yeemachine
How to replace a new live2d model
Thank you very much for sharing a fun project！
I collected some new live2d models from the live2d website, but replaced them with a new live2d model path in the code, and I can see the new live2d model, but I can't make the new live2d model move with my face movements。
After analyzing the intermediate running results of the rigFace function, no obvious abnormalities are found, you can obtain the live2d model parameter values and assign new values to the live2d model parameters.
So I would like to ask what the reason for this phenomenon may be. How can we use other live2d models？Looking forward to your reply。
Hi! Please refer to the pixi-live2d-display library for any questions about how to work with Live2D models on the web.
Other live2D models can be loaded by URL.
You will need to modify the rigFace function to suit your Live2D's parameters. Live2D parameters can be whatever the rigger named it, so be careful there.
Another question, do you know how to reduce the resource consumption when the face motion drives the Live2d model, such as reducing the resource consumption of CPU, memory, GPU, etc.
Because the current face driving a Live2D model consumes nearly 10% of the CPU, 35% of the GPU, and nearly 1GB of memory, I hope to take certain methods to reduce resource consumption.
Hope to hear your good advice on this subject @ @yeemachine
@yeemachine Can you answer this question? Looking forward to your reply。
This is sometime I'm actively exploring as well so I don't really have a solution for this. Maybe you can ask the Tensorflow JS/Mediapipe community since most of the processing power is consumed by their mocap models.
I made a serious estimate, most of the CPU, memory, and GPU resources are spent on pixijs to create application objects and add live2d models to the stage, which consumes 26% of the GPU, and the AI model calculation consumption is relatively low, only 14% GPU,
so in terms of creating application objects and adding live2d models in pixijs, is there an improved way to significantly reduce GPU resource consumption. looking forward to your reply @yeemachine
Unfortunately, that is outside the scope of this repo, which only handles the calculations from ML results. If you have any rendering questions, please ask the owners of pixijs or pixi-live2d-display. They will have much more experience.
