v0.7.0 Patch 1 🐾...and Special Notes Regarding Gemini TTS

Project N.E.K.O. · build 22084523 · published

On SteamNotes only · this build shipped before we watched the app

Hello Humans! Following the major v0.7.0 release, the Dev Nekos have prepared Patch 1 to smooth out some initial bumps:

  • Screen Sharing Fix: Resolved a critical bug where some users couldn't share their screens. Neko can now clearly see your desktop again!

  • Important Note: Screen sharing is only active when "Proactive Vision" or "Screen Sharing" is manually enabled by the user. (For proactive chat with screen share sources, two switchers - proactive vision, proactive chat->screen share, must be BOTH turned on) Your privacy is our top priority. If you suspect that N.E.K.O. has accessed your screen without permission, please report it to us immediately so we can conduct a full investigation. Thank you for your trust!

  • Resizable Dialogue UI: Is the text box getting in the way? We've updated the Dialogue UI to be freely resizable, allowing you to adjust it to your liking.

  • Instant Agent Task Cancellation: Added an immediate cancellation feature for Agent tasks. If Neko gets a bit too carried away, you can now stop her instantly.

  • Russian Language Support: Official Russian language support is now live! A massive welcome to all our Russian-speaking Masters, nyan!

SPECIAL NOTES

In the past week, the Dev Nekos have received reports regarding significant delays or total generation failures with Gemini TTS. After extensive testing in our "N.E.K.O. Lab," we would like to clarify the following:

  • Root Cause: We have confirmed that this issue originates from the Gemini server-side. Since Gemini TTS is currently in its Preview stage, it is highly unstable. It frequently encounters errors where it returns text instead of the requested audio stream (effectively performing "text-to-text" instead of "text-to-speech"), causing Neko to stay silent for too long.

  • The Patch: Referencing other open-source implementations, the Dev Nekos have patched the calling method to explicitly request audio generation from Gemini TTS. This should partially restore the voice experience for most users.

  • Clarification on Performance: Some users suggested that our previous mini patches degraded Gemini's voice performance. However, the Dev Nekos believe this is not the case. The issues likely stem from Google's internal updates to the Gemini TTS Preview model itself, rather than our local modifications, nyan.

The Dev Nekos will keep a close eye on Google's updates. Thank you for your patience!