends on your device, browser, API project, or selected model.

Gemini API

Gemini API failures

Use this when card generation fails, the response is empty, JSON cannot be parsed, or fallback output appears instead.

WebLLM mobile

WebLLM does not run on mobile

Check secure context, WebGPU availability, browser support, memory pressure, and model size.

PNG metadata

PNG metadata is not detected

Diagnose missing embedded card data, recompressed images, format mismatches, and manual recovery options.

Start with the symptom

Generation fails: check Gemini API setup, model settings, quota, network state, JSON parsing, and fallback behavior.
Local AI does not load: check HTTPS or trusted local URLs, WebLLM spelling, WebGPU support, browser compatibility, and device memory.
PNG imports as an image only: check whether the original PNG still contains character card metadata and whether another app recompressed it.

Keep the workflow intact

These pages do not replace the existing apps, guides, examples, or articles. They help you return to the right place after a problem is identified.