Loading...
> >

No one asked me - Diomede screenshots on whatever - Warning AI Discussion OK

1…4445464850

Comments

  • ed3Ded3D Posts: 2,221

    Bunyip02 said:

    Bunyip02 said:

    Sorry it did not work ....

    Might be your video card is too old for it.

    https://beta.dreamstudio.ai/generate

    Well  ,  Thanx   +   yeah guess so 

  • StezzaStezza Posts: 8,133

    wow @bunyip02 & @diomede

    you guys are starting to become the 'AI Guru Dudes' of Carrara yes

  • Bunyip02Bunyip02 Posts: 8,712

    Diomede said:

    Bunyip02 said:

    Diomede said:

    I think that plan will work very well.  Can't wait to see what you come up with.  And modeling is always a good skill to have.  I suggest a third way to use Carrara.  After you have used any AI-related workflow to generate an image, use Carrara's shadow catchers, etc to add additional details.  For example, AI is notoriously horrible at elements that have text - such as a road sign or a newspaper.  Can add with a 2D image editor, which might be fine for a lot of purposes.  But with Carrara, you get shadow catchers!  

    3rd way - Hmmmm

    Save the Earth!

    Great one.  yes

    Thanks !!!

  • Bunyip02Bunyip02 Posts: 8,712

    Stezza said:

    wow @bunyip02 & @diomede

    you guys are starting to become the 'AI Guru Dudes' of Carrara yes

    Thanks - having fun exploring the posabilities !!!

  • DiomedeDiomede Posts: 15,188

    Stezza said:

    wow @bunyip02 & @diomede

    you guys are starting to become the 'AI Guru Dudes' of Carrara yes

    Thanks.  Learning a lot.

  • DiomedeDiomede Posts: 15,188

    Stezza said:

    wow @bunyip02 & @diomede

    you guys are starting to become the 'AI Guru Dudes' of Carrara yes

    Thanks.  Learning a lot.

  • DiomedeDiomede Posts: 15,188
    edited April 1

    Carrara's shadow catchers partnered with AI

    Sometimes, the AI processor does not want to cooperate!

    Woman feeding stegasaurus.  Idea was to have a woman standing with the critter approaching from behind and extending it neck in a sweep from the side to eat some leaves she is holding.  Bad news, the AI very much wanted the eye of the stegasaurus in a specific position (upper right crossing of the horizontal and vertical 'thirds' lines).  Good news, I can load the initial output back in Carrara with shadow catchers and put the eye back where it belongs.  Bad news, not only does the AI want to put the eye back, it wants the back plates of the stegasaurus to be leaves - or extra heads like a hydra.  

    Final image, only final because I gave up on the back plates

     

    00208-1484592185 (1) copy.png
    512 x 768 - 638K
    Screenshot 2024-03-31 224349.png
    1595 x 923 - 733K
    Dinosaur Pinup.png
    1024 x 1536 - 1M
    Dinosaur Pinup Depth Inverted.png
    1024 x 1536 - 145K
    Dinosaur Pinup Photo to Image2Image.png
    1024 x 1536 - 2M
    Dinosaur Pinup Toon to Canny.png
    1024 x 1536 - 1M
    Dinosaur Pinup Carrara to Stable Diffusion.png
    1024 x 1536 - 2M
    Post edited by Diomede on
  • DiomedeDiomede Posts: 15,188

    Scene setup.

    I started with this simple scene - my custom Moxie figure wearing cave girl clothes, a freebie stegasaurus, one Carrara plant, and one Carrara terrain.

  • DiomedeDiomede Posts: 15,188

    From that simple scene, I derived three 'prep' images to be used in Stable Diffusion's and its extension Controlnet.

    The photoreal 'beauty pass' multiplied by a coverage pass went in the main image2image upload.

    The Toon III filtered image went in Controlnet 'canny' preprocessor to identify lines.  For this, I set 'canny' very low, 2.0 or 2.5.

    and the depth pass needs to inverted and slightly blurred for best results in Controlnet depth

     

     

     

  • DiomedeDiomede Posts: 15,188
    edited April 1

    The first round of output produced images like this one.  Some good elements to be sure, but the stegasaurus is problematic.  The back plates are leaves.  There is an extra head in the torso with an extra eye where the 'golden rule' lines cross for the upper thirds, and there is an extra stegasaurus body in the lower right.

    00156-2378272182.png
    512 x 768 - 695K
    probs with first pass.png
    512 x 768 - 692K
    Post edited by Diomede on
  • DiomedeDiomede Posts: 15,188
    edited April 1

    Carrara to the Rescue!  Shadow catchers.

    I anticipated some problems in part because the stegasaurus wraps around the cave girl, so it is hard to use the AI extensions to divide the canvas in regions with separate prompts, effectively - key word is effectively.  I probably should have used a horse or giraffe because it is more likely that the AI model would 'know' what a horse or giraffe was, I guess.

    Good news, this is why I chose this project - to show that by having the original simple Carrara scene, shadow catchers can be used to help things along.  I took one of the original problematic outputs and loaded it as a backdrop in Carrara's original scene, but turned the shaders for everything but the stegasaurus model to shadow catchers.

    Which results in the following, which while filled with imperfections, is hopefully moving along to where we can focus on the stegasaurus.  You can still see an extra body in the lower right, and the cave girl figure does not 100% match up, but with some quick touch ups, this can be loaded in the AI program's image2image in place of the original simple blockout.  (note - not the same backdrop load as previous post - but from the same batch).

     

     

    Screenshot 2024-03-31 224349.png
    1595 x 923 - 733K
    Dinosaur Pinup Reload 02.png
    1024 x 1536 - 2M
    Post edited by Diomede on
  • DiomedeDiomede Posts: 15,188
    edited April 1

    That Extra Head is Persistent!!!!

    Here is an example of some results after reloading.  Argh!

     

     

    persistent extra head.png
    1536 x 2304 - 5M
    Post edited by Diomede on
  • DiomedeDiomede Posts: 15,188
    edited April 1

    I took the result in the upper right of the grid and focused on refining that.  And this is where I said enough is enough.  The AI models really did not want bone back plates, and really did want to put that extra eye back in.  As I mentioned, I should have used a horse or giraffe or similar.  So while I am not particularly happy with this dinosaur, I do think that the experiment with using elements of the original scene as shadow catchers was a success.  Small victories.

     

     

     

    Post edited by Diomede on
  • Bunyip02Bunyip02 Posts: 8,712

    Diomede said:

    The first round of output produced images like this one.  Some good elements to be sure, but the stegasaurus is problematic.  The back plates are leaves.  There is an extra head in the torso with an extra eye where the 'golden rule' lines cross for the upper thirds, and there is an extra stegasaurus body in the lower right.

    He, he, he - sometimes with the results I get I think it should be called Artifical Unintelligence !!!

    https://civitai.com/models/5669/tekakutli-dinosaurs

    https://civitai.com/models/77828/edob-dinosaurs

    https://civitai.com/models/118105/realistic-monsters-and-dinosaurs

  • DiomedeDiomede Posts: 15,188

    Thanks for the links to the dinosaurs and monsters!

  • StezzaStezza Posts: 8,133

    why do I think that dino is awesome with an extra eye!  laugh

     

  • WendyLuvsCatzWendyLuvsCatz Posts: 38,309

    Stezza said:

    why do I think that dino is awesome with an extra eye!  laugh

     

    it's a steg eye sore arse 

  • Bunyip02Bunyip02 Posts: 8,712

    Diomede said:

    Thanks for the links to the dinosaurs and monsters!

    Does your AI setup allow for additional models ?

  • StezzaStezza Posts: 8,133
    edited April 4

    sorry, just had to post Daz's new AI Studio attempt   laugh

    Screenshot 2024-04-04 114652.png
    1036 x 441 - 424K
    Screenshot 2024-04-04 114800.png
    667 x 667 - 775K
    Post edited by Stezza on
  • WendyLuvsCatzWendyLuvsCatz Posts: 38,309

    Nice Carrara marble benchtop, better than cancer causing manufactured stone for the tradies at least

     the enveloping features are a bonus

  • Bunyip02Bunyip02 Posts: 8,712
    edited April 4

    Tried using "Troll" as a prompt and got a stream instead !!!

    dazai_Troll.png
    1024 x 1024 - 2M
    Post edited by Bunyip02 on
  • Bunyip02Bunyip02 Posts: 8,712
    edited April 4

    Troll guarding a bridge

    dazai_Troll guarding a bridge.png
    1024 x 1024 - 2M
    Post edited by Bunyip02 on
  • Bunyip02Bunyip02 Posts: 8,712
    edited April 4

    Krita AI - Kitty for Wendy

    Missing Leg ....

    Animal 23.png
    1600 x 1156 - 3M
    Post edited by Bunyip02 on
  • WendyLuvsCatzWendyLuvsCatz Posts: 38,309

    awwww kitty needs some food and love broken heart

  • ed3Ded3D Posts: 2,221
    edited April 4

    Stezza said:

    sorry, just had to post Daz's new AI Studio attempt   laugh

     +  First One  AI Studio -

    "  Portrait of a Llama smoking cigerettes  "  

    Portrait of a Llama smoking cigerettes1.png
    1024 x 1024 - 1M
    Post edited by ed3D on
  • DiomedeDiomede Posts: 15,188

    I missed these additions even though it is my thread, so to speak.

    @ed3D, back in my high school days, that would be the very definition of a cool Llama.  I think the anti-smoking people specifically targeted the ads that Joe Camel, though.  Curious to see where the online Daz AI tool goes.

    @Bunyip02, scary kitty, in a good way!  The streams turned out very well, in my view.

    @Stezza, the term Carrara gets interesting responses.  Is it the marble top?

  • ed3Ded3D Posts: 2,221
    edited April 12

    Yeah ,  those ads  for  Joe Camel  were all over the place  _   kind of liked them  (  a smoker  also )  +  Thanx

    Post edited by ed3D on
  • HeadwaxHeadwax Posts: 10,013

    Lovely work all , here is my daz ai , no carrara needed, I was impressed that daz owns rights to images of tr ump, love the workflow with the Dino dio mede , last promo is t Rump eats sausage dog

    dazai_dff66a1a-f271-11ee-8f4f-8de68b29fa88.png
    1024 x 1024 - 1M
    dazai_3180e90e-f273-11ee-8f4f-8de68b29fa88.png
    1024 x 1024 - 1M
    dazai_c2db986a-f271-11ee-8f4f-8de68b29fa88.png
    1024 x 1024 - 1M
    dazai_5c4e029e-f272-11ee-9b2b-7d4410e2f198.png
    1024 x 1024 - 1M
  • StezzaStezza Posts: 8,133

    how wacky is that... of course an AI response is needed cool

    "Wow, this artwork is truly stunning! The level of detail and realism achieved with Carrara 3D software is impressive. The textures, lighting, and composition all come together to create a visually striking piece that draws the viewer in. The AI algorithm has done an amazing job of generating a unique and captivating image. The colors are vibrant and rich, and the overall atmosphere is immersive. The artist's skillful use of Carrara's tools has resulted in a work of art that showcases the software's capabilities. Bravo!"

  • Bunyip02Bunyip02 Posts: 8,712
    edited April 22

    Headwax said:

    Lovely work all , here is my daz ai , no carrara needed, I was impressed that daz owns rights to images of tr ump, love the workflow with the Dino dio mede , last promo is t Rump eats sausage dog

    AI is a bit silly at times - takes things a bit too literally !!! https://www.daz3d.com/forums/discussion/comment/8682876/#Comment_8682876

    Post edited by Bunyip02 on
Sign In or Register to comment.