@micah yeah i definitely want to play with it more. I think it could be a lot of fun. I kind of kept hitting walls today where like everything is way better documented and supported on Windows and even the most performant capture method is windows exclusive right now AFAICT
Found a PNGTuber app that works and made a very basic avatar. When I get this elementary blog post up we’ll test it out and do our first elementary stream in quite a while 😅
Diving down this rabbit hole further I learned some things:
Live2D avatars seem to be the most popular and best looking but are also the most complicated and it seems like it almost requires Windows + iPhone to use them optimally
PNGTubing is a lot more accessible and there’s more software available, with the one major caveat that it only does volume tracking ie lip flaps but no face tracking
3D VTubers seem a lot less popular/more gimmicky probably not worth spending time in this direction
It seems like the skills developed by creating PNGTubers can translate to creating Live2D VTubers in terms of drawing a model with layers that you slice up and rig etc. With far less complexity but there is a surprisingly high ceiling with the right software!
The most basic thing you need is a 2D character with two sprites: mouth open and closed. The practical baseline is adding eyes open and closed sprites. So with 4 sprites you’re tubing!
Gazo Tuber works perfectly on Linux and I was PNGTubing with it in just a couple hours. So if you want the MVP VTubing experience, this is my recommendation to get started! https://risunz.itch.io/gazo-tuber
Where it seems like my next steps into this world are is with PNGTuber+. This app also works on Linux and offers a more complex approach similar to Live2D where you’re separating your model into individual body parts with physics and parts can be sprited for animations etc: https://kaiakairos.itch.io/pngtuber-plus
This morning I exported my model into a bunch of layers and started playing with giving her some bouncy hair physics. The only problem I’m having right now is it seems like the app is focus stealing so if I can’t workaround that it’ll be a show stopping bug, but it’s promising so far!
I’m open sourcing my model, which right now isn’t saying much but hopefully can be useful if I keep having fun and making it more complex: https://github.com/danirabbit/Tuber
Currently I’m using Inkscape because that’s what I know best, but it sounds like Krita might have built in tools for exporting your layers so I’m planning to explore that. Otherwise I might end up adding a meson build system to generate layers and sprites etc from the SVG
@egeexyz The background is a Canva template, but otherwise yeah! I definitely referenced a lot of stuff and her design isn’t super original or anything but yeah drawn in Inkscape!
I now have a Pinterest board and a note with a bunch of thoughts about palette, visual motifs, usernames, etc haha. So I think for the next version of my PNGTuber I’ll have a lot more art direction and a more cohesive brand and style. I’m also thinking about things like Twitch emotes and sub badges etc. this is a deep rabbit hole if you want it to be 😱
I wanted a nicer chat box on my waiting screens but I didn’t want to pay $15 so I made my own. Turns out StreamElements has a bunch of custom widget examples so it was super easy to fork that and make a custom chat box. This isn’t anything special, but feel free to use it or fork it or whatever :) https://github.com/danirabbit/streamelements-chat
I think I’m going to get into head turning today and learn about Group Meshes (I think it’s the equivalent of Deformers in Live2D). I kind of have wondered if it’s worth streaming this learning process but tbh at this point I think it’s probably boring to watch. It’s a lot of reading, searching discord, watching YouTube videos, trying stuff, it not working, repeat until I realize I need to change my model, etc etc. it’s very chaotic.
I think right now my plan is to learn my foundations, apply the things I’m learning to this model as an iterative process, and then once I understand how to model, cut, rig, etc then I can go back through and make a nicer and more complicated model and stream that process. It would be a more interesting stream if I vaguely know what I’m doing and probably more valuable for others to learn from if it doesn’t include too many mistakes and uninformed decisions
Haha okay so we found out today that running Inochi Viewer and playing a video game and streaming all at the same time is a bit too much for this computer. I’ve been doing all this on a mini pc. Like basically an Intel Nuc 😂 So I think I need to basically dedicate one computer to streaming and inochi and then another computer for the game and do HDMI capture
Last night I stayed up pretty late drawing and cutting new eyes. During my lunch break today I meshed and rigged them. I’m pretty freaking happy with that. I watched so many videos haha
@lw64 I was just using PNGTuber+. I get the idea and the basics of like slicing up the model etc now so my next step is just getting the same basics here and then eventually working my way up to deforming meshes etc
@danirabbit interesting. configuring a 2d character sounds like a lot of work to me. I looked once in using vpuppr with a model create with vroid studio. vroid studio is properietary, but you can create models for free and use them freely.
Add comment