1stbees Trials and Piles promptly

some misc. ai gen postings.

Gotta love the outfit details and the realistic hair.

 

«13

Comments

  • FirstBastionFirstBastion Posts: 7,760

    Is there an "uncanny valley"  to AI generated art? 

  • linvanchenelinvanchene Posts: 1,382
    edited April 27

    FirstBastion said:

    Is there an "uncanny valley"  to AI generated art? 

     

    My impression is that the "uncanny valley" can be found within some image results you get with the SDXL 1.0 base model.

    If you use a "realistic" checkpoint you pivot towards a heavily retouched photoshop style.

    If you use an anime or 3D checkpoint you lean towards the safer stylized versions.

     

    XY Plot example:

    prompt:

    photo of a woman, long brown hair, brown eyes, jewelry, earrings, armor, shoulder pads,

    rim lighting, depth of field,

     

     

    Daz AI Studio without LoRA:

     

    The influence of training

    We will have to wait and see how the style of  Daz AI Studio develops over time with more training,  different  LoRA & base models.

     

    Examples:

    A checkpoint trained on only Genesis 9 shapes yields different results than a checkpoint trained on toon figures like Aiko.

    A checkpoint trained only on images rendered in Iray yields different results than a checkpoint trained with 3Delight in the mix.

    A checkpoint trained with images with themed environments yields different results than a checkpoint trained with promo images featuring monochrome backgrounds.

     

    Maybe Daz will release different types of checkpoints & LoRA in the future? Some more realistic, some more stylized? 

     

    The "uncanny valley" may be a matter of personal taste?

    Where do you draw the line?

    XY-Plot_1120805326016108_example.jpg
    5070 x 6359 - 3M
    base_dazai_57d04404-04bd-11ef-b9e1-e9c8be489245.png
    896 x 1152 - 2M
    Post edited by linvanchene on
  • FirstBastionFirstBastion Posts: 7,760
    edited April 28

    That is always the question, what falls into the range of "acceptable" which is probably different for each person who uses these tools.

    1st-concern.jpg
    750 x 750 - 116K
    Post edited by FirstBastion on
  • FirstBastionFirstBastion Posts: 7,760
    edited May 1

    forest hike

    1st-ranger.jpg
    750 x 750 - 89K
    Post edited by FirstBastion on
  • FirstBastionFirstBastion Posts: 7,760
    edited May 2

    Medieval

    1st-medi.jpg
    750 x 750 - 90K
    Post edited by FirstBastion on
  • FirstBastionFirstBastion Posts: 7,760
    edited May 3

    game world

    1st-gamewor.jpg
    750 x 750 - 97K
    Post edited by FirstBastion on
  • FirstBastionFirstBastion Posts: 7,760
    edited May 4

    Figured I better try out the flagship Vic9.

    1st-v9realistic.jpg
    750 x 750 - 71K
    Post edited by FirstBastion on
  • FirstBastionFirstBastion Posts: 7,760

    gracefully...

  • FSMCDesignsFSMCDesigns Posts: 12,752

    Curious, why do you cover the images you post with a large AI GEN stamp?

  • FirstBastionFirstBastion Posts: 7,760

    I am strongly in the "NO AI" camp for the ethical issue that were brought up re: training and infringement,  but recognized that nothing is stopping the tidal wave of this new tech,  so have been using and developing the skills since the end of 2022. 

    I just feel AI art made by a computer should be clearly labelled as such.  Even if I prompted the thing into existence.

  • FSMCDesignsFSMCDesigns Posts: 12,752

    That is kind of what I thought since covering it with the stamp eliminates any viewing pleasure, which seems to be the point now.Thanks for the reply.

  • linvanchenelinvanchene Posts: 1,382
    edited May 6

     

    I just feel AI art made by a computer should be clearly labelled as such. 

    Coming from a media background declaring the source has always been deeply ingrained.

    On the other hand, in history it has sometimes led to discrimination when it was enforced to wear clearly visible labels.

    Media outlets have embraced the practice to declare sources below an image or in movies at the end of a film.

     

    Because this is a forum dedicated to Daz AI Studio I just assume that any images posted here were generated by AI.

     

    prompt:

    A blue teddybear holding a sign with the text "bear" written on it, aviator sunglasses, clown outfit, in a circus, spot light, depth of field,

     

    If some post images who are not created with Daz AI Studio it could help to list the the base model, checkpoints, prompts or other settings used for generating their images to provide some context.

    But no one should feel obligated to make that extra effort.

    Thank you for sharing your images, I enjoy returning to the forum to check the latest creations.

    dazai_a1eb759c-0beb-11ef-ba01-e9c8be489245.png
    1152 x 896 - 1M
    Post edited by linvanchene on
  • FirstBastionFirstBastion Posts: 7,760
    edited May 6

    Obviously our 3D renders are also "made by a computer"  and photohop filters and actions are also "made by a computer".   It is all a little bit much to be singling out AI generation.  All those other tasks though, required some technical skills or creative talent to achieve the results.  The artist had to work at it.

    With AI generation,  it literally is a "make art" button.   So any hack can now output reams of content in no time at all.  That's the thing that bothers the artist most.  All those years of work and dedication  developing a craft and talent is now moot.

    So self labelling these AI Generated images  is my small attempt to acknowledge the artists that are affected by these tumultuous changes.  No biggie. We all get to the future eventually.

    Teddy bear holding a bear.

     

    1st-bearholding.jpg
    750 x 750 - 88K
    Post edited by FirstBastion on
  • linvanchenelinvanchene Posts: 1,382
    edited May 7

    All those years of work and dedication  developing a craft and talent is now moot.

    I wrote in another post that it took me 6 months to come to terms with the situation that AI is here to stay.

    But  my internal struggle and worry what this means for 3D may be an ongoing process.

    I still hope that Daz can provide smart solutions to combine AI and 3D so all that invested time and money can still be put to use in the future.

    So self labelling these AI Generated images  is my small attempt to acknowledge the artists that are affected by these tumultuous changes. 

    I can understand this sentiment. I assumed there must have been quite some discussions between PA and Daz behind the scenes. 

    a blue teddybear and a pink teddybear sitting on a new york rooftop, view from behind, jacket, enjoying the sunrise, hoping for the best, golden hour, rim lighting, depth of field,

    dazai_eec82264-0c01-11ef-ba01-e9c8be489245.png
    1152 x 896 - 1M
    Post edited by linvanchene on
  • FirstBastionFirstBastion Posts: 7,760

    princess

  • FirstBastionFirstBastion Posts: 7,760

    Knitted

  • FirstBastionFirstBastion Posts: 7,760

    special day

     

  • FirstBastionFirstBastion Posts: 7,760

    Out hiking

  • FirstBastionFirstBastion Posts: 7,760

    Armored

  • FirstBastionFirstBastion Posts: 7,760

    car

  • FirstBastionFirstBastion Posts: 7,760

    hero

  • FirstBastionFirstBastion Posts: 7,760

    evah aftah

  • FirstBastionFirstBastion Posts: 7,760

    Falling

  • FirstBastionFirstBastion Posts: 7,760

    Nature Calls

  • FirstBastionFirstBastion Posts: 7,760
    edited May 17

    try to fly

    Post edited by FirstBastion on
  • FirstBastionFirstBastion Posts: 7,760

    Defend

  • FirstBastionFirstBastion Posts: 7,760
    edited May 23

    Finding the Words

    Post edited by FirstBastion on
  • FirstBastionFirstBastion Posts: 7,760

    Beret

  • FirstBastionFirstBastion Posts: 7,760

    Curious

  • FirstBastionFirstBastion Posts: 7,760

    Can't Back Down

Sign In or Register to comment.