Voxta 145 Beta: Reasoning Models, GPT-Style Chat + Video Support, and a lot more!

Hey everyone,

Beta 145 is here, and it's packed! This update brings some big changes, focusing on making Voxta smarter and more useful for different tasks, not just roleplay. We've added reasoning model support, a new Assistant view (think GPT-style chat!) turning Voxta into a serious productivity tool, much easier video integration for scenarios, and support for the latest Nvidia 50x series GPUs!

Here’s the breakdown:

🧠 Productivity Power-Up & Smarter AI

• Assistant View & Reasoning: Voxta can now be your productivity buddy! The new Assistant view (with full markdown support!) combined with "thinking" models helps you write, code, or brainstorm. See the AI's reasoning process right in the UI!

• Your Choice: Local or Online Power: We don't lock you in. Use your own local LLMs for full privacy, or connect to monsters like Gemini 2.5 Pro – you have the freedom, no compromises.

• Fine-Tune Control: Added options for custom system prompts per-character/scenario (add to, replace intro, or fully replace) and better handling of text formatting (like line breaks) from models.

• Direct Image Input: Multimodal models can now often use raw images directly (using 'Normal' formatting style), simplifying workflows.

🎬 Easier Scenarios & Better Visuals

• MP4 & AVIF Avatars: Big news for scenario creators! You can now use .mp4 videos for avatars – no conversion needed, just drop it in and script it! We also added .avif support for super-compressed, high-quality video assets – perfect for large scenarios.

• Life-Changing Inspector View: Seriously, if you build scenarios, this is huge. The new Inspector view shows everything that's happening – events, actions, scripts firing, flags changing – making debugging way easier.

🛠️ Workflow & Scripting Goodies

• Multi-Audio Tracks: Layer background music and ambient sounds using app triggers, with individual volume control and stopping.

• Folder Watcher Vision (New Service): Point Voxta at a folder, and it'll automatically send any images added there to computer vision.

• Scripting Upgrades:

  • Scripting: Allow using arrays and objects in chat variables

  • Scripting: Simpler and more reliable way to get assets: e.character.assets.get(path), chat.scenario.assets.get(path) and help methods oneOf(regex) and oneOrNoneOf(regex) - Use this with SetBackground, SetAvatar, PlayMusic, PlayAmbient, PlaySound and PlayVoice app triggers.

  • Scripting: New @voxta/utils package with oneOf function (choose randomly from a list)

  • Scripting: chat.on("", () =>{}); can be used instead of chat.addEventListener.

• Scenario Controls: Ability to disable character bootstrap messages inheritance (-) and prevent them from running for characters disabled on start.

⚙️ Hardware & Core Stuff

• Nvidia 50x Series Support! Yep, if you've got one of the new Nvidia 5000 series GPUs, the update to Torch 2.7 & CUDA 12.8 means Voxta should now work smoothly for you.

• ExLlamaV2 Updated: Running the latest v2.9.0.

• Key Fixes: Patched up issues with speech start events, scenario character loading order, and a few rare crashes.

📱 Mobile & UI Polish

• Looks Better on Phones! The Avatar view is now better organized for small screens (mobile devices) and includes a nice typewriter effect for messages, making the experience much smoother.

✨ UI Polish & Quality of Life (Desktop & Web)

• Paste Images in Prompt: Easily add images by copying and pasting them directly into the chat input box.

• Form Improvements: Better display and handling of default values, plus improved validation messages in configuration forms.

• Smarter Dropdowns: Dropdown menus now open more intuitively and clearly show your current selection while Browse.

• Preset Saving Fix (Desktop): Fixed issues with saving presets using Ctrl+S in the chat's preset tab.

• And More: Lots of other small tweaks like improved avatar hover effects for a smoother experience.

Important Notes: 🛠️

• Those reasoning models can be hungry! Running them locally alongside other AI processes might need a decent amount of VRAM/RAM.

• Remember, this is still beta. We rely on your feedback! Please give the new Assistant view a try for productivity tasks, test out the .mp4/.avif video support, and let scenario creators tell us how the new Inspector view feels! Hit us up on Discord or comment here.

Links:

• How to install Voxta server app: https://youtu.be/1I9VkJ8tTlo

• How to update Voxta server app: https://youtu.be/5aa7sducwoc

Thanks for being awesome and supporting Voxta!