Remixing your art with AI

WendyLuvsCatz

jmucchiello said:

At the risk of being contentious, all I see here is some folks need to learn how to do lighting better.

mine are mostly composited using png layers, not that that is any excuse  

jmucchiello

WendyLuvsCatz said:

jmucchiello said:

At the risk of being contentious, all I see here is some folks need to learn how to do lighting better.

mine are mostly composited using png layers, not that that is any excuse  

I'm not claiming I'm a lighting master. But when I see my lighting is off, I reach for some spotlights, or some colored gels, or I create a ghost light, or I tweak the tone mapping, or... I don't think AI. Because I don't learn about lighting if I fix it with a filter (whether AI or just mathematical). 

 

jmucchiello

Barefoot Upto My Soul said:

i am loving this forum commentaries with the before and after DAZ to ai enhanced that i had to add to my watch list. Here is one of mine. The top rendered in 3DLight about 20 minutes and the bottom having it enhanced with ai in 30 seconds.

But the eye color changed from brown to blue. Doesn't that bother you?

jmucchiello

This was thirty seconds in GIMP playing with levels and exposure. I don't claim any ownership of the image.

Barefoot Upto My Soul

jmucchiello said:

Barefoot Upto My Soul said:

i am loving this forum commentaries with the before and after DAZ to ai enhanced that i had to add to my watch list. Here is one of mine. The top rendered in 3DLight about 20 minutes and the bottom having it enhanced with ai in 30 seconds.

But the eye color changed from brown to blue. Doesn't that bother you?

Brownish yellow to greenish for a redhead. i can tell the ai what colour the eyes should be along with hair colour, etc.

Yes it would bother me if it changed without my permission from one shot to the next. That is where cloning in post in photoshop would come in. Or having the ai either rerender or having the ai fix the error of the new image it rendered.

If it doesn't get it right in the first 3 to 5 renders, it won't get it right. In which case i have to move on to something different or shut down for the cache to clear.

Barefoot Upto My Soul

high quality, clear, realistic, cinematic, cinematic color grading. Keep the camera static and no zoom, no dolly movement either. Have the woman wear this Daz outfit and accessories. Ensure this looks natural and realistic, while preserving the woman’s original facial features, blue eyes, body features, proportions, original hair features and blond hair color.

Barefoot Upto My Soul

So this would be putting a photo of my friend into a Daz outfit. Notice the fingers get screwed up if i don't uncross them before render. Would have to fix in post. Any clothes would do really, but i love these designs i bought at Daz for V2 V3 V4 G1 G2 etc.

Barefoot Upto My Soul

jmucchiello said:

This was thirty seconds in GIMP playing with levels and exposure. I don't claim any ownership of the image.

3DLight didn't get it rendered the way i wanted or maybe i should have had her wear two of that hair as i notice some people do with iray. Still, i was quite pleased with the result at the time without fixing it.

Your edit in GIMP is nice of the HDRI however she still looks like she is going bald like a man, and the nostrils look too bright. i just looked in the mirror at me, and my nostrils are dark like in the ai render of her. You are welcome to try again if you like. :)

jmucchiello

Barefoot Upto My Soul said:

jmucchiello said:

This was thirty seconds in GIMP playing with levels and exposure. I don't claim any ownership of the image.

3DLight didn't get it rendered the way i wanted or maybe i should have had her wear two of that hair as i notice some people do with iray. Still, i was quite pleased with the result at the time without fixing it.

Your edit in GIMP is nice of the HDRI however she still looks like she is going bald like a man, and the nostrils look too bright. i just looked in the mirror at me, and my nostrils are dark like in the ai render of her. You are welcome to try again if you like. :)

I can't fix issues with the render. I was only addressing the lighting issues. It isn't my fault the figure has no expression in the render. And it's a 3DL render. 3DL is decades old technology. Using modern assets would help with the hair. The AI adding expression is a flaw IMO.

Barefoot Upto My Soul

csaa said:

Barefoot Upto My Soul,

I haven't put much effort into combining generative AI and Daz3D. Mostly I use AI in my professional work; with any spare time I have for creative hobbies, I'd like to explore pre-AI art. 

That said, I do appreciate this thread! There's so much contention surrounding AI, much of it based on misinformation or narrow POVs. The work that people share here goes a long way to support much needed counter-narratives.

Just a suggestion: Please continue sharing your before-vs-after renders. Nothing carries more weight, or gets the idea across more succinctly, than a well made image. More so than a wall of text here in the Daz Forum. 

Cheers!

Wow, yeah for generative ai i was trying out backgrounds in some ai and character actors in other ai and combining in another ai adding music and voice in more ai to get this short clip. i can imagine the wonderful work you do.

https://youtube.com/shorts/jTBLaIq4W4Y?si=41Srsfc8_hwTXB7w

Barefoot Upto My Soul

But if i am mixing Daz assets and enhancing and animating with ai, this is what i would get. https://youtube.com/shorts/Xd5AANKuTaA?si=U87MpHSBYQsJufWs

Barefoot Upto My Soul

wsterdan said:

Nano Banana Pro (and Gemini) handles things this way as well; if I say, "redraw the image showing character xxx..." there's a chance background might be altered, but if I say, "show character xxx..." the background is usually intact.

Because I like to do a lot of "bridge on a starship" type images (inspired by Trek) I've worked out a method to generate a single high-res image of the bridge and the main characters. Once that's done, I'll use that image as my starting point for bridge interations and advice NB Pro to "show the captain turning his head and talking to science" and I'll get an image of just those changes.

The second part of my system – once the base image is set – is to name the characters and use those names going forward by copy/pasting the naming section.

For example, here's an old Smay/Chibi verson of some Justice League members:

I asked NB Pro to "redraw the characters in a modern, 2d animated film style;" and got this:

 

Now, I name the characters and use the names to change *some* of their poses leaving the others intact:

"For reference, the names of the characters can be used for instructions: the male character in green on the far left is "Hal"; the male character second from the left in red is "Barry"; the female in the centre is "Diana"; the fourth charcter from the left wearing grey an blue is "Bruce" and the character on the far frght is "Clark".

With the five characters, show "Dianna", "Bruce" and "Clark" laughing at a joke; no word balloons"

And get this:

 

Then, have the other two changing while the first three remain intact:

 

"For reference, the names of the characters can be used for instructions: the male character in green on the far left is "Hal"; the male character second from the left in red is "Barry"; the female in the centre is "Diana"; the fourth character from the left wearing grey and blue is "Bruce" and the character on the far right is "Clark".

with the five characters, show "Hal", and "Barry" talking to each other;  no word balloons"

and get this:

Finally, to show that the backgrounds remain intact I make a very youth-oriented version of the guys sitting around a table in a futuristic meeting hall:

 

and prompt: "

For reference, the names of the characters can be used for instructions: the male character in green on the far left is "Hal"; the male character second from the left in red is "Barry"; the female in the centre is "Diana"; the fourth charcter from the left wearing grey an blue is "Bruce" and the character on the far right is "Clark".

with the five characters, show "Dianna", "Bruce" and "Clark" laughing at a joke; no word balloons"

and get this:

You might notice that when some of the characters are asked to perform, the others might show slight reaction to them like eyes turning or slight smiles.

Of note: I started to try this using Gemini but was told, "There are a lot of people I can help with, but I can't depict some public figures. Do you have anyone else in mind?", no doubt in response to trademarked characters like Clark, Bruce and Diana.

Those are serious and well done instruction prompts to follow.

wsterdan

Barefoot Upto My Soul said:

wsterdan said:

Nano Banana Pro (and Gemini) handles things this way as well; if I say, "redraw the image showing character xxx..." there's a chance background might be altered, but if I say, "show character xxx..." the background is usually intact.

Because I like to do a lot of "bridge on a starship" type images (inspired by Trek) I've worked out a method to generate a single high-res image of the bridge and the main characters. Once that's done, I'll use that image as my starting point for bridge interations and advice NB Pro to "show the captain turning his head and talking to science" and I'll get an image of just those changes.

The second part of my system – once the base image is set – is to name the characters and use those names going forward by copy/pasting the naming section.

For example, here's an old Smay/Chibi verson of some Justice League members:

I asked NB Pro to "redraw the characters in a modern, 2d animated film style;" and got this:

 

Now, I name the characters and use the names to change *some* of their poses leaving the others intact:

"For reference, the names of the characters can be used for instructions: the male character in green on the far left is "Hal"; the male character second from the left in red is "Barry"; the female in the centre is "Diana"; the fourth charcter from the left wearing grey an blue is "Bruce" and the character on the far frght is "Clark".

With the five characters, show "Dianna", "Bruce" and "Clark" laughing at a joke; no word balloons"

And get this:

 

Then, have the other two changing while the first three remain intact:

 

"For reference, the names of the characters can be used for instructions: the male character in green on the far left is "Hal"; the male character second from the left in red is "Barry"; the female in the centre is "Diana"; the fourth character from the left wearing grey and blue is "Bruce" and the character on the far right is "Clark".

with the five characters, show "Hal", and "Barry" talking to each other;  no word balloons"

and get this:

Finally, to show that the backgrounds remain intact I make a very youth-oriented version of the guys sitting around a table in a futuristic meeting hall:

 

and prompt: "

For reference, the names of the characters can be used for instructions: the male character in green on the far left is "Hal"; the male character second from the left in red is "Barry"; the female in the centre is "Diana"; the fourth charcter from the left wearing grey an blue is "Bruce" and the character on the far right is "Clark".

with the five characters, show "Dianna", "Bruce" and "Clark" laughing at a joke; no word balloons"

and get this:

You might notice that when some of the characters are asked to perform, the others might show slight reaction to them like eyes turning or slight smiles.

Of note: I started to try this using Gemini but was told, "There are a lot of people I can help with, but I can't depict some public figures. Do you have anyone else in mind?", no doubt in response to trademarked characters like Clark, Bruce and Diana.

Those are serious and well done instruction prompts to follow.

Thanks, every day is still a learning process just trying to keep up with new features, but a lot of fun and I think well worth it. 

WendyLuvsCatz

another hybrid video

am going to upload fully AI animated renders to my AI channel next and fully 3D CGI ones to my main

will embed when done

see here

 

wsterdan

Fascinating stuff, as always Wendy!

Barefoot Upto My Soul

WendyLuvsCatz said:

wsterdan said:

Wow, lots of stuff happening here, love the soundtrack.

How long are the clips you generate? Most of the video generators I've been looking at only make 15- or 30-second clips.

4 seconds, I queue them usually 30 videos in one go

121 frames 30fps takes 25mins on my PC 1280x730 pixels and 2X upscaling  

its considerably faster than rendering the images I animate, in DAZ studio 

I simply do poses along the timeline in D|S with some camera movement

a different pose each second and render a series 30 of them with the default timeline 

the fps is irrelevant 

playing it back would look ridiculous but that's not the intention as I am actually letting the AI decide how to get from one pose to the next over 121 frames

the AI also adds camera movement whether I want it or not, if I don't have any it's actually worse

adding subtle movement between my frames reigns it in a bit

Oh, i thought you were doing first frame last frame renders in daz to import into ai. Letting the ai do the movie render. But just that thought of 1st and last from Daz blew my mind. It's what i do when i make ai movie clips... it just didn't occure to me to do the same here with Daz. Seriously blew my mind :)

Barefoot Upto My Soul

Top in daz of V3 Emilie and the bottom where i put her in the subway metro with backlighting. Rendered in Gemini Nano Banana ai free. Added headphones too.

WendyLuvsCatz

Barefoot Upto My Soul I do both

added my other videos separating AI and 3D on other page

this thread moves faster than I can upload 

thanks wsterdan

Barefoot Upto My Soul

Was having fun putting my friend in a daz scene via ai. i hope i am not spamming this board thread.

csaa

Barefoot Upto My Soul said:

... via ai. i hope i am not spamming this board thread.

For all its worth, keep this in mind: good signal to noise ratio. The life (or death) of a forum discussion hinges on that. May the thought spread throughout the four corners of the world, and echo down the ages. 

Cheers!

WendyLuvsCatz

ooops, I didn't mean the comment about the thread moving fast in that way at all

WendyLuvsCatz

3DIO said:

WendyLuvsCatz said:

am going to upload fully AI animated renders to my AI channel next and fully 3D CGI ones to my main

I watch a lot of your videso since I'm always curious how the AI interprets the renders and animations you feed it.  I'm surprised, though, that you never make split-screen viideos where the original animation (or sequence of images) is playing on the left of the screen, while the AI interpretation of it is playing (in sync) on the right.

 

not particularly interested in doing tutorials

I just do this for fun 

there are actual talented people who can do that stuff and explain it all, and do

Barefoot Upto My Soul

Okay animated the three images as first last and extend video to first last. Added some sound. Left out the cardboard wide image so aspect ratio remains square.

Here it is on my DeviantArt gallery. https://www.deviantart.com/barefootuptomysoul/art/EXT-CYBERPUNK-CITY-DAY-1360530219

WendyLuvsCatz

smooth 

Barefoot Upto My Soul

WendyLuvsCatz said:

smooth 

Thank you :)

i hope you don't mind that i added you to my watch list. Seeing you and the others here on DeviantArt makes it feel less lonely there. Lots of cool renders everyone is posting there.

WendyLuvsCatz

Barefoot Upto My Soul said:

WendyLuvsCatz said:

smooth 

Thank you :)

i hope you don't mind that i added you to my watch list. Seeing you and the others here on DeviantArt makes it feel less lonely there. Lots of cool renders everyone is posting there.

 

I don't mind but I never go there, not uploaded anything in years

YouTube is my platform  

Barefoot Upto My Soul

3DIO said:

Barefoot Upto My Soul said:

Okay animated the three images as first last and extend video to first last. Added some sound. Left out the cardboard wide image so aspect ratio remains square.

Here it is on my DeviantArt gallery. https://www.deviantart.com/barefootuptomysoul/art/EXT-CYBERPUNK-CITY-DAY-1360530219
 

Can't remember the name of it now. but there's an AI product for ComyUI that allows you to control the ease-in, ease-out and paths for animations like that.  It had a limit of five seconds per take due to memory requirements, but it worked extremely well.

Any of these perhaps? It sounds like you're thinking of DragNUWA or the CameraCtrl / Wan-Move (Wan ATI) nodes for ComfyUI.

​Here are the main custom node suites that fit that exact description:

​DragNUWA / MotionCtrl: Uses drawn path trajectories with point-level motion control. It is famous for strict 5-second (around 16–24 fps) generation limits per take due to video diffusion memory constraints. ​ComfyUI-Keyframed (or FizzNodes): Provides explicit math-based curve interpolation like ease_in, ease_out, sine_wave, and exp_decay across motion vectors and prompts. ​FL_PathAnimator / Wan ATI Trajectory Control: A visual path editor node in ComfyUI that lets you draw keyframe paths and bezier curves directly on the frame canvas for object movement.

​DragNUWA is usually the top one people misplace by name when talking about drawn motion paths and that tight 5-second take limit!

Barefoot Upto My Soul

Sometimes it's a real bitch slap getting the ai to render an image the way i want it. So i have to say close enough and just go to sleep for the night. These were putting my friend and me into a daz scene.

WendyLuvsCatz

this was too long for my Tycath channel without YouTube verification 

so having to break up the 3 tunes into a playlist instead 

My Madcatlady subscribers hate AI so will get heaps of dislikes and some will unsubscribe 

csaa

WendyLuvsCatz said:

My Madcatlady subscribers hate AI so will get heaps of dislikes and some will unsubscribe 

WendyLuvsCatz,

There'll always be nay-sayers. 

It is not the critic who counts; not the man who points out how the strong man stumbles, or where the doer of deeds could have done them better. The credit belongs to the man who is actually in the arena, whose face is marred by dust and sweat and blood; who strives valiantly; who errs, who comes short again and again, because there is no effort without error and shortcoming; but who does actually strive to do the deeds; who knows great enthusiasms, the great devotions; who spends himself in a worthy cause; who at the best knows in the end the triumph of high achievement, and who at the worst, if he fails, at least fails while daring greatly, so that his place shall never be with those cold and timid souls who neither know victory nor defeat.

 Cheers!