Marahzen's Musings on Creating Characters (and other matters)
paulawp (marahzen)
Posts: 1,371
Through most of Q1 each year, my time is otherwise occupied with a recurring gig that I do, but every year around this time, as I get near the point where I can proudly hand over the final product ... and return to life, already in progress ... I start making grandiose promises to myself about what I plan to accomplish before another gig season starts at the end of the year. In recent years, those promises have been centered around finishing my work-in-progress book. I have been trying to finish the second book in my story series for the last three years, which coincides exactly with detouring into an attempt to create art for said book. I decided to try a different tactic this year. This spring, I'm going to document some of my journey in trying to create characters in hopes that it will inspire me to finish some of these characters already and get back to work writing the damn book itself. If anyone else finds it interesting or useful, feel free to join in. I don't claim to be an expert or even necessarily good at it - but I've put a lot of time and effort into the subject, and I've tried a lot of ideas and products along the way.
Inevitably, it's also a retrospective on a newbie learning to use Daz. All I can say is that I did the best I could with what I knew at the time.
Just briefly about me, I have zero background in art. I am - or at least, I once was - a fiction writer. But life happened and I put away a lot of stuff I used to do, like write stories, on a shelf for years and years. Once in a while I'd twiddle with an old project, or write a few notes for a new idea, but I hadn't done any meaningful amount of writing in some 20 years when suddenly in the spring 2020, I had the urge to pick up an old project, revamp it, and work on it... and I finished it that fall. Or at least, I finished the first of what is now planned out to be five books in the series. The thing that inspired me to finish this book that I originally started - not joking - in the late 1980s was stumbling across Artbreeder. I started tinkering with Artbreeder portraits and suddenly, there were characters that I had envisioned for decades and had just been unable to get out of my imagination and into reality because I stink at drawing. Just having those simple 2D images was enough to help make these characters real to me, which helped inspire me into finally finishing that book. But then, partway through writing the second book, I realized that I wanted more than a single, static 2D image. I wanted to make 3D characters. And thus began my humble foray into Daz.
Daz 3D is part of 
Connect
DAZ Productions, Inc.
7533 S Center View Ct #4664
West Jordan, UT 84084
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.
Comments
Before I get to Daz itself, I should start with a conversation about what got me here to begin with: Artbreeder. Even now, four years later, Artbreeder still has specific features that are more helpful to my purposes than any other comparable site that I know of. There are other places you can get a 2D image to create a character from, but Artbreeder's claim to fame - as its name implies - is an array of tools to reproducibly and predictably fine-tune a starter image. If you have in mind an idea of what a character looks like, or you need to create a bunch of relatives, Artbreeder is a tool that can do the job.
Here's a quick tour of my favorite Artbreeder features - shown with the old-school legacy interface that I prefer.
The main screen shows the currently selected image and any child images (here, seen at the bottom). There's three options. You can edit genes, as seen here, which is fairly self-explanatory: make the current image older or younger, more masculine or feminine, etc, with precise dials; no futzing around with prompts. Save if you get something you like.
You can do children just from the current image. It will create random variants til the cows come home, and you can save any that seem interesting.
Or you can crossbreed. You can crossbreed multiple images elsewhere or here, just pick an image. In this screen shot, the current crossbred image is at left, with the original image in the middle and a random other image from my collection picked for crossbreeding. The dials determine whether the resulting face is more like one parent image or the other, or the tone, hair and other stylistic features are more like one or the other. Again, you get reproducible and predictable results that you can save and work further with if you get something that's a step in the direction you want to go.
And all that saving, whether you're crossbreeding, using children or genes, makes it so that you can fine-tune, for generations upon generations, to move toward what you want. This is a pedigree of that particular character, where you can see that it's a product of a previous version of him crossbred with a character that caught my eye, created by some other random Artbreeder user.
In doing this, you can create older/younger versions, relatives who are different but bear a resemblance to each other, etc. Artbreeder is great if you want relatives, variants, and have an idea in mind of what a character looks like.
If you just have a generic idea for a one-off character, one of the other sites, like thispersondoesnotexist, could be best. You don't have to sign up; just refresh until you get a face you like. There are new AI sites popping up every few minutes it seems like, and they could be great, too - I just stick with what works for me.
And that's all it takes to get a source image.
OK, no, it's not, but that's where I was when I set out to learn Daz to create characters.
Actually, I had bought Poser for this very purpose way back in the Poser 5 era. I installed it and upgraded it once in a while and never used it once. Also long ago, I installed Daz and likewise never used it until somewhere I found out that one could use FaceGen to create a character from an Artbreeder image. In late May 2020 I bought FaceGen* and fired up Daz long enough to create a sufficiently decent character to be a proof of concept that it could work. Then I went back to writing until the last week of May 2021 - almost exactly a year later - when I was ready to commit to learning to use Daz. And I discovered to my horror, that now that I was ready to get serious about this hobby, my characters were inexplicably terrible, looking nothing like they had a year earlier.
[*In May 2020, I was able to buy FaceGen from the Daz store, but now, one needs to go to the company's website and obtain it directly from them. It's an external link, but it's easy to find.]
I was wandering through Dazland, completely alone and bereft of clues about what I was doing. Luckily, IRL, I do software support (and training, and documentation, and QA) so I was able to work out the problem. I'll tell you what the problem was, but first, let's take a look at FaceGen.
FaceGen is pretty straightforward to use. It's an external executable which lets you select a source image from which it creates textures and a shape for Daz. This is my current FaceGen, which is the latest version - upgraded to create G9 characters. Otherwise, it was functionally the same product back then.
Basic usage of FaceGen is just to select the image, set the dots as suggested, then create it. Once created, you can export it for Daz to use.
The really cool thing that FaceGen does is to allow you to select the base maps to use. (It doesn't make as big of a difference with the face itself as one might think, but the rest of what it produces will take on the characteristics of the mat files you pick, which can be anything you have that's correct for the Genesis version you're creating.)
When done, it offers instructions on what to do next. Easy peasy.
So there I was in my first hours of using Daz, and I ran right into this. (G3 at left, G8 at right)
(That is a recreation of the dilemma I ran into; I'm onto my second dedicated art computer so those earliest files are long gone, or essentially unusable given path differences from that first installation on my workhorse PC.)
I discovered that the problem only occurred if I created a G8/G8.1 character. It turned out that a year earlier, I had tested FaceGen by creating a character with G3. When I figured out how to compare the surfaces of the two figures, I realized that G3, using the AoA_Subsurface shader, looked reasonable. The problem was when I returned a year later, I started creating characters with G8.1 and its mysterious Iray Uber shader.
Yes, I actually did consider creating G3 characters, but ultimately, in my first days of using Daz, my kick-the-can solution was to use AoA_Subsurface shader ... on G8.1. I planned to circle back later to revisit the subject when I was better at using Daz, but in the meantime, I wanted to start creating stuff - and from the beginning, I wanted to use my characters, not someone else's characters.
The best I could muster in June 2021 ...
This one, from September, I consider to be my favorite of all of my FaceGen era images. It's also one of my first ambitious efforts with a new BFF - Ultrascenery. Please overlook the various newbie issues as I struggled here with shader and dForce situations, and placing of the characters on this new-to-me USC surface.
Then, in October, I switched to Face Transfer.
This has been a super interesting read. I'm always fascinated by artists' reflections on their processes and motivations. Looking forward to more!
Before I talk about why I turned my back on FaceGen and walked away, I suppose I should start off with a conversation about Face Transfer. [Here of course, I mean the "original" Face Transfer Unlimited, vs Face Transfer 2, which recently became available. Despite sharing a name, FT2 is significantly different in output than Face Transfer - a game changer that will get its turn in this story when the time comes.]
https://www.daz3d.com/face-transfer-unlimited
I had actually bought Face Transfer months before I first put it into serious use in late October 2021 - back in the summer, around the two-month mark of my Daz career - but like many people who try Face Transfer once, I didn't realize the critical necessity of using Face Transfer Shapes, without which you cannot reasonably use Face Transfer. The dreaded "Voldemort face" ...
Oh, and you also have to be really good at Photoshop, which I was not.
For me at the time, it was not the only Daz toy that I'd bought that I totally didn't get how to use, so I shrugged and moved on, figuring I'd revisit later on when I was better at Daz.
One day, the time had come to go back and take another look at Face Transfer, and this time, I tried Face Transfer Shapes - which solves the problem of the freakish flat profile.
https://www.daz3d.com/face-transfer-shapes-for-genesis-8
And while this particular source image was not the best one to start with, I noticed that where it wasn't badly discolored, the skin was a much more reasonable tone than my initial experiments with just replacing the Base Color mat on a stock G8/G8.1 figure. I tested with others among my major characters and some were ... well, not as bad as this one.
But also at this moment, I stumbled across a product called Altern8, which I had just picked up during a sale.
https://www.daz3d.com/altern8--skin-shader-system-for-genesis-8
I had no clue whatsoever how it was supposed to be used, and I didn't have any of the characters it references (at that time, I had no use at all for characters and had only bought a couple of Pro bundles, just to - well - have all of the parts, even if I never used them for anything). I bought this Altern8 thing and for some reason, by lucky chance, I had a newly created Face Transfer character and after I randomly clicked a bunch of options, I had something that, on a good day, was a quantum leap from my original generation of characters. I wouldn't say I had mastered the mystery of Iray Uber, but at least I finally had characters that used it.
I'm not even going to discuss the steps I followed, other than to note that it was totally off-label usage of the product that the PA surely wouldn't endorse.
But, it worked for me for the better part of the next year.
This was created shortly after the switch to FT/FTS and my new approach to finagling FT textures. I did what everyone does - the classic group shot...
(Yes, it's dark ... everything these kids are doing is under the cover of darkness. It's not a happy story.)
A couple of favorites from this era of character creation ..
So ... FaceGen. That singular thing that literally brought me here and sustained me through five months of newbie fumblings, allowing me to learn Daz while using my characters, not someone else's characters, just like I wanted.
Mostly, I broke up with FaceGen because Face Transfer - once I figured out the Face Transfer Shapes thing - produced a better starter character for me. At the time I still had no idea how to fix a stock G8 character simply updated with FaceGen textures - which looked like it had been staked out in the desert for six months. When my hacky Altern8 workaround fell out of the sky within days of revisiting Face Transfer, I took that chance new pairing and ran with it. I'm always looking for newer or better ways to do something, and if I find one, I move on, with no hard feelings about yesterday's "new better."
But to be honest, I also left FaceGen because there was another problem that was starting to bug me.
I am the kind of person who is glad to rave about things I love; who is happy to talk about what works for me. What I don't want to do, especially in context of a hobby that is supposed to be fun, is clutter it up with the negative energy of complaint or of being unhappy with things. If I don't like something, I'd rather turn my back and walk away than linger and be unhappy with it.
So that said, here it is: FaceGen creates wide squirrel cheeks and sometimes long faces/noses, and as time went by, I ran across more and more characters that just looked terrible as FaceGen characters. At a certain point, I finally squirmed and edged away from blaming all of it on my newbieness. Some of them were fine; in fact, some of them I like better in some ways than their subsequent FT/FTS version.
And then, some of them looked like this one.
That just wasn't going to do.
So I moved on to FT/FTS/Altern8 and produced a version of him (and most of the others) that I could live with.
While Face Transfer/Face Transfer Shapes resolved some issues I had been facing, it introduced a big new one, which has already been mentioned (and seen!) in these ramblings: ugly relics on the resulting texture files. FaceGen produces much cleaner texture files than does Face Transfer Unlimited; and what I began to refer to as "Face Transfer stains", stretches and washouts now stared at me front and center.
For reasons that were (and, frankly, still are) unclear to me, my Altern8 hack seemingly helped minimize the problem a bit, but on a closer look, most of my new characters had stains on the side of the face ...
... or obnoxious light/white areas ... (the white around her eyes and cheeks)
... or stretch marks. (you'll see literal stretch marks in various places on the side of his face)
These are relics from the source image and the transfer process. The stains, most commonly on the lower half of the side of the face, are from shadows on the source image and the white areas are from highlights. These are normally present on any image that looks like a natural photo - it looks weird without them, in fact. The stretches seem to represent transfer issues from the edges of the source, often areas that are a bit out-of-focus or hazy.
On top of those predictable trouble spots, the original Face Transfer was very finicky on what it could handle. Remember my snarky comment way back about getting a source image? In fact, I had to do a bunch of updating of some of my original source images from 2020 to make them more FT-friendly. To create a starter set of textures from a source image with the original Face Transfer, you have to start with a head-on image, with a neutral expression, the right kind of lighting, and no glasses, no long hair, no facial hair ... nothing that gets in the way of the basic face that you wouldn't want as a permanent part of the textures.
I started off with rudimentary skill at best with Photoshop, and thanks to the University of Google, spent a lot of time reviewing training videos to teach myself how to use the various tools, to create masks, and perform various other Photoshop wizardry. If anyone cares, maybe one day I'll illustrate my favorite Photoshop tips and tricks that any total amateur* can do but as a general comment, I'll just say that I've done my time of fixing source images and the resulting textures. [*I was and still am a total Photoshop amateur. I'm hoping that one day I'll cross paths with someone who isn't, who can teach me how to properly fix these things.]
For a long time thereafter, my character creation process hardly changed at all. Toward the end of 2022, I retired Altern8 in favor of a custom process for doing skin textures that a friend who was pursuing his own Daz projects had created. But other than that, the year went by without much modification to my character creation.
And then everything changed.
Two gig seasons came and went, and spring 2023 found me at a crossroads. Genesis 9 had come out the previous fall, and I had by then invested in some G9 shape/morph utilities and even a few commercial characters. (Not for the characters, but because there were times when there was such an attractive, stacking deal on the bundle that it was worth it to me, even if I never used the characters themselves for anything.) I was getting close to two years spent on Daz, and I hadn't actually created any new final characters in over a year. I think I had been looking for that "next great thing" and the arrival of G9 threw an additional complication into the mix. Of course, 2023 dawned with no easy path to creating G9 characters but I knew it was only a matter of time before that changed, and I was going to have to decide whether to stay the course with G8 or switch to G9 (or, possibly, use both).
Of the major players, it was HeadShop that was first out of the gate to offer the ability to create G9 characters. I had not done anything with HeadShop yet - I had not strayed from Face Transfer since switching in late 2021. Given a lucky opportunity to test drive it, I found it was an useful experience and created a few interesting characters. But for all of my careful, evolving procedures with creating characters in G8, I had almost no tools in the toolbox to reach for in creating a G9 character. It was, in so many respects, like starting over again.
This one is retrospectively my favorite of the HeadShop creations. It's a cute variant of one of my major characters, but it wasn't "the thing" I was looking for, and I moved on to a new challenge - converting my G8 Face Transfer shapes to G9.
Sorry, meant to say hello and thanks! I wasn't sure if anyone else would be interested, but really felt like I needed to stop and take a backward glance at the journey I've taken so far.
No sweat! I figured you were just busy working on more posts, which is fine by me. I'm still having a ball reading this thread, but don't have much to add, for the same reason that it's so interesting to me -- I don't use any of these tools (other than DAZ Studio itself, of course) so you're giving me a window into a workflow that's unfamiliar to me.
I'd like to echo Ninefold and express my appreciation of your efforts here, Marahzen. I mostly lurk on the forums, but I'm here several times a day and I've read all of your posts in this thread with great interest. My origin story is a lot like yours. I came to Daz with the desire to create my own characters and use my art for my book covers. After so many years of cobbling together stock photos and never being satisfied that the results represented the images of the people living in my head, I'm finally able to create what I want to my perfectionistic specifications. I'm always fascinated by the journeys other artists take and I love seeing their processes in achieving goals similar to mine. Like you, I don't think I'll ever be done tinkering with details, but I wanted to let you know that I'm reading along. Thanks for this!
Hello! Welcome to my little travelogue!
In the spring of 2023, there was chatter on multiple threads here at Daz about creating G9 characters with the major character-creator options (FaceGen, Face Transfer, HeadShop), and I found myself having exchanges with a committed FaceGen user. He had a couple of fair points on his side: Face Transfer Unlimited had not been updated since it came out in 2019, and there was no obvious sign at the time that it would be updated to do G9. Meanwhile, FaceGen was still in active development, and its developers had specifically indicated that a G9 version was in the works. Moreover, FaceGen created cleaner textures and also had the option to pick custom mat files.
It was something that caused me to rethink my process in a fundamental way.
Having given a fair test to all three options - FaceGen, Face Transfer and HeadShop - I continued to prefer the shape created by Face Transfer/Face Transfer Shapes, but I agreed that FaceGen produced the best textures.
And therein were the first seeds sown that changed up everything by the end of the year.
I wish I could say that, having opened my mind to the possibilities, the pieces fell neatly into place and I began creating the characters of my dreams shortly thereafter, but that's not what happened. Real life is messy, so what really happened is that I spent most of the rest of the year scrambling around, trying multiple new products and processes, traveling down different avenues of experimentation at the same time as I continued to test both G8 and G9.
I also was otherwise occupied at the time in buying everything in the Daz store, which is part of a different story. It did provide an ongoing excuse to create new characters, though - a constantly revolving cast of what I came to refer to as "insta-people" who I would quickly create - via whatever creation process I was trying out at the time - to use for testing new stuff ... of which there was lots.
Early in 2023, my "insta-people" were constructed something like this one. I started with a stock G8.1 figure with a Face Transfer shape. I'd apply the skin settings that I was using at the time, and then update the textures. After all the time I'd invested a bit earlier in Photoshopping the Face Transfer texture and creating my own bump and translucency files, all insta-people got were the original Face Transfer texture and base male or female bump and translucency files. All those morphs I'd bought - and I didn't bother with them for these mass-produced characters, unless I was testing a new morph.
There are better examples than this one - truthfully, I haven't made a character this way in a long time - but it only took minutes to make one and it ultimately had one job, for which it was good enough: testing new Daz stuff.
In the spring of 2023 came an entire reconsideration of skin textures, and the possible role of my old once-spurned friend, FaceGen.
Of course I was aware that most - ok, yeah, probably all - other people who use Face Transfer substituted textures, generally with textures from Daz characters, and I'd firmly refused to do that. After all, the stated mission of my whole operation was to create and use my characters, not somebody else's characters. But it was only in context of these conversations with the staunch FaceGen supporter that it occurred to me to try using FaceGen textures of my characters on Face Transfer shapes created from the same source images.
The last time I'd tried to use FaceGen textures - in Iray anyway - it was a useless disaster, but it had been two years since then and I had new tricks to try.
First I tried to substitute FaceGen textures in place of the stock Face Transfer textures in already-created characters, like this one. It's different than the original, when you can compare side-by-side, but not objectively better:
But ... it proved that it could be reasonably done, so I tried further experimentation over time.
I got a lot of mileage out of new product 1-Click PBRSkin.
https://www.daz3d.com/1-click-pbrskin
By the time I picked up 1-Click PBRSkin late in the summer, I'd been doing a lot with G9 and as my testing on G8 characters continued in parallel with G9 characters, I wanted to test out PBR on my G8-era textures. This is a quick touch-up of the last one, after converting to PBR:
The end point of testing in G8/G8.1 was making use of the same technique that I had been using in G9. I would create the shape, apply it to a stock figure, apply the mat files of a commercial character, substitute the base color mat with my FaceGen-sourced textures (in some cases, using the textures of the character whose mat files I had applied, since FaceGen lets you do that) and then convert to PBR.
Still, it's an insta-person. It took me a few minutes to create these, and off they'd go, to help test out new stuff.
That series of examples was just whipped up yesterday, since there was never a real-life moment in which I methodically worked from beginning to end with one example. It was more organic than that. In fact, there wasn't really a true straight line from one process to another, as I kept circling back and retrying old things even as I was testing newer ones.
There were better renders along the way. Here are a few of my favorite G8.1 Face Transfer-based insta-people images from 2023 - my year of wandering and product testing. The earliest ones - the bathroom scene and the girl on the horse - used Face Transfer textures and the later ones, aside from the young mother, use FaceGen textures.
This character is really elusive to capture as a 3D character. This version of him is not especially recognizeable, but I still like the render itself.
Under Sun-Sky lighting conditions - important for me, since a lot of my story takes place outside.
OK, the Mechasar baby is not a Face Transfer product.
I just realized that almost all of these - and my example from yesterday - feature a girl that is ultimately from the same source image. Don't worry, I have dozens and dozens of NPCs - I literally call them that - and they get a lot of use with a new product that gets mentioned soon.
Meanwhile, like lots of other Daz users, I was busy exploring G9. There had been so many truly great bundle discounts on the G9 characters, that despite never planning to use the characters for anything, I had practically all of the basic human* G9 main releases. [*Not to be ... well ... some sort of "ist", all of my story WIPs involve physically ordinary people in ordinary worlds, so I have no need for - or interest in - non-realistic humans and animals. That's why there's almost no presence of toons, anime, mythical creatures, monsters, aliens, gods and demons, elves and their fantasy brethren, etc, in my stuff.]
A look through my gallery shows a handful of renders that are just simply G9 characters around that time; they were something new to be tested like all the other new stuff, and sometimes something came out that I liked. But my main priority was getting my characters onto a G9 figure, and when I first broke through with success in that regard, it was thanks to the information here:
There's a bunch of steps involved, but it can be done. She was my first converted character:
And then this guy, who is based on the same source image as the guy in the bathroom up in the last post.
When the RSSY G8/G9 converter came out, I had a working manual process so I thought, I can just do it manually. By the time that ManFriday's Figure Converter for G9 came out a bit later, my response was, "Yes, please." I was ready to try a utility:
https://www.daz3d.com/figure-converter-for-genesis-9
It was vastly quicker and easier than doing it manually.
This was one of my first G9s created using this utilty.
Within a short period of time, everything I needed came together. Having sufficiently convinced myself that I could reasonably use FaceGen textures on my Face Transfer shapes, I went ahead and updated to the new FaceGen that does G9. And that was how I started creating my "Franken" characters. They were Face Transfer shapes converted to G9 with the ManFriday converter, with textures provided by FaceGen, run on the same source image and exported straight to G9.
Another handy thing I picked up around the same time was Dial Fusion for G9:
https://www.daz3d.com/dial-fusion-genesis-9-edition
I would then immediately create a dial for my newly converted G9 shape and save it, then load a fresh stock G9 figure, dial the shape, apply the textures and ... voila, G9 insta-people.
This was where the final step of using store-bought texture underpinnings - that tactic that also crept into my G8 characters later in the year - came into play. The technique I had been using in the first few months of the year was specific to Iray Uber, and meant for G8/G8.1. There was a moment sometime along the way when I realized that I could just apply the mat for one of the many store-bought characters I'd accumulated when buying all those bundles, and then go back and replace the base with the texture file created by FaceGen.
Yeah, eventually I plan to go back to creating my own translucencies, bumps/normals, etc, just as I plan to analyze what settings go into skins that work best with my process and create my own. But through the latter months of 2023, when I was streamlining all of this, it made usable test characters.
For the first year or so, I had been creating characters in small numbers, but during the "new stuff testing season" of 2023, I created dozens of characters, both G8s and G9s from converted G8s, and the telltale "look" of Face Transfer characters started to nag at me more and more. Through the year, I began to tinker with shaping products - not in an attempt to create different characters but to introduce some variation to the stock cast of characters I had.
Of course I had bought morphs - lots of morphs - from the beginning of my Daz dabblings, and back in the early days, had earnestly done lots of twiddling with the various morph sets as I tried to fine-tune both my original FaceGen characters from the first few months and the first generation of Face Transfer characters that started appearing late in 2021.
But as I expanded the use of conversion to G9 and creating dials, I started trying all sorts of things to dial down the recognizeable "FT" element of my characters. I started generating shapes from the same source image in multiple transfer utilities and then experimenting with characters that were less than 100% Face Transfer, in which I'd dialed in a touch of the shapes from dFace - another transfer utility that was around for a while in 2023 - and even FaceGen itself.
By the end of the year, I was making larger use of a resource to which I really hadn't given any previous thought. From all of those pro G9 bundles I'd been buying, I also had numerous shapes. Character shapes, elder shapes, portly shapes ... all manner of shapes and "kin." I found that they were perfectly usable on my Franken characters - though they had to be used judiciously as they were optimized for a completely dissimilar figure and looked freaky if cranked up in large amounts.
These two are G9 Franken characters with 100% Face Transfer, with some shapes I took from characters:
... varying, 40-60%, Zack Elder and Zack Buff shapes for the body and face.
... about 20% Dain (also, some 3DU Time Machine).
And then, of course, everything changed again from one minute to the next with the entirely unanticipated release of Face Transfer 2 in late November.
https://www.daz3d.com/face-transfer-2
OK, it wasn't actually available on the day it appeared in the Daz store, but it wasn't long before a bunch of us were happily testing and posting the results of our experimentation on a Daz forum thread that remained active for weeks.
https://www.daz3d.com/forums/discussion/661746/face-transfer-2/p1
It was immediately clear that FT2 was a significant improvement over the original Face Transfer. I'd spent two years finding ways to address the many foibles of FT and they were instantly gone with FT2. It still doesn't produce a really usable image if the source isn't a head-on image, but it shrugs off some of the other stuff that its predecessor could not handle at all. While you're arguably still best off starting with a traditional ideal source image - neutral expression, good lighting and nothing intruding on the facial surface like glasses, hair, facial hair, hats and such - FT2 can produce a usable output even with those problems.
Obviously this one needs some more work before he would be ready for prime time, but at least there's a starting point. In the original Face Transfer, this would be an non-viable source image.
This is what the original Face Transfer creates, here after applying some level of Face Transfer Shapes to at least give him a human profile. (Obviously, no further work was done on the surfaces or lighting here.) The glasses are literally on the texture, as are the smile lines. You can go to posing and close the mouth, but it will never close correctly on the edges.
While the shape - in particular, the side profile - created by Face Transfer 2 is vastly improved over the entirely unsalvageable "Voldemort face" of the original Face Transfer, there are still some shaping issues. The profile was still rather vague and FT2 introduces a new weirdness that crops up sometimes - bad ears. Luckily, soon thereafter, Face Transfer Shapes 2 came out, offering not just a basic fix but also some additional shape starters.
https://www.daz3d.com/face-transfer-shapes-2-for-genesis-9
So from beginning to end, those were all of the major pieces that have gone into almost three years of working on creating characters in Daz from source images - and using the textures from the source images, which as far as I have seen, almost no one else does. Most of my time since FT2 came out late last year has been occupied with my side job, but I've still found a little time to tinker, and I'm still working on the path forward, given all of the tools that are available now. Here are a couple of my current favorite WIPs, featuring major characters in my story:
You've only seen her a hundred times in this thread - she turns up everywhere. She is the main character, so ..
He's much more elusive, and very rarely turns out either recognizeable or decent.
Those two were before Face Transfer Shapes 2 came along. This one is of more recent vintage and includes FTS2 and a smattering of other shapes as I experiment with just bit of variation to the FT2 shape.
And a few random scenes I created with FT2 insta-people while testing "new stuff", that aren't in my gallery (link is in my signature) .. Most of what I've done in the last few months is an endless parade of head shots with generic backgrounds, but there were a few fun product-testing images.
This one pre-dates Face Transfer Shapes 2; you can see an example of the "bad ears" I mentioned; also the heavy roundness of her lower face makes its vintage obvious.
IIRC, I was testing Dial Fusion for G9 at this point, confirming that there's no problem having two unrelated FT2 figures in the same scene.
Too over the top to put out in public, I thought. But, checked some stuff off the new-stuff list.
It occurs to me that for the possible benefit of anyone else who wanders across my ramblings and wants to see more conversation about the subject of FaceGen/Face Transfer/etc, I should include some of the other Daz forum threads I've been involved in, in addition to the FT2 thread mentioned above. Surely there's others, but here are ones I found in my bookmark list:
https://www.daz3d.com/forums/discussion/597876/face-transfer-on-genesis-9#latest
https://www.daz3d.com/forums/discussion/381026/facegen-vs-face-transfer-vs-headshop-12-pros-cons#latest
https://www.daz3d.com/forums/discussion/571066/face-transfer-so-much-better#latest
https://www.daz3d.com/forums/discussion/623396/can-any-facegen-users-help-me-here#latest
https://www.daz3d.com/forums/discussion/621496/face-transfer-unlimited-g8-expressions-morphs#latest
https://www.daz3d.com/forums/discussion/351636/is-there-a-face-transfer-thread#latest
https://www.daz3d.com/forums/discussion/421111/artbreeder-and-prototyping#latest
Since I last checked in, I have finished and handed over my side gig work for the 2024 season, so now I have a few months to play.
As I had documented the latest steps of my creation process, I figured it might be time to tackle that challenging and controversial topic - G8 or G9?
If you hang around the Daz forums, you're surely familiar with the main "Getting on the 9 train, or not" thread over in The Commons, which has been perpetually active since G9 appeared on the scene in the fall of 2022.
https://www.daz3d.com/forums/discussion/598286/getting-on-the-9-train-or-not/p1
Over the last year and a half plenty of people have made their case for why they are staying with G8/G8.1 (or even earlier) or why they are moving on to G9.
I came to the conclusion that whichever figure works best for a particular user, and his or her art and process, is the right version for that user.
I don't think there are many people who have invested more time and money in parallel testing of G8/G8.1 and G9 than I have and ultimately - barring catastrophic usability issues in terms of poses or clothing or something else of central importance in using Daz - the most important factor for me was which one made better versions of my characters. It was the same consideration that came into play when I went back and forth between FaceGen and Face Transfer, back to FaceGen (in part), on to new Face Transfer 2; and when I adopted and moved on from various approaches to skin textures.
So here they are - G8 (the brunette) and G9 (the blonde) from the same source image, using just FT and FT2, respectively, with 75% of the respective "female" fixes for each verson. (G8, being half a foot taller, is also scaled down to be a similar height and I've done routine head propagation tweaking on G8, which comes in after FT with disproportionately tiny heads; somewhere around 4-6% gets a match in dimensions to a stock G8F figure.) Here, I'm actually using the same skin treatment on both - I've applied a commercial G8 skin to both figures and then used textures sourced from FaceGen as the base.
This was the original source image:
I could do better, but the goal here was a quick comparison of Face Transfer/G8 and Face Transfer 2/G9 characters on the most level conditions I could reasonably muster. This particular skin material set, being out-of-the-box G8, is Iray Uber, which I haven't done so much with lately, and I didn't do any cleanup of the "stain" from the source image shadows yet. Putting a G8 skin directly on G9 is accomplished by use of the "Material Suit," which is pretty cool toy if you want to put G8 textures on a G9:
https://www.daz3d.com/rssy-genesis-8-and-81-female-material-suit-for-genesis-9
https://www.daz3d.com/rssy-genesis-8-and-81-male-material-suit-for-genesis-9
So am I planning on doing much application of G8 textures to G9? Not sure, but for a few bucks, I like having the option.
And as for the main question - G8 or G9?
I'm still not sure but have been leaning G9 for a while now. I just do not like the oval android head of G8F in particular; G8F face transfer characters also tend to have disproportionate lower face features, especially a heavy jaw line, and that's harder to morph away than the long head itself.
One day in the 1980s, I walked up to my English teacher and presented him with a stack of pages that represented some portion of what was then my latest, greatest effort - what is now the first incarnation of an idea that would eventually become my current WIP. After some amount of time, he returned it to me with his review. Since this was about 40 years ago, I don't remember exactly what he said about it, but it went something like, "This is good and that's good, but it moves a little slow." I accepted his gentle and well-intentioned constructive criticism with grace and took my story home and put it in a drawer, where it sat untouched for years.
Was he right? Of course he was. I write incredibly long, inaccessible books that follow basic rules of fiction writing only minimally at best - books with no audience whatsoever. No one will ever read this book, or any of the others. But that didn't stop me from writing it, all 167,000 words in the first book alone. (There's four books to go in the series.)
At a certain point, later than my teens to be sure, I asked myself:
Self, are you more interested in writing the book or having other people read the book?
It surely wasn't the first or even the second time that I asked myself this question that I realized that the answer was, I am more interested in writing the book that I want, even if no one else ever reads it. And once I was able to be honest with myself about that, I went on writing.
Writing fiction doesn't pay the bills for me, so I have no obligation to anyone other than myself when it comes to writing books - not on what to write, how to write it, or when to write it (and more importantly, when to finish writing it). I've seen my name in print - let's just say that realistically, there's no new life experience to be gained publishing a fiction book. The fiction writing is for fun. It's supposed to be for fun. And it is fun, until I question and doubt myself.
It doesn't surprise me that I sense myself going the same way with art: I see the earliest glimpses of art being fun until it isn't, when I start doubting myself.
I'm not really sure that what I create now is materially better than what I created a year or two ago. I know that I have done next to no creation of art for any of my books, hardly any since my earliest, most cringey days of Daz-dom. I created more story-related art in the first month of using Daz than I created in almost three years since.
I feel like I've lost my way.
Hi again Paula!
This last render so perfectly expresses what you say in your last sentence, that you must be doing something right!
The pose, the clothing, the DOF, and the expression, all convey that meaning so well.
If you were to do nothing else other than produce expressive renders like this, you would be creating real art that embodies emotion.
But of course the very act of producing this kind of art, is the fruit of the innumerable hours you have spent so far on characters, and also their expressions and posing - so there has certainly been purpose in your experimentation and tinkering, that is, honing your skills through this, although it might not have seemed that way at the time!
You are right in saying that it is only you who determines what you should do in, and with, your writing. Some people keep a journal (I do this at times), which is a way of getting down on paper, electronic or otherwise, what is inside their head and heart. I guess your writing is another way of doing that. Having said that, the day may come when you actually want to publish it; but such a day is clearly not anywhere near the present time, and there certainly is no obligation to do so.
If you were to begin again producing illustrations of the scenes and characters from your story, and what their feelings are, I think you would do very well at that!
Thanks for dropping by and checking out my ramblings. This has been an interesting exercise to stop for a look back at all the ground I've covered in my Daz venture to date. I do need to concentrate on being pleased with myself at how far I've surely come in that time, rather than let it trigger angst at the sense of passing time and the lack of output.
Hi Paula, this is a really fascinating and helpful thread I just stumbled on today - I do hope you will keep on exploring! I'm always amazed by your custom characters when you post renders of them elsewhere. I'm dabbling in the same area occasionally, but haven't gotten nearly as far. But then, I'm not nearly as persistent with it as you either ;-)
As to the writing - well. I'm not sure if you're even looking for advice at all, so please feel free to skip all the rest. But I can relate to your experiences I think, and to your feeling of being lost atm. I've been published as well, years ago, by smaller and also by one mega publishing house, and I hated 90% of the experience (the few good things, like lovely reader emails, are sure VERY good though). So I've come to the conclusion after some years that I'm just not cut out for the traditional publishing process, which is fine by me.
But the thing is, I believe stories want to be read, not just written. And this desire to reach out is, to me at least, also an integral part of the writing itself. Stories IMO want you to write them so that they can be told, and spread, and evolve. I understand that your stories are difficult to read, and do not follow conventions and are mainly a work of love for yourself, but I do feel that maybe this kind of closed-in system you're writing in is why you're stuck atm. From your willingness to show the characters you create to us here, I also get the feeling that there's at least a part of you that is actively trying to reach out and make your story heard.
Sooooo ... maybe rethink the not-to-be-published rule, and give the good old self-pub a try, where you can control most of the parameters of publishing? Even if it's just a handful of readers who find you, their reading will reflect back on you and your writing and help bring all of it into clearer focus. Maybe this could help finding yourself again? - And as an avid reader myself, PLEASE do, I'm drowning in booktok smutty dark romances and sadistic psychothrillers here ....