LurkerNoMore: Fooocus AI Image Experiments

[obsidian]

Prompt:
A great white shark surfing like a professional surfer on a surfboard, riding a big ocean wave, wearing stylish sunglasses with clearly visible side arms resting above its eyes, realistic shark anatomy with a wide toothy grin, cinematic ocean spray, vibrant summer lighting, ultra-realistic, detailed textures, action shot, dynamic angle

Negative Prompt:
dog face, mammal mouth, cartoon, extra limbs, missing sunglasses arms, distorted sunglasses, melted glasses, blurry, low detail, deformed head, incorrect shark anatomy, dolphin

Model:
dynavisionXLAllInOneStyl ized_releaseV0610Bakedva e.safetensors

[LurkerNoMore]

That is what Midjourney is giving me basically.  But when trying to make it into a sticker style with border (so it can be cut out), it won't do it. 

When I use Gemini with the same prompt but with "sticker style" added at the start and it produces a more 2D flat image that is more easily used for stickers and tshirt transfers.

[B_B_C]

  But when trying to make it into a sticker style with border (so it can be cut out), it won't do

despite all your cutting posts here, are you not able to use a scissors?

[obsidian]

This is what ChatGPT gave me based on these revised prompts:

Prompt

A great white shark surfing like a professional surfer on a surfboard, riding a big ocean wave, wearing stylish sunglasses with clearly visible side arms resting above its eyes, shark body with realistic muscular human-like legs replacing both lower side fins, both legs bent in a natural surfing stance firmly planted on the surfboard, wide toothy grin, cinematic ocean spray, vibrant summer lighting, ultra-realistic, detailed textures, action shot, dynamic angle

Negative Prompt

dog face, mammal mouth, cartoon, missing sunglasses arms, distorted sunglasses, melted glasses, blurry, low detail, deformed head, incorrect shark anatomy (except for the human-like legs), dolphin, one leg only, asymmetrical limbs, leg off the surfboard

[obsidian]

That is what Midjourney is giving me basically.  But when trying to make it into a sticker style with border (so it can be cut out), it won't do it. 

When I use Gemini with the same prompt but with "sticker style" added at the start and it produces a more 2D flat image that is more easily used for stickers and tshirt transfers.

I am sure Fooocus can create images like that, with the right prompts and checkpoint models. Have you tried to see what ChatGPT or Grok can generate for you?

I upscaled your Gemini image in Gigapixel by 2X. Then I loaded it in Photoshop and selected the outer white region. Expanded the selection by 8 pixels. Deleted it all. Then I chose the smart selector and it selected the shark, waves, etc. I then turned that selection into a mask. I am attaching the result with the background as gray so you can see how it masked it. I saved a transparent PNG file but it is 2 MB in size and that will kill Ron's hard drive space if I try to upload - lol!

[obsidian]

This is what ChatGPT gave me based on these revised prompts:

Prompt

A great white shark surfing like a professional surfer on a surfboard, riding a big ocean wave, wearing stylish sunglasses with clearly visible side arms resting above its eyes, shark body with realistic muscular human-like legs replacing both lower side fins, both legs bent in a natural surfing stance firmly planted on the surfboard, wide toothy grin, cinematic ocean spray, vibrant summer lighting, ultra-realistic, detailed textures, action shot, dynamic angle

Negative Prompt

dog face, mammal mouth, cartoon, missing sunglasses arms, distorted sunglasses, melted glasses, blurry, low detail, deformed head, incorrect shark anatomy (except for the human-like legs), dolphin, one leg only, asymmetrical limbs, leg off the surfboard

Here's the same image with more "stylistic guidance" - per ChatGPT. It added more yellow to the waves and board.

[LurkerNoMore]

despite all your cutting posts here, are you not able to use a scissors?

It would take too much time to cut out 50+ stickers to be mailed everyday.

[LurkerNoMore]

I am sure Fooocus can create images like that, with the right prompts and checkpoint models. Have you tried to see what ChatGPT or Grok can generate for you?

I upscaled your Gemini image in Gigapixel by 2X. Then I loaded it in Photoshop and selected the outer white region. Expanded the selection by 8 pixels. Deleted it all. Then I chose the smart selector and it selected the shark, waves, etc. I then turned that selection into a mask. I am attaching the result with the background as gray so you can see how it masked it. I saved a transparent PNG file but it is 2 MB in size and that will kill Ron's hard drive space if I try to upload - lol!

Basically what they are doing with the images - hence why add "isolated on white background" is bringing the image over to PS and selecting the white background, inverting the selection, Ctrl+J for a new layer and then hiding the original layer.

That prompt you have with the Negative Prompt criteria, I have not tried that route.  I mean, I really have been putting it into Gemini as simply as that one sentence and Gemini has smashed it.   I did try Chatgpt but it takes at least 15 minutes for the output to be generated.  It always says something like "this will take some time due to heavy traffic right now" or something like that.

The prompt for this image in Gemini was :
A sticker style image of a hillbilly abduction by aliens isolated on a white background


Nothing else added.  Midjourney fumbles the ball so bad trying to make cinematic outputs.  Gemini saves output at 2048 x 2048.  Was using Chainnr for the upscale but discovered XnConvert which is way more awesome and simpler to use.  Just click three buttons and it is done in 3-4 minutes for a batch of 40-50 images.  Chainnr would take about an hour for that.  Could be better at higher resolutions, but I never went beyond 3x the standard.