Hello everyone,
Voxta v1.6.0 is here. This release is built around three big things: a top-to-bottom UI rework, custom reply suggestions that turn chat into a choose-your-own scene, and the groundwork for HY-Motion - our new module that will let your characters move as you speak to them. On top of that, Voxta Cloud got a lot of attention with prompt caching, new frontier models, and a much better model selector.
🎨 New Voxta UI - Total Rework
We rebuilt the UI from the ground up. Same Voxta, much sharper.
Redesigned home page and a refreshed wizard + registration flow
Improved styling throughout the app - consistent chrome, cleaner cards, better hierarchy
Two-pane preset editor so you can scan all your presets and edit on the right
Improved Voxta Cloud model selector with model details and pricing visible up front
Improved reasoning UI
Vision playground streaming
Audio input ripples while speaking - small thing, big feel
Audio input/output picker when using NAudio or OpenTK — switch devices without leaving the chat
Local muting instead of fully disabling services in-chat (mic/speaker buttons no longer unload your models)
Mobile UX pass - Voxta is much more usable on smaller screens
Improved Voxta Lite UX
🎭 Custom Reply Suggestions
You can now author your own reply suggestions per character - including custom colors. Drop in anything you want ("rope play", "tell her the truth", "draw your sword") and they show up as picks in the chat, like a PR game. Choose how you want to respond instead of typing every line.
Enable it in Apps → Voxta Talk.
🕺 HY-Motion - Animations service
A brand new animation module type. HY-Motion generates real character motion from text in real time, so your characters actually move while they're talking to you instead of standing still. The generated motion ships as SMPL animation data that drives your character — the first integration is VaM, with more coming.
This is a whole new service category in Voxta (alongside TTS, STT, LLM, image gen), so other animation backends can plug into the same slot down the line.
☁️ Voxta Cloud
Prompt caching across supported models - should noticeably cut your usage cost
Latest models added: Claude 4.8, latest OpenAI, and more - the catalog is a lot fresher
Model Max Context override for Assistant chats, separate from the context window
Dedicated out-of-credits modal when Cloud returns PaymentRequired (no more confusing "unauthorized")
🚀 Backend & Module Improvements
MCP: one module can now expose multiple tools, with a much better UX
MCP fixes: zero-argument tools, image rendering, and general robustness
llama.cpp: shared vision + chat model — lower memory footprint when you're using both
llama.cpp updated to 0.3.39, ExLlamaV3 updated to 0.0.38
Namespace-aware enum mutex - state.outfit.dress and state.pose.sitting can finally coexist instead of overwriting each other
Orpheus / Sesame: automatic ref_text transcription — drop a reference clip, we transcribe it for you
🛠️ Key Fixes
Minecraft Companion: fixed ArgumentOutOfRangeException (zero-character cases now work)
Lovense: compatibility fixed with the iOS Lovense Remote app
LocalDiffusers: xformers kernel crash, weighted-embedding device mismatch, and VAE decode device mismatch all fixed
NAudio: device-listing memory leak when "Could not find output device" warnings fired
Deepgram: now correctly reports "unauthorized" instead of "payment required"
Orpheus: fixed the access-violation-reading-0x...5400 crash and a continuity bug
STT playground: no more "Unexpected recording call while recording is disabled" when leaving
Configurations canvas: connections and node positions now actually save
Thinking models: token counting fixed so you can control both thinking and reply length
📦 Voxta Installer
Faster install - we precompute the install size at build time instead of scanning at runtime.
Thank you all for the continued support and feedback. Whether you're enjoying the new UI, writing your own reply suggestions, or watching the animation module come together - we hope you enjoy v1.6.0.
After subscribing to the required Patreon tier, you can get the latest version at: https://portal.voxta.ai/
If Hy-Motion isn’t working, please install Orpheus TTS. It includes the dependencies that Hy-Motion needs to run. This will be fixed in the next hotfix. Also, in order for a character to perform animations, you need to use the attached scenario: HY-Animation.1.0.0. voxpkg.
To test if the module works correctly in the VAM-Voxta chat, type:
/perform character is doing jumping jacks.