Tried some 17MP ideogram 4 images for fun
The images are larger than reddit or imgur allows so i'm hosting them on imghippo. the comfyui workflows and json prompts are embedded in the images if you want to view them.
the hardest part about these high mp images is that you really cant tell how the image composition will look until you are quite a few steps into the gen, and at this size that takes 10-15 minutes. so these are just first attempts at each prompt, really just sent it.
I use Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive (IQ4_NL) to generate the json prompts via LM Studio and the YANC_LMStudio comfyui node.
I have a laptop 5090 (sorta like a desktop 5090 that got kicked in the CUDA-coresack and less vram) and it takes between 55 minutes to 1 hour 15 minutes per image depending on steps and cfg. Generally I stuck to 28 - 40 steps. image sizes are around the ballpark of 5472x3264 or thereabouts. I know these won't be everyone's cup of tea, but wanted to share :).
warhammer40k-esque ship: https://i.imghippo.com/files/cgu9505EFA.png wanted to see what grand scale with micro details might look like. i'm impressed with the cityscape on the top and bottom of the ship, at 2mp it looked like pure mush. at least now its semi discernable. has that 40k vibe going on though which is sweet. I made quite a few others of these with even larger ships (continent sized).
millenium falcon: https://i.imghippo.com/files/wuD3658IWM.png two radar dishes, RIP... decent but its one of those cases where I couldn't see the composition until it was far along so I said F it. still, its crisp and the composition is nice
random sword wielding pistol holding mysterious figure: https://api.imghippo.com/files/Vsrc3199uBU.png I found an image similar to this on civitai and had qwen convert it into json.
I upscaled all 3 of these to 31MP as well but I cant find anywhere to host those as they are 100mb+ each, and probably not worth it. but its cool we can native gen such high resolution images with this model. im loving it man.
https://redd.it/1u0qf2z
@rStableDiffusion
The images are larger than reddit or imgur allows so i'm hosting them on imghippo. the comfyui workflows and json prompts are embedded in the images if you want to view them.
the hardest part about these high mp images is that you really cant tell how the image composition will look until you are quite a few steps into the gen, and at this size that takes 10-15 minutes. so these are just first attempts at each prompt, really just sent it.
I use Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive (IQ4_NL) to generate the json prompts via LM Studio and the YANC_LMStudio comfyui node.
I have a laptop 5090 (sorta like a desktop 5090 that got kicked in the CUDA-coresack and less vram) and it takes between 55 minutes to 1 hour 15 minutes per image depending on steps and cfg. Generally I stuck to 28 - 40 steps. image sizes are around the ballpark of 5472x3264 or thereabouts. I know these won't be everyone's cup of tea, but wanted to share :).
warhammer40k-esque ship: https://i.imghippo.com/files/cgu9505EFA.png wanted to see what grand scale with micro details might look like. i'm impressed with the cityscape on the top and bottom of the ship, at 2mp it looked like pure mush. at least now its semi discernable. has that 40k vibe going on though which is sweet. I made quite a few others of these with even larger ships (continent sized).
millenium falcon: https://i.imghippo.com/files/wuD3658IWM.png two radar dishes, RIP... decent but its one of those cases where I couldn't see the composition until it was far along so I said F it. still, its crisp and the composition is nice
random sword wielding pistol holding mysterious figure: https://api.imghippo.com/files/Vsrc3199uBU.png I found an image similar to this on civitai and had qwen convert it into json.
I upscaled all 3 of these to 31MP as well but I cant find anywhere to host those as they are 100mb+ each, and probably not worth it. but its cool we can native gen such high resolution images with this model. im loving it man.
https://redd.it/1u0qf2z
@rStableDiffusion
This media is not supported in your browser
VIEW IN TELEGRAM
AI is watching a film via Marlin(visuals), Whisper(audio), and Pallaidium. Input video by avataraim.
https://redd.it/1u1026k
@rStableDiffusion
https://redd.it/1u1026k
@rStableDiffusion
ComfyUI XY Plot & Find/Replace Nodes + GitHub Suspended My Account (Ep21)
https://www.youtube.com/watch?v=_M-IGh38UkM
https://redd.it/1u17qyb
@rStableDiffusion
https://www.youtube.com/watch?v=_M-IGh38UkM
https://redd.it/1u17qyb
@rStableDiffusion
YouTube
ComfyUI XY Plot & Find/Replace Nodes + GitHub Suspended My Account (Ep21)
In this tutorial, you’ll learn how to use the new Find and Replace Pixaroma node to quickly edit AI-generated prompts, replace subjects like animals or objects, and control prompt changes before sending them into your image workflow.
You’ll also learn how…
You’ll also learn how…