I've read through most of the posts on the hub and looked through the website. I've also looked through the discord. I am still a bit confused as to what exactly Voxta with the VaM plug-in will do.
- Are you only able to use it with specific scenes?
- Can I use this in any scene including self created ones?
- Does/can Voxta also control the movement of the character in VaM?
- If I want to run this locally do I have access to all of the features?
- Do I set personality and other settings outside of VaM for the individual character? Does it save or do I need to change it every time?
- Is the VaM character able to be aware of thier in game environment?
- Does the VaM character respond to touch or interaction within the game environment?
- Voxta can only be applied to one character in scene?
- Does voxta with VaM only use speech or can it also provide sounds like moaning?
- Instructions on setup and use are lacking (or I couldn't find them).
If you could answer these questions or guide me where to find them, that would be great. I've been wanting to build a VaM plug-in with these features but, it is a huge undertaking and I opted not to. So, I very much appreciate you taking this on.
I'd be happy to support you on your adventure with AI. I just want a better understanding of what you have and are working on.
Thank you
Hey! Here are your answers!
1. Are you only able to use it with specific scenes?
No. The plugin is just an MVRScript you add to any Person atom via `Voxta.cslist`. Demo scenes are a convenience, not a requirement.
2. Can I use this in any scene including self-created ones?
Yes. Load your scene, go into Edit mode, select a Person atom -> Plugins -> add `Voxta.cslist`. Open Custom UI, pick a character from the dropdown, and it'll connect to your Voxta server.
3. Does/can Voxta control the movement of the character in VaM?
Not directly - Voxta doesn't animate bones itself. What it does expose:
- A State storable that flips between `idle`, `listening`, `thinking`, `speaking`.
- An On State Changed trigger and an On Action trigger.
- A CurrentAction string whenever the AI decides to invoke a custom action you've defined (jump, wave, blush, etc.).
You wire those triggers into Timeline (for animation clips) or Scripter (for scripted logic) - that's the pattern shown in the README. VaM's built-in Auto Behaviors -> Lip Sync handles mouth movement from the TTS audio. So movement is "AI decides -> your scene executes."
4. If I run locally, do I have access to all of the features?
Yes. Voxta server runs on your machine (default `
http://127.0.0.1:5384`). It supports local LLMs (KoboldCpp, Oobabooga, Llamasharp, Exllamav2, v3 etc), local TTS (XTTS, Piper, Coqui, etc.) and local STT (Whisper). The VaM plugin doesn't care where the AI lives - same SignalR connection either way. The quality of the experience depends on the local models you pick, but no feature gets locked off.
5. Do I set personality outside of VaM? Does it save?
Yes. Characters (name, personality, scenario, system prompt, voice choice, example dialogue, memories) are created in the Voxta server's web UI. They persist server-side. In VaM you just select a character from the dropdown and the plugin stores that selection in your scene/preset - you don't redo it every time.
6. Is the VaM character aware of their in-game environment?
Partially, and it's up to the scene to feed it. The plugin exposes Context slots - string fields. E.g. your scene can write "User is in the bedroom. Lights are dim. It's 2 AM." into a context slot and the AI sees it as part of its prompt. There's also a Flags system (set/read flags) for AI-visible state and conditional actions. So awareness is as rich as your scene logic makes it - nothing automatic.
7. Does the character respond to touch or interaction?
Not out of the box, but it's designed for exactly this. You hook VaM's collision triggers (or Scripter) into Voxta's TriggerMessage storable (injects a user message) or RequestCharacterSpeech (makes the character say something). You can also set flags or update context when touched. Example: a collision on a breast trigger -> Scripter writes "User is touching your chest" into Context, AI reacts accordingly. Again, plugin is the bridge; scene provides the inputs.
8. Voxta can only be applied to one character per scene?
No. A single plugin instance supports up to 3 characters (character 1, 2, 3 with separate roles and names). And you can add the plugin to multiple Person atoms if you want truly independent brains. Multi-character conversations are supported server-side.
9. Speech only, or can it also do moans/sighs/etc.?
Depends on the TTS engine you pick in Voxta, not the plugin. Expressive models (XTTS, ElevenLabs, some fine-tuned voices) will vocalize moans, sighs, laughter, gasps when the LLM emits them - either as inline action tags or via audio-tag syntax the TTS supports. Basic TTS engines won't. So: prompt your character to use those cues + pick a capable voice = yes, it works. It's emergent from LLM output + TTS capability, not a dedicated plugin feature.
New storables — last message metadata
