How Nana prepares their model for Live2D (Part 9)
Continuing from where we left off last time we are going to have a brief introduction to the texture atlas and we will export our model for the very first time.
Let's go through what the plan is here
We are going to talk about the texture atlas.
We are going to talk about the limitations in the free version of Live2D.
How to fix or add parts to your project.
We are going to have a first export of our model.
We are going to have a quick look in prprlive .
Without further ado, let's start.
The texture atlas
Now before you can start exporting your model, you first need to create a texture map of your model, this might sound familiar for people who've been messing around with 3D models before and the idea is fairly the same.
You can gain access to the texture atlas from the modeling menu under texture. As a free version user you will encounter a pop-up notification here, indicating that the free version only supports textures up to 2048x2048 px - don't worry about this, since under normal circumstances, 2048x2058 px is more than enough.
Setup your texture atlas file by giving it a name, and set the width and height to 2048 (or higher if you are using a pro version). Once you are done, click ok.
Now you will see a window where a preview of the texture will be generated. You can manually change the positions of the separate parts. Most of the time, you will notice some slight misplacing of the parts (aka they are overlapping). It is important to make sure that non of them are overlapping. So fix that first. You can also see if you've forgotten to add a mesh to a part.
It looks like that this part has been corrupted when I was telling Live2D to automatically create the Mesh. this is a great showcase of why this function cannot be fully trusted. ALWAYS check if it has created the mesh properly. As you can see, this part has a mesh, but the texture part has been lost and cannot be retrieved. but this is also the perfect opportunity to explain how to add new parts to an existing project. In our case it's replacing a corrupted part.
Replacing / adding new parts in your Live2D project
First of all, adding new parts to Live2D is fairly simple. you just have to drag and drop the .psd file with the new parts into Live2D while the existing project is opened. Obviously your new .psd file should only contain parts which you'll need. So in my case here I will simply create a new .psd file with only the layer of the now corrupted part.
As you try to drop it, you will see a new pop-up window asking if the .psd file you have dropped into Live2D is a new model or an addition to the currently open project. Select the project and click ok.
Next Live2D will ask you if you want to add all layers as new ArtMesh, click okay. This will add all the new parts into your project where you can continue where you've left off.
Once you've put the part into the right order and assigned it to where the original part was assigned to. Open the texture atlas one more time and add the part to your texture. You can also select automatic layout for the program to place it. But at this point, you should ask yourself if you should trust it. Double click on the part and position it. After 1 final check click okay.
You will immediately notice that the resolution of your model has changed. This is due to the resolution you've set for your texture. You can always see the original resolution by hitting the T-key, but be aware that this is the resolution this model will have once you export it.
Try to export your model now for the first time by going to file > export for runtime > export as moc3 file. You will now see some options, the most important one and the only one I will go into here is the export version. This is important depending on what you'll be using this model for, as some software out there might only support certain versions. I'll be using this model for Vtubestudio and prprlive, both support the newest built, which is why I'll be going for the SDK 4.0/Cubism4.0 here. Once you've selected it, click okay.
You can now name your .moc3 file, save it.
Go to your folder and you should be able to see the following files:
a folder with the texture;
a .cmo3 file, this is your Live2D file (07 - TCG Nana v3 in my case);
a .moc3 file, this is the primary file which will be read by Vtubestudio and prprlive;
a model3.json and a physics3.json file, which are secondary files which will be read by Vtubestudio and prprlive.
It is finally time to try out the model on the mentioned programs.
Prprlive is very simple to set up and one of the two popular face tracking software for live2d models out there. It was originally developed for streaming on bilibili and the software was originally in simplified Chinese. Due to that, there are plug-ins out there which aren't exactly suited for users who cannot read simplified Chinese.
However, the core software is fully available in English and has several advantages over Live2D when both are run on PC:
Very easy to setup;
A very good and responsive face recognition;
60+ fps (if there is enough resources available);
Tracks eyes separately;
Simple to use;
Voice tracking using vowels is possible;
Bilibili Gift options.
To be fair, I will also list some of the not so great traits of prprlive:
Limited options (for instances only 2 background, no props although there's a plugin in Chinese for it);
The good tracking can be a downside for certain setups (it's very sensitive);
Settings not as rich and detailed as Live2D;
Moving your model around requires a bit of a learning curve;
I've read before that the performance can drop or even make the keyboard experiencing delays when the stream is intensive (not confirmed for myself).
How to setup prprlive
As mentioned before, prprlive is super easy to setup. All you need to do is to follow these steps:
Click on the first button on the left side. Click on Load, select your model3.json file.
Click on the second button on the left side, there you can setup the physics parameter and animation keys.
Next, after the model has loaded, click on face capture (4th button on the left). You will have some settings there, but for now, just click on start capture. The capture should be working now.
If it doesn't work, check under select camera whether your webcam has been selected as the tracking camera or not.
Try the tracking out, if it doesn't look right, try out the calibration settings
You can check out the settings on the left out for some limited fine-tuning.
There's also a lip sync, which you can find on the top right of the same menu overlay. If your model was created with mouth movements for vowel movements, you can also set them up there, it works quite well if your model has that feature.
Once you want to actually stream, you will have to go to the video streaming menu (6th icon on the left). Install the virtual camera and the transparent streaming plugin for OBS. You may mess around with the other related settings there.
Disable the background with the first button of the lower row on the right side of the screen, this will turn on your green screen. If your character has green in their color scheme, you want to switch the background to a different color, that can be done in the video streaming menu.
And that's pretty much it, there are some other options and settings available in prprlive, but they won't be relevant for most users.
Moving your model in prprlive
If you have been using Vtubestudio and this is your first time using prprlive, there's something about moving your model which will suddenly become apparent, it won't respond to your drag and drop. This is similar to vseeface for our friends out there who are using a 3D model. In order to move your model in prprlive, you will need to remember the following key combinations:
Moving around: Drag and drop while holding down the Alt key and the left mouse button.
Resizing the model: Rotating the scroll wheel while holding down the Alt key.
Rotate the model: Rotating the scroll wheel while holding down the Ctrl key.
Here's a short footage sample of the model in prprlive (physics have been set to low):
This blog entry has been going longer than expected, which is why I will continue with explaining how to setup simple toggle features and give you a brief introduction to Vtubestudio in the next entry.