Introduction
This tutorial will focus on how to create content to use in foto2vam to allow you to come up with a good result and have a nice base model that will inspire you for original content.
It means I won’t cover how to install or convert your photos with foto2vam. The subreddit has a few tutorials about that. You can find some links in the external resources chapter at the end of this tutorial.
I will put an archive of the last release of foto2vam I’m using for windows in the downloads.
Initially I wanted to talk about look-a-like, but after months of usage. I realized that just like in video game or movies, creating a good “replica” requires a lot of work, and in it’s current form foto2vam is not good enough to allow a “two click clone” like a lot of users think it would.
If you're interested in good looking replicas, I would recommend vecterror's work. Or even to support him on patreon. His work shows how much time and dedication you need to get a good look-a-like. And I'm 100% sure that he is not using foto2vam.
The software
foto2vam 1.11
Available as a download in this tutorial for convenience with a folder containing my sources for Liv the character I built this tutorial upon.
The original software was released by hsthrowaway5, and forked by multiple developers. I cannot even find who did the 1.11 release. I would gladly credit him/her if someone knows.
What to expect
Before starting to search for your sources and thinking that foto2vam is a magic software that will transcribe your quick screenshot of a character in a tv show into an astonishing replica that will work inside VaM… stop.
foto2vam is a tool that will allow you to get a base character to work with. If you are into good or very good quality characters for VaM… it will not get to that with only a couple of pictures.
You will have to work on your character after getting it into VaM to get to a great result.
With that out of our way, let’s check out the the process I’m going through
Process
I have 7 major steps in my process listed in the picture below.
Until I’ve reached step 6, I can always come back to step one to generate another iteration after changing some of my source pictures if I’m not happy with the results.
Find your pictures
So, we want a great base to work with. Let’s feed foto2vam good sources from the beginning.
We need to prepare two pictures, one that is called angle0. This is a picture of your character’s face facing you. One that is called angle35, which is a picture of your character’s facing you but with her head turned about 35 degrees to your left (meaning if you were at her place you would be looking at your right).
Angle 0
For this shot, pay attention to your orientation. Be sure to have a shot that is the most straight possible. Meaning : when the picture was taken (or the character created / posed) your camera has to be on the same level has her head.
Having a perspective issue could lead to misinterpretations by the software which will impact the result.
Angle 35
For this shot, respect the same rule as the angle 0. But one thing that could improve your result is to test some angle that are way more exaggerated than 35° (but not too much).
Here are the rules I use for the angle0 and angle35 that I consider my main pictures.
- Have a similar resolution ( 300x300 pixels or more )
- Have a similar ratio for the face and head ( approximate same size )
- Avoid big shadows and details on the face
- Avoid harsh shadows and prefer a soft light that still keeps the details of the face visible
- Clean pictures : not too blurry, not too grainy and so on
- Neutral expressions or if you don’t have, the most neutral possible
Now that our two main pictures are ready. Let’s feed a few additional shots to foto2vam.
We can now be way less demanding on how we treat the pictures, but let’s keep a few rules to get the best out of them :
- Same approximate resolution as the main pictures
- Good quality pictures
- Neutral expressions or if you don’t have, the most neutral possible
If you get this, throw away the picture and replace it with another one.
Generate your json file
You just created your folder with your pictures inside, launched foto2vam and waited for the conversion. Here we are, we have our result.
With 1.11, you have two options : use the Output folder or the Output_Merged folder. The difference is : Output_Merged merges the result with the base character of VaM.
I do not like using the merged version because it kind of skews my perception of the result I get. I prefer to work with the base model with no hair and start the next steps.
Open it in VAM, add some light
Load your scene. Start by adding some light. With VAM 1.20.x you can now load the subscene with the 3 light setup. It is perfect to tweak lights on the fly and see how the face of your character behaves depending on the lighting.
This is what I get.
Not bad. I do see a lot of genetic similarities with my source choices in my source pictures, I’m happy with the result.
But wait… you did the same thing and your result is bad like… bad bad. Why ? This is the magical power of tutorials : my pictures were not working as good as this on the first try. I had to iterate a bit through different pictures and find the best ones for the angle0 and angle35.
With the 1.11 of foto2vam you can input more shots which will give you better results. But still, I can’t stress this enough : you will often have to try a few times before getting the result you want.
As I was saying earlier. Until I’ve reached step 6, I can always come back to step one to generate another iteration after changing some of my source pictures
Check all skins and try a few haircuts
The second step is important too. You have to select a skin that you find great. It can change drastically the appearance of a character. Sometimes, even if my first impression of the base model is bad, I do try to switch skins, I’ve been amazed how much it can change.
I’ve played around with the different skins and a few haircuts here is what I like :
Try a few adjustements...
Now that we have a nice setup. Time to tweak a bit.
Try a rough pass on major morphs for the eyes, mouth and nose. If by just tweaking rapidly, you start to find something that you like… now is time to work a bit more in depth. BUT, if you’re having doubts, don’t hesitate to go to step one and try another iteration.
Refine, detail, finish your character
Here comes the hard part. Basically, the tutorial is finished. The main process to get here is over and now you are just going to tweak your morphs and get where you want with your base model.
I’m not gonna lie to you, the first part of the process is the easiest and the quickest when you are used to your workflow, aka : getting and preparing your pictures, trying the results with VaM and setting up your initial model.
Your final goal is what is going to make this final step longer. If you want to make something really close to your source material, you will have to work a lot with the morphs, maybe create a custom texture or a few decals here and there. Create customs hairs and so on. If your goal is to create an original character inspired by your source material, you have less restrictions and more creative freedom.
During this process, I’m saving a lot. Generally, I’ve got at least one save file per “feature” (nose, chin, cheeks, mouth, eyes, head… etc). Don’t be afraid to do it… when I wasn’t doing that, I’ve ruined a few models by modifying way too much some features of the character without having the possibility to go back on an older save.
Anyway, if you are here, you already know about morphs and how to use them. So here are the way I’m working when I try to get closer to a source picture : OnTopReplica.
OTR is a free software that allows you to overlay any window on top of another. Let’s just say that this way of working is only for creators who are working on desktop.
You can find the link at the end of the tutorial.
The mechanic is really easy :
- First, save your current character and pose it like your source picture ( you can use whatever angle you want, you’re not foto2vam )
- Load your source picture in OTR
- Tweak your morphs to get closer and closer to your source picture
To avoid having to reposition myself to be aligned with OTR after checking. I’m generally saving my position, saving my file. Then I check my angles. If everything is fine… I reload the file.
Do this process, as much as you want… as soon as you’re ready, you’re on the final stage…
Add a few clothes, save… and share!
You like your model, now it is time to make some beauty shots and to share it.
For Liv, NoStage3's haircut does a lot of the job... please, check out his work.
Conclusion
Even if a lot of people would like foto2vam to be a magic tool to output perfect clones based on a few photos, it is not.
Making exact replicas takes a lot of time, a lot of tools that are way beyond the simple use of a software like foto2vam that anyone can understand. And even with this kind of tool, you need to put a bit of time and effort to get a good result.
For me, foto2vam is a perfect tool to create a base for my characters that inspire me and come up with original ones. For instance here’s what I use to create nice bases :
If you’re crazy you could even gather a few sources from these websites then put them inside portraitai.com to get even more different results.
Don’t be shy, don’t hesitate to mix and match your source pictures with similar faces… you’ll get some awesome results just by trying different combos. You can even try like I did for Liv to mix and match real pictures, CG and AI generated faces… be creative.
Liv
You can now find the released character from this tutorial here. And also a gallery.
External resources
foto2vam releases
https://mega.nz/folder/OXgXiaRC#0SbdCBmlDtu2VEkFkKxPAQ/folder/Ka5ikC7I
foto2vam miscs threads
Thread foto2vam 1.10
Thread training foto2vam models
Thread foto2vam release 2
OnTopReplica
https://github.com/LorenzCK/OnTopReplica
