I've just finished my Action List Manager plugin and I'd like to share it with you. You can link actions into a tree of choices to build a consistent story. It supports triggers and multi-context - location/clothes/custom injecting into Voxta.
Instructions:
The flow is like so:
LLM inferres action -> Voxta -> On Action Trigger -> ActionListManager -> executes triggers and changes ActionsList and Context in Voxta
If you defer the action until the character stops speaking (checkbox in the action definition), then ActionListManager memorizes the action and calls it on getting the idle state update from Voxta.
As I don't have the possibility to do mocaps, timeline animations for new scenes are not possible for me so I do (Pose + Synergy) pattern for idle movements. It works really good alongside Voxta. I'll share a new version of this plugin soon with a demo scene for an ease of use. I'm waiting for the permission from the Synergy creator to make changes in his plugin to allow loading config on a trigger.
A word of warning:
This plugin is made specifically for Voxta, which is currently a paid experience. If you want to make use of it, you need to be a patron of this initiative, which I highly encourage you to do.
If you have a hard time understanding either my plugin or Voxta just drop me a message on Discord. We can talk or even do a screen share.
If you like what I'm doing consider subscribing to my Patreon.
------------------------------------------------------------------------------------------------------------------
Character for testing action inference:
------------------------------------------------------------------------------------------------------------------
The LLM model I recommend:
------------------------------------------------------------------------------------------------------------------
More tutorials:
Layers Tutorial
Transition Actions Tutorial
Instructions:
- Add the plugin to the Person atom along with Voxta.
- Define a tree like the one in the picture below linking actions together.
- Add custom contexts for those actions. The general rule is that when a field is empty it does not substitute the value that already exists. It applies both to context and linked actions.
- Context will be constructed like so: "{{char}} is feeling shy. {{char}} is in {{user}} living room. {{char}} is wearing pink blouse and jeans"
- In Voxta plugin "On Action Trigger" call TriggerActionChange on my plugin.
- In Voxta plugin "On State Change" call TriggerStateChange on my plugin
The flow is like so:
LLM inferres action -> Voxta -> On Action Trigger -> ActionListManager -> executes triggers and changes ActionsList and Context in Voxta
If you defer the action until the character stops speaking (checkbox in the action definition), then ActionListManager memorizes the action and calls it on getting the idle state update from Voxta.
As I don't have the possibility to do mocaps, timeline animations for new scenes are not possible for me so I do (Pose + Synergy) pattern for idle movements. It works really good alongside Voxta. I'll share a new version of this plugin soon with a demo scene for an ease of use. I'm waiting for the permission from the Synergy creator to make changes in his plugin to allow loading config on a trigger.
A word of warning:
This plugin is made specifically for Voxta, which is currently a paid experience. If you want to make use of it, you need to be a patron of this initiative, which I highly encourage you to do.
If you have a hard time understanding either my plugin or Voxta just drop me a message on Discord. We can talk or even do a screen share.
If you like what I'm doing consider subscribing to my Patreon.
------------------------------------------------------------------------------------------------------------------
Character for testing action inference:
------------------------------------------------------------------------------------------------------------------
The LLM model I recommend:

TheBloke/MLewd-L2-Chat-13B-GPTQ · Hugging Face
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
huggingface.co
------------------------------------------------------------------------------------------------------------------
More tutorials:
Layers Tutorial
Transition Actions Tutorial