New features
- Added VoiSona Talk as a text-to-speech engine. Enter your username and API password in the settings to show its voice list, then pick a voice to speak with (Japanese and English UI).
- You can now sort the model list by newest release or by name, and it scrolls to the model you're using when you open the screen.
Improvements & fixes
- In the first-launch model download prompt, the recommended model is now tailored per language (Gemma 4 for Japanese, English, European languages and Russian; Qwen for Chinese and Korean).
- Tidied up the model list: each model now shows its release date after the name, and the descriptions were revised.
- Minor bug fixes.