Voxta v1.13.1: Plugins, a New Interface & Stage Building (Patreon)
Content
Hello everyone,
This release brings a plugin system: most modules are now downloaded when you add them, and you can create and install your own. The interface has been rewritten around a rail of sections, with a new look, a Play library and guided tours, so it is easier to find your way whether it is your first chat or your hundredth. The new Stage and Places editors let you build multi-location scenarios without scripting. Music generation, 3D models, pose control, voice design, version history for your characters and scenarios, and a Linux desktop app fill out the rest.
This covers v1.13.0 and the v1.13.1 fix release that followed it. There is a lot in this one, so here is what changed.
🧩 Plugins & Modules
Releases now ship only the Lite and Full modules. The rest are downloaded when you add them, and you can build your own.
Module Registries from Voxta Cloud or GitHub repositories (private ones too, with a GitHub token), with updates, Upgrade All and rollback to the previous version
Releases verified by Voxta Cloud are marked Official; third-party, flagged and blocked releases say so before installing
Preview what a module source offers before adding it, and install from voxta://modules/install links
Release builds download the official modules without a Voxta Cloud account
Modules load and unload while Voxta runs; a restart is only asked for when needed
Cancel a module while it loads, or kill one a chat holds
Nothing installs on its own: update checks run once a day when an administrator signs in
Add and Manage Modules: categories, update badges, search, and why a module did not load
One install page for the setup steps, settings and models of the modules you add
Module settings and presets in one popup, with Export, Import and Uninstall
Building your own:
Load a module from a dev folder, a zip or a DLL, and reload it as you build it
With the Module Dev Tools module, Work chats get skills, examples and tools to create, build, install and test a module (augmentations, chat bridges, game integrations)
Modules can ship skills, setup steps and scenarios, and declare their Python requirements, category and author
The SDK examples and the module template on GitHub are the place to start
🧭 A New Interface
The UI has been rewritten to be easier to navigate for first-time users and power users alike.
A rail of sections with their pages beside it, recent chats and projects in the side panel (with favorites), and swipe from the edge to open panels on phones
Play gathers Games, Saves and Integrations; a scenario's page resumes its last chat, with Continue and New Game for games
Move to… places a character, scenario or memory book in Chat, Play or Work
Guided tours from Help › Take a tour: Your first chat, Configuration, Working in a folder and Creating a character
New look: Inter font, flat forms with short help behind an "i", quieter dropdowns, menus, dialogs and toasts, and plainer, shorter text
Theme palettes: Midnight, Ember and Black on dark, Paper on light
Character, scenario, collection, memory book and package editors share one layout: a sections column, a centred form, a save bar, a Tests tab, Versions, and their assistant in the header
Pages lay out against their own width, so they fit beside a docked right panel, which can be resized
Sign in, registration, the setup wizard, import and startup pop-ups are redesigned
The setup wizard gives Full builds a manual setup path, and its category steps install downloadable modules
Start a chat: a compact list of switches, the module behind each service, and the services a scenario needs with a Fix button
The Configuration page shows every service in one table, per medium, with its module, preset and model
🎭 Stage & Places
Build multi-location scenarios without scripting.
Places with versions, per-place context and cast, arrival stories, repeatable places, story lines, travel notes and transition screens
The Stage editor: stand characters on the stage, lay out and edit the HUD, landscape and portrait layouts, setups that apply with flags, snapping and undo
Hand-placed spots for each role, and off-screen roles that talk but never stand on the stage
Buttons and hotspots: flags, markers, a sound on click, "Only once", and a line said by a role
Hotspots: smooth edges, a white outline preset, label position, a pixel font, and placement behind the characters
In the chat, places and overlays show on the stage with Look around, and the Characters panel says where each one is
Audio environments (cavern, hall, muffled, phone, radio and underwater) for voices and sound effects, per place or from a script
Skip becomes Continue when the next message waits on the reader
📝 Scenarios & Scripts
The scenario editor has Places, Stage, Variables and Run tabs, and Ctrl+G goes to any script, test, event, button or action
Variables can be declared with a value, constant or server-only, with JSON variables (optional schema) and tables (chat.table()) with indexed columns
Script tests: Jest-style specs that run each test in a fresh chat, with coverage and lint rules, run inline with a play button per test
Scripts can choose who takes the next turn, let a role speak with chat.roleMessage, show buttons by name and limit setContext to a role
chat.secret() and chat.note() can address some characters only, and memory books match only what each character could read
Scripts can create and delete characters during a chat, with a matched voice and an avatar drawn from a prompt
Scenarios can recommend or require voice, image, video and music generation, and Start a chat warns when the profile does not run them
The scenario assistant builds the stage, places, variables and tables, adds roles and the narrator, draws pictures, makes music, and runs tests and lint
Scenario test chats: the assistant plays the scenario itself, presses buttons, answers choices, and reports what each script changed
🎵 Music
Voxta can now make music, with a new service type and two modules.
ACE-Step, with LoRAs, takes, stems, repaint and remix
LeVo 2
/music and /song write and play a song from the story
The Music Playground: write, adjust and remix songs, with a small multitrack session
🎨 Images, 3D & Create
MOSS-Image, a new local image module with LoRAs, Turbo presets, INT8 and low video memory options
Pose control for MOSS-Image and Krea 2: region, depth and OpenPose skeleton maps, a VRM figure posed over the image (drag joints, undo, frame slider), or a pose prompted through the animation service
Krea 2 can edit pictures through edit LoRAs (Identity Edit, General Edit, Style Reference)
3D model generation is a new service type, with the TRELLIS.2 module making textured models from a picture, and a 3D Playground
GGUF models and text encoders for Wan 2.2, LTX-2.5, MiniMax H3, Krea 2, MOSS-Image and Qwen Image Edit, Q4_K_M by default
Create projects keep your image, video, music and 3D generations with their prompt and settings; restore a result's settings, star pictures, animate one into a video, or fork a new project
Projects and Gallery pages, and projects travel in backups
Picture prompts are written in the style the image preset asks for
The image and video prompt assistants share one editor with LoRA tools, and can see the last picture and generate
Right-click a picture or thumbnail for its menu, including Send to Create Image
🎙️ Voices
The new audio.cpp module: local text to speech (Breeze TTS 2, Qwen3-TTS, Ming 3.0, PocketTTS, Kitten TTS 2, Echo-TTS, VibeVoice, OmniVoice, Fish S2, with audio tags), streaming speech recognition (Nemotron), and ACE-Step music
The Ming-3.0 TTS module: cloned, built-in and described voices with emotions
Voice Design: make up a voice from a description with ElevenLabs, Voxta Cloud or audio.cpp
Voice Studio: design a voice, keep the generations, and save them to the library or a character
Characters have one voice sample shared by every cloning voice service
Voices show each provider's price and quality, and a Source tag for the voices that ship with Voxta
The TTS playground plays the voice in each audio environment
🤖 Assistants & Characters
The Character Builder runs on the Assistant service, saves at the end of each reply, upscales, compresses and cuts out its pictures, and draws the character for the stage
Assistants can read your content, look at pictures and at the stage, pick voices and restore versions
AI Help: type while the assistant works, see how big its context is, and continue a failed reply
The App Assistant shows on screen where something is, switches augmentations and keeps its chats
The AI character creator is one short form, with the steps as a checklist
Add a character offers templates (Companion, Roleplay, Storyteller, Blank), Create with AI or Start from scratch
Collections get AI Help, which can hand character fixes to the Character Builder
🗂️ Versions & Undo
Characters, scenarios, memory books and collections keep their history.
Compare any version with the previous or the current one, and step through the changes
Restore a whole version, or one field or row at a time
🎮 Games
A Launcher service type, with Steam, GOG and Epic games detection on Windows and Linux, and a Game Shortcut module for any game
Modules can serve HTTP endpoints that game mods call with an API key, and run headless chats with game audio and turn steering
Chat bridges can ask for speech, so a game can voice its replies
💬 Chat
Up to 3 chats at once, so one can run in its own window
Events, voiceless lines and typed-out messages take their turn, so each finishes before the next one plays
Stop one running tool call without ending the reply
Reasoning effort gets a Minimal level and a per-chat override
Tool calls and thinking show as one quiet trail of steps, including Claude Code subagents
The Chat Explorer has a message timeline with the chat's facts, and the Chat Gallery shows the chat's pictures, videos and music
Chats are listed one row per chat with day headings, and Home shows running chats with the characters' faces
Right-click the mic, the speaker or the dictate button to pick the audio device
Rename chat can name it for you
🛠️ Workspaces
New projects from templates or git repositories, with a git branch per workspace
.voxta/workspace.json can ask for augmentations
Any chat can use a folder with the Workspace augmentation
Augmentations, including MCP servers, can be enabled for Never, Characters, Assistants or All, with a per-character switch for default augmentations
⚙️ Models & Performance
Models can live in several folders, searched in order, with rules for where downloads go; free space and the download size are shown before a download
The Models page filters by disk and moves files between models folders
One model picker for installed, recommended and downloadable models
Image, video, music and 3D generations unload their module when done
The System Monitor shows memory and GPU load per module, with Unload unused
Startup warns when Windows holds RAM no process owns, which freezes GPU modules until a reboot
Updated to .NET 11 and EF Core 11, ExLlamaV3 1.6.6, llama-cpp-python 0.4.5, LLamaSharp 0.30.0, ONNX Runtime 1.31.1, Transformers 5.22, Diffusers 0.42, FFmpeg 8.1.3 and more
🔐 Content & Access
Hide explicit content marks every resource as adult the same way on every save, and hides Create projects with a profane prompt
A Safe character gets the profanity check
Resources and packages carry a minimum Voxta version, and the Content Hub greys out Install for packs that need a newer Voxta
API keys get an Edit services scope, and resources.write manages resources
An app on the Voxta computer can ask signed-in users to approve it, without typing a code
Character from a picture strips the photo's metadata before upload
🖥️ Voxta UI
Account: one button for the picture, name and credits, with Theme and Palette in its menu
Settings, modules, workspaces, API keys, backups and users as plain rows that open their own pages, and one Security page
Library pages: a single toolbar and skeleton cards while loading
Playgrounds: the result fills the page and the composer is docked under it
Asset explorer: move a file to another folder or clone it under a new name
Middle or Ctrl+click on the rail opens a section in a new tab
Ctrl+S saves the field being typed in and keeps focus in it
Credits warnings when draining fast, almost out, or sending a large prompt
Service Tests can queue tests during a run, and the Tools Playground shows arguments as a form
Choice fields with a few options show as a switch
🪟 Voxta Desktop
A Linux desktop app (x64 and arm64), as a tarball or a deb
Chat windows open as their own desktop windows, and avatar overlays follow them
📱 Voxta Mobile
Web fetch and an Android shell tool
☁️ Voxta Cloud
The Module Registry is the official source of downloadable modules
Voice Design through ElevenLabs
Image and video models come with their prompting style, and image models show their resolution tiers
A reply whose thinking used the whole budget gets a follow-up instead of coming back empty
🧰 Voxta SDK
Voxta.Sdk.Modules packs a module for a GitHub release, with the Voxta.ModulePublish dotnet tool
Module API: user contexts, HTTP endpoints, headless local chats, next speaker and interrupt, and tool results that end the turn
Module traits are flags, so new ones never break a published module
Voxta.Model compiles in Unity 6.5 and later
🆓 Voxta Lite
New to Voxta? Lite is free and runs on Voxta Cloud, so you can create an account and start chatting without any local setup. If you like it, a subscription gets you the full version, with a large library of integrations, custom modules, and the option to run everything on your own hardware.
This release also carries a long list of fixes, including a reply never ending when its speech arrives faster than it plays, a reply running on into another character's turn, and exporting a package as PNG.
Thank you all, as always, for the continued support and feedback. Whether you are installing your first plugin, finding your way around the new interface, or building a scenario that moves between places, we hope you enjoy this release.
Voxta is available to subscribers at: https://portal.voxta.ai/