That's not something we can help with from the Kroma side. Kroma
connects you to whichever AI model you've configured (Gemini, Claude,
OpenAI, or a local model). We don't train, run, or improve any of
them — they belong to their respective providers.
Two things usually help:
- Edit the fields by hand. The AI's reading is a starting point, not the final word — every field is editable. You have the last say.
- Use a more capable model. If you're on a free tier (e.g. Gemini Flash), a paid tier of the same provider often reads photos far better. That usually means an API key with credits — costs vary by provider, and what you pay goes to them, not to us.