Hello everyone,
Voxta v1.3.2 is now available! This update brings a massive boost to generation speeds with cloud integrations, creates a more dynamic browsing experience with video previews, and refines the actual chat experience with a polished Visual Novel aesthetic. We have also overhauled our default content to make your out-of-the-box experience smarter and more engaging.
✨ New Features & Integrations
ComfyCloud Integration
We are excited to introduce the ComfyCloud module. You can now access any ComfyUI workflow directly through Voxta using your own ComfyCloud API key. It is extremely fast and capable of running heavy workflows (tested with Illustrious and Pony models) with lightning speed.
Shared Hugging Face Module
We have solved a long-standing request: centralized token management. You can now add your HF_TOKEN in one place (Profile & Settings). This allows you to easily download and use gated models from Hugging Face across any supported module without manual configuration hacks.
Video Thumbnails
Your library just got a lot more lively. We have added support for video thumbnails on hover for both Character and Scenario cards. Browse your collection and see your characters move before you even start the chat.
⚙️ UI and Experience Overhaul
Visual Novel Style & Typewriter Effect
The Stage and Avatar views have been significantly reworked.
Visual Novel Aesthetic: The Stage View now features a clean, dedicated text display area, giving your roleplay a cinematic feel.
Typewriter Effect: Text now streams in with a satisfying typewriter animation as the character speaks.
Attachments: Full support for image attachments within these views.
Modernized Editors & Lists
Memory Books: The editor has been modernized for a better user experience.
Model Lists: Added a refresh button and clear indicators for supported model types.
Character Organization: "Assistants" are now separated from other character types in the list for better organization.
Augmentations Visibility: Active augmentations are now visible directly in the character lists.
📦 New Content & Smart Defaults
New Built-in Characters
Say hello to Vox, Zen, Kod, and Kat! We have added these four new built-in characters with updated personas to showcase Voxta's capabilities.
Smarter Memory Books
The default memory book has been improved with over 50+ new entries. This means default characters are now "self-aware" about Voxta—they can actually help answer basic questions about how to use the app!
🛠️ Key Fixes & Technical Enhancements
Local Diffusers Optimization
A major optimization to VRAM management. We've reduced memory usage significantly (from 17GB down to ~12GB), making high-quality local generation accessible to more GPUs. We also fixed issues with SDXL Embeddings.
Voxta Cloud Enhancements
Image Gen: Added support for Image Generation via Voxta Cloud (SFW only).
Smart Filtering: The cloud list now only shows models relevant to your needs (Computer Vision vs. Image Gen).
Module Updates
ExLlamaV2: Updated to 0.0.19 with support for legacy tokenizers.
Orpheus: Updated llama-cpp-python.
Grok: Fixed reasoning models not supporting stop tokens.
OpenRouter: Cost is now included in diagnostics logs.
Linux & System Fixes
Fixed PATH overwriting on Linux to ensure NVIDIA driver access.
Fixed race conditions in the STT (Speech-to-Text) service recovery.
Fixed soundfile dependency issues for KittenTTS.
Thank you for being part of the community. Whether you are generating art in seconds with ComfyCloud or chatting with the new cast of characters, we hope you enjoy v1.3.2!