This update improves custom character textures and updates the local AI options.

- Improved shading and hair detail on custom characters, and fixed washed-out colors and texture seams in the cafe.
- Updated the AI options: High prioritizes quality, Medium prioritizes speed, and Low now uses the smaller LFM 1.2B model to reduce memory use.
- Low is a lighter option, with lower reply quality than the larger models.

All the best,

WhyDuck Studio

MoChat still uses the local AI model on your PC by default. This update adds an optional way to use a cloud model instead, plus a few fixes.

1. Windows players can now connect their own API key from OpenAI, Anthropic (Claude), Google (Gemini) or xAI (Grok) in Settings > AI Model Quality. While this option is on, your chat messages, recent chat history, saved chat memories and the character's profile are sent directly from your PC to the provider you picked, and usage is billed to your own account. The key is encrypted by Windows and stored only on your PC. It is never sent to us. You can delete the key or switch back to the local model at any time.
2. If a local model is too slow on your PC, MoChat now tells you while the model is still loading, and suggests switching to the Medium model when replies take a long time. The recommended model also takes your GPU memory into account.
3. If Steam asks you to accept the Workshop agreement before you upload a character from the in-game creator, the game now opens it for you. After you accept, upload again and it continues the same item.

Warm regards,

WhyDuck Studio