Lemmy seems to scale the image down. The original resolution is: 3072x2048

The comfyui workflow should be embedded into the picture. For this one i wanted to try the outpainting process in comfyui.

I used chatgpt to generate the Prompt:

“Please give me a descritpion of a surreal unsettling cartoon style comic page”

The Prompt:

Prepare to venture into the unsettling depths of the unknown with “Eerie Enigma,” a surreal cartoon-style comic page that will leave you with a sense of profound unease and curiosity. This disconcerting page draws you into a world where reality unravels, and every panel is a portal to a bizarre and unsettling dimension.

The page is dominated by a nightmarish dreamscape, where the sky is a sickly shade of green and jagged, twisted trees reach out like skeletal fingers. Moonlight casts long, distorted shadows that seem to have a mind of their own, morphing into eerie shapes that crawl across the ground.

Characters are equally unsettling: a faceless, cloaked figure with hollow eyes that follow your every move, a humanoid creature made of fragmented, disjointed limbs that seem to float in mid-air, and a spectral, translucent child who whispers cryptic messages that send shivers down your spine.

The narrative unfolds in a non-linear, disorienting fashion, with fragmented, surreal sequences that defy conventional storytelling. In one panel, a clock melts into a puddle, its hands spinning wildly as it oozes into a black void. In another, a door opens to reveal an infinite corridor of twisted, impossible architecture.

The color palette is dominated by sickly greens, blood-red accents, and deep, inky blacks, creating an oppressive atmosphere that seems to close in on you. Unsettling sounds—whispers, distant screams, and eerie laughter—seem to emanate from the very page itself.

“Eerie Enigma” is a disconcerting journey that challenges your perception of reality and invites you to confront the uncanny. It is a nightmarish, surreal experience that lingers long after you’ve turned the page, leaving you with a profound sense of unease and a burning desire to unravel the enigma that it presents.

I used comfyui as Generator.

Stable Diffusion XL is better at processing a lengthy prompt. So i used a XL Model

• Model: fenrisxl_801Photorealistic
• Positive Prompt: See above
• Negative Prompt: signature, Text, credits
• Negative Embedding: unestheticXLv31
• SDXL Refiner: None
• CFG 7
• Sampler first KSampler dpmpp_3m_sde
• Sampler all following KSamplers: ddim
• Scheduler: Karras
• Steps: 40

Also:

i´ve integrated FreeU for allegedly better images into the process and a latentgarbagecollector for better performance.

If you pull the process into your Comfyui instance you will have to install some custom nodes.

The Process:

In the first step, i generate a 1024 x 1024 Image with above parameters:

In the next and follwing steps, i enlarge the canvas and do an inpainting with the same parameters in the enlarged area:

I do this for several iterations until iget a nice larger picture. For this image i did not take the last image but one in the process beause in later steps the picture got a bit wild:

I think outpainting is a nice tool you can use in the comfy workflow.

Theoretically you can generate extremly large images with outpainting.

You could also use a completly new or modified prompt for the outpainting which should give you some interesting results.