Voxta 136 Update: F5 TTS, Azure Wake Words, Chain of Thought, and More!

Hey everyone! 

We’ve got some updates to share with you in Voxta Beta v136. We’ve been working hard to bring you new features and improvements. Here’s what’s new:

This release introduces two new TTS services:

F5-TTS: Finally, we have a high-quality local text-to-speech service with voice cloning capabilities! With just 15 seconds of audio, you can clone any voice and enjoy incredibly realistic and natural-sounding results for your scenarios.

Kokoro TTS: A fast and lightweight text-to-speech service with high-quality built-in voices. While it doesn’t support voice cloning or advanced intonation, it provides almost instant responses.

Azure Wake Word Integration

Activate Voxta by saying a keyword like "Hey Voxta." You can also train your own custom keywords using Azure for even more customization. Additionally, you can set up deactivation keywords such as “cancel” or “stop listening” to easily control when Voxta stops responding.

Chain of Thought

Your character can now think before replying, and you can even peek at their thought process. It feels like you’re reading their mind, giving you a unique glimpse into how they approach conversations and decisions.

Auto Continuations and Follow-Up

Characters can follow up on conversations or continue storytelling automatically. They’ll keep going until you interrupt, making it easy to explore long narratives.

Custom Background Images

This is a huge update for anyone who loves stories and visual novels. You can now set custom background images for your chats, creating the perfect atmosphere for every scenario.

Privacy with Ephemeral Chats

Enable ephemeral chats to ensure conversations aren’t saved. Perfect for maintaining privacy and control over your data.

Dialogue Suggestions

If you’re stuck or need ideas for how to reply to your character, Voxta can provide some suggestions to inspire you and keep the conversation flowing. Additionally, you can manually type and ask the AI to generate ideas on how to reply.

Vision

Share your screen or specific windows with your AI. You can now select specific windows for AI analysis. This ensures the AI focuses only on the relevant areas of your screen, avoiding unrelated visual data and providing more accurate and context-aware responses.

Technical Improvements and Changes

This update also includes important technical upgrades:

Memory Enhancements: Better handling for improved context awareness.

UI Updates: Added chat favorites and more streamlined navigation.

Framework Upgrades: Updated to Python 3.12 and Torch 2.5. Note: Re-installation of services is required after upgrading.

Breaking Changes

Hugging Face Models: Models now require the "hf:" prefix in settings. This is applied automatically during the upgrade.

Folder Changes: Directories for Whisper and Florence-2 have moved to `Data/Models/Whisper` and `Data/Models/Florence-2`.

This isn’t everything we’ve added! There’s plenty more in this update—check out the changelog for a complete list. We hope you enjoy exploring these new features and can’t wait to hear your feedback!