AI Part1 – Generative Images – (240813 CMA042)


Caught My Attention
240813 = 2024Aug13



I’ve been wanting to write a post on Artificial Intelligence, but every time I think about what I’d talk about, the realization that there are so many aspects to AI, that it seems logical that this becomes a series of posts over time. I started down the AI rabbit-hole with image generation, so it makes sense to start there.

In the summer of 2023 I went to what is probably longest running computer graphics tradeshows, SIGGRAPH (Special Interest Group on Computer Graphics and Interactive Techniques). I’ve gone to or worked at this event on and off for the last 30 years. At the 2023 show I got to see how AI was being used in a way that I could understand for real commercial purposes – till then I was just seeing things as being neat or cute, not really “getting it”.

The demo I watched was at the ShutterStock booth and a guy was demostrating how he had created imagery of “models” that were then photoshopped into ad layouts for a cosmetics brand with other actual photographed elements of the products. The big takeaway for me was how he said it could take hours of itterations in the description requests to the ai engine to finally create the image that the client was looking for. Each tiny attribute could have many subattributes, the variables could seem endless.

There are several image generators out there and I tried a few out. To create anything in AI, the computing resources are huge and can take a lot of energy and compute time. ie it’s not cheap, so free online generators will limit your usage (and at the moment I don’t see roi for myself to spend money to get access).

Two of the generators that I had some success with are Stable Diffusion and Microsoft Bing. Here are examples of the commands that I was able to design to generate the images seen below.

Stable Diffusion:

  • photo of a cortado in a glass on a saucer, in a coffee shop that has windows looking out to the ocean with palm trees in the foreground
  • older indonesian man with goatee and retangular frame glasses, in a coffee shop, holding cappuccino, wearing flannel shirt, baseball hat and bomber jacket, natural light, magazine photo, 5 0 mm, zoomed out 2x
  • young adult japanese woman reading in a window seat in a coffee shop, 4k, lifestyle, light, mood, cinematic lighting rendered, zoomed out 5x
  • John’s Japanese Hot Sauce on a shop counter that has windows looking out to the ocean with palm trees in the foreground, kawai aesthetic, ambient occlusion, ultra realistic, warm overhead lighting, magazine photo, 5 0 mm, zoomed out 3x

Bing:

  • create a ultra realistic photograph of a cortado in a glass on a saucer on a counter, in a coffee shop that has windows looking out to the ocean with palm trees in the foreground
  • an ultra realistic photograph of an older japanese man with goatee and retangular wire frame glasses in a coffee shop wearing flannel shirt, natural light, magazine photo, zoomed out 3x
  • ultra realistic photograph of young adult japanese woman reading in a window seat in a coffee shop, 4k, lifestyle, light, mood, cinematic lighting rendered, zoomed out 5x
  • John’s Japanese Hot Sauce on a shop counter that has windows looking out to the ocean with palm trees in the foreground, kawai aesthetic, ambient occlusion, ultra realistic, warm overhead lighting, magazine photo, 5 0 mm, zoomed out 3x

Stay tuned for more on AI. I think maybe next time I’ll show some of the failed attempts 🙂

Leave a comment

Your email address will not be published. Required fields are marked *