Daz 3D is part of 
Connect
DAZ Productions, Inc.
224 S 200 W, Suite #250
Salt Lake City, UT 84101
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.Daz 3D is part of
Connect
DAZ Productions, Inc.
224 S 200 W, Suite #250
Salt Lake City, UT 84101
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.
Comments
That wolf really looks great. Last time I saw somebody else's wolf the gait and weight of the legs looked off, but this really looks good.
Thanks for stopping by and commenting @TigerAnne and @Noswen - it's much appreciated!
Props go to Alessandro for the wolf. That other wolf may have been one of mine I posted while playing with IBL Master when it first came out lol.
- Greg
Would be really cool to see a short 8-12 fps animation sequence rendered with those algos!
That was my first thought, too! It would make a wonderfully unique creation, so unlike the plastic look of current mainstream animation.
I'm keen to hear Algovincian's thoughts on how well the algorithm outputs could work as an animation
Research papers I've read indicate that each frame has to use the previous frame as a guidance in order to smoothly transition without flickering.
I'm busy with other projects at the moment (my network is accounted for), but in the coming weeks I'll schedule the rendering of a couple of animations:
1. Some camera movement in the enchanted forest (to test the parallax effect):
2. a figure animation of some martial arts combos perfromed by Dead Blade
Since many different styles are generated as output every time I run the algos anyway, I'll create the animations in more than one style. I've been keeping animation in mind, so I'm fairly confident that there will be continuity from frame to frame - none of that dreaded shimmer from random stroking patterns, etc. . . we'll see lol
Here's an animation that was rendered in October of 2014 - one of my first projects, and not that long after I started using DS and the Genesis figure. I still remember being blown away by the abillity to morph the figure. The BVH run cycle was adapted (made seamless in time and smoothed using Excel lol) and applied to the Genesis figure as it morphed:
http://algovincian.com/temp/mrc.mp4
Hard to believe that was five years ago!
- Greg
Just catching up. Great stuff here, Greg. I loved the animation. Very cool! :)
Thanks, KM - I'm taking off for the 4th, so the plan is to get things cookin before I go. With any luck, all the rendering for the anims will be done and ready for assembly when I get back.
- Greg
Really diggin the B/W output:
Still finding it challenging when the subject matter hasn't got a ton of detail to work with, though.
- Greg
Looks great! My fav is the middle image but all look great. :)
Leaving tomorrow morning until next week, so I wanted to get a quick animation of DeadBlade by @daveyabbo cookin tonight:
https://www.daz3d.com/deadblade-for-genesis-3-female
I used 3 poses from the excellent Fan Ling set by Capsces to create the animation:
https://www.daz3d.com/poses-and-expressions-for-fan-ling-and-genesis-8-female
Here's a quick OpenGL hidden line render of the anim:
http://algovincian.com/temp/kick.mp4
Don't have any real experience with this yet, so figured it would be best to keep this test simple/short (because a lot is likely to change). As it stands now, I set the base stance at frame 0, a leg up pose at frame 10, the kick pose at frame 20, the leg up pose again at frame 30, and the base pose again at frame 39. The video shows the frames played back at 12 fps.
Of course the right foot moved all over the place, so I wrote a quick script to make the necessary adjustments to anchor the foot throughout the frames. I'll render out the analysis passes overnight and run the algos before I leave.
- Greg
So here's links to the quick & dirty animation in 2 different styles:
http://algovincian.com/temp/kick-color-12fps.mp4
http://algovincian.com/temp/kick-bw-12fps.mp4
Left them on my server for DL rather than uploading to youtube because the forced compression/size is so awful for any type of line art. Both are 1080p and about 5MB.
For the most part, the smoothing algorithms I wrote did a good job on the color animation. The lines are coherent/consistent/persistent from frame to frame, with the exception of some of the shading on the boots. Not surprised by this, as the shading of darker areas (the boots are a darker color) was anticipated as being problematic.
The high contrast B/W animation is more problematic.
The 2 animations were created by simply assembling the individual frames at 12fps and compressing. Now that the frames of this animation are all rendered for these 2 styles (and many more), I'll begin looking at including time in the equation to further "smooth" things.
- Greg
This looks too good. I'm calling the render police.
Love the animation tests! The color version is reminiscent of the videogame Borderlands, while the black & white version reminds me of Frank Miller's comic/graphic novel look in Sin City.
Your pictures are blowing my mind. Never had thought, that something like this is possible from a render. Truely amazing Art.
Edit: Just watched the two videos. Wow, just wow. As Kaleb said, it looks really like something from Borderlands or Frank Miller.
Thanks to you all for taking the time to post - it's much appreciated. The plan is to focus more on animation going forward. Given its time-consuming nature, it's more important than ever to make sure things are headed in the right direction.
The kick test has already provided valuable feedback with regards to both the raw rendered output, as well as subsequent processing. There's some time-warp functionality/code I've been developing, and the kick anim showed without a doubt that the deltas are to big at 12 fps :(
As soon as I can free up some cycles, running the algos on the camera motion test for the Noodles in the forest can begin!
- Greg
Absolutley incredible work! It's been fun checking this post as the images keep coming. The mummy character and the 2 demons were incredible. As you say the more detail is available in the normals and the bump maps you get so much more detail. What I like most about this is the organic feeling to the line work. Some of it reminds me of the art from 2000AD, which remains one of my favourite comics! Looking forward to seeing the animations
Thanks, @jepsonpeteCMT - your comment about the line work feeling organic is nice to hear. Many years were spent trying to make them look a little less stale than typical NPR output.
It's been my experience that, as far as comics/animation goes, 3D rendered output has been frowned upon commercially. Although this is just a hobby for me at the moment, the hope has always been some commercial applications eventually if it's good enough. We'll see . . .
- Greg
I'm blown away by everything in this thread. W-O-W
Greg,
Please...please tell me that at some point you'll release at least the D/S-compatible portion of this process to the rest of us plebs...the style your system generates would be a flawlessly-perfect addition to my graphic novel toolbox.
I get chills just thinking about it.
Wow! Those are outstanding, Greg! I love the animations and the other renders you've created since I came by. The black and white animation is super cool! The xeno-love image cracks me up - it came out great too!
To be honest, it's hard for me to get motivated to polish just the analysis pass shaders/scripts enough to release them as a product. Maybe if I ever get some free time and have nothing else to do lol. Really - it could happen ;)
Thanks, Diva. I've just finished rendering out a few frames of the Noodles in the forest - the cats are static but the camera is moving. The process of developing an animation workflow has just started and it's all pretty new still.
Work has started on some tweening algorithms to ensure smooth motion. "Cheating" with tweens generated post render is crucial because not only does it help with smooth playback, but it shortens render time (fewer frames), helps make the compression more efficient, further limits shimmer, etc.
In this short test, 102 frames were rendered, an additional 102 tweens were generated, and they were assembled & compressed for playback @ 60fps:
http://www.algovincian.com/temp/noodles.mp4 (21MB)
I didn't want to compress it too much because I wanted to judge the results of the tweening. It's so challenging because everything effects playback - the frame rate, the resolution, the compression, the data rate, etc. It's hard enough to get it to look/playback well on my computer, let alone different devices (which may have more or less trouble). Ideally, I'd like it to look and playback well on my phone as well as a 65" 4K TV.
Overall, not bad for an early attempt! There are some issues, but I am pleased with how 3D it looks.
Next on the agenda is to re-render the kick test at 30fps instead of 12fps and apply the same processing.
- Greg
This makes me both very happy and supremely sad. I'd love to be able to just play with the scripts and see what I could produce. Hopefully someday!
I can relate to finding it difficult to get motivated to do tht drudgery of preparing and packaging. It seems so non-creative a use of time. I think you have done some stunning things since last I commented here. The cats and the motion video are quite something. It makes me sorry I hadn't been back more frequently but I have many good surprises now, so there's that!
^__^
I still love the sketchy style of your work.
The cats are honestly some of the best quality stuff I have seen around here! That should be an animated short film. Is that what you are working towards?
-Rak
Thanks Rakuda - means alot coming from you. I only wish I could get the algos to sketch like the few examples you posted in your thread!
I'd love to find a commercial application for the technology so I could spend all day messing with this stuff. I've been exploring the idea of an animated short, as well as comics and creating 2D assets for games in Unity. That being said, they all require additional skills that are much different, and I'm not sure that I have. At the very least it would require a major shift in gears.
For now, since it's just a hobby for me, I can afford to be whimsical in how I spend the available hours working on it, which is what makes it fun!
- Greg
Hmm, interesting idea, filtering things down to an algorythm that is reproduceble. There certainly are steps to sketching like
etcetera...But, then again, I don't always follow the same path.
Those are things that could be set in an algorythm. I have a nephew who loves Python and is into the mathematics of such things. He likes going on about light rays and shading, and is only 14. I am less of a mathematician and more visual.
Of the images above, I like the linework on the tattered mummy most. The cross-htching lends a nice texture.
So are your algos something that you are working on to release as a product?
:)
Brian
Sounds like you and your nephew compliment each other perfectly and make a great team!
Thanks for sharing a bit about your process. For me, it all always starts with developing a workflow to produce something manually, and then pinning it down to specifics in order to try to automate it for the general case. Your list sounds like the pinning it down part of that process, which is really interesting to me.
Have you ever considered doing a time lapse video of your work? I'd bet I'm not the only one who would enjoy watching you create something like your coffee shop drawings. For those that haven't done so already, check out the thread - tons of 2D/3D goodness:
https://www.daz3d.com/forums/discussion/comment/4688891/#Comment_4688891
My algos definitely seem to work better with lots of detail to chew on - whether comes geometry or maps. The flip side is they've always had issues with close-ups of normal/smooth human faces. It's all part of the process of developing a style to draw different subject matter at different scales, which I'm sure you know all about.
As far as releasing anything as a product goes, in short, unfortunately no. It's a 2 step process:
1. The scene is analyzed in DS, a series of new shaders is dynamically applied and parameters set, and the analysis passes are rendered out
2. The output is sent to algorithms outside of DS and the info is used to govern the decisions made (strokes that are made by fractal algorithms, etc.)
The second part includes complex neural networks that run on a distributed network (many boxes), as well as other integrated software. This is the part that can't really be packaged up and redistributed.
- Greg
That is a really good idea. Sorry to be so long in replying. Got caught up in some other busy work and a few trips to a dentist and whatnot... :)
Hope to get back in the swing and do something like that.
:)
Brian
No need to apologize - we've all got a lot going on, I'm sure. Looking forward to checking it out if/when you do.
As far as the NPR animation goes, I learned a ton from the short animation tests. But, as is the case so often, the results raised as many questions as they answered. The experience got me thinking about some additional ways to utilize neural nets in the process . . .
While trying to wrap my mind around all of that, I've been spending my free 3D time getting my hair on. I know this is OT for this thread, but hopefully the OP won't mind lol. Anyway, I've been busy creating new tools & procedures to make strand based hair (focused on animal fur). Here's a reference I made for my own use that shows the base styles which are possible so far:
1, 2, 3, 4, and 5 are the base styles. 6 shows what 4 looks like with 1 underneath (1 can be placed under any of the styles making them all look fuller). 7 and 8 show the two curly options when applied to 4 (the curls can be applied to 2, 3, 4 or 5). These will be used to create the complete styles for the critters.
I've spent quite a bit of time with the shader, too. Wanted to make soft, luxurious looking fur without taxing system resources too much - believe I've found somewhat of a sweet spot. It's been a lot of fun to work with, and it hasn't crashed on me a single time! In the past, the instability of some other hair tools in DS has been an issue that was a huge turn-off.
In the meantime, just to make sure my NPR chops don't get too rusty, here's a fun render playing with some posing & panel elements:
https://www.daz3d.com/kayma-for-genesis-8-female
https://www.daz3d.com/maxx-hd-mega-pack-for-the-brute-8
- Greg
Love Herschel's creatures, and just got around to rendering this one:
https://www.daz3d.com/hh-ravager-drone
Not sure I like what the algos did colorwise, although it's accurate to the original.
- Greg