Well his Song “Eat it” kinda has this theme. But he made many other songs with other themes.
Well his Song “Eat it” kinda has this theme. But he made many other songs with other themes.
They think they are pretty fly when they pick on Al. Probably also white guys.
I don’t understand why everyone suddenly hates Al .
I found a way to make the outpainting seams less obvious.
The adapted Workflow is embedded in this picture (Catbox)
“Melody in the Shadows of Doom”
Made with comfyui. The workflow is embedded in this picture (Catbox)
The basic idea was to create 3 Main pictures and to connect them via inpainting/Outpainting
The base Images and the prompts:
barefooted Tomboy, age 16, sitting on a rock in forest, smiling dreamly, red hair, freckles, green eyey, playing guitar.
a forest, atomic explosion in the sky
A group of undead zombie children sitting in a circle around a pentagram on the ground in a forest
I connect the mages via Outpainting and add a bottom layer to get a bigger picture and a more eerie felling of decay and destruction.
The whole process works with latents. No VAE Decoding and endcoding is needed in the process, which prevents quality loss in the generation process.
I still struggle a bit with the outpainting seams. They are still too visible.
I am curious what your Theme would be if you would win a challenge. I fear the day when i will have to go into the depths of civitai for sailor moon loras.
Interesting read, thank you! I was coincidentially going into this direction. The idea was to create a still life picture without the constraints of a room while still fulfilling the basic parameters of the challenge.
This would have been more appropriate for the concept (still struggling with text):
Still-life with a window in a forest.
Made with comfyui. The Workflow is embedded in this picture (Catbox)
sortiment of exotic fruits on a small rock in the center of the picture in a forest
wooden sign on a post in a forest that says (“Art”:1.2), logo, text
pictore of an window on an easel in the forest (yeah i mistyped, but the outcome as nice)
Workflow:
I am still trying.
I´ve asked the ipadapter to integrate five. But of course he is distorting the reality…
I had a stern talk with the wolf. He promised to do better, but he lost his fire.
Low effort toilet ron.
What’s up with the broken tree defying gravity?
Your choice Magic or bad outpainting
The wolf is looking into the camera
I will reprimand him for ruining the shot
What is the boy thinking?
what do youthink that he is thinking?
Not intentionally. but he has some resemblance.
Cat’s in the Cradle - Ugly Kid Joe
Will definitely look into that. Thank you!
Some months ago there was a comfyui extension for regional prompting which was quite visual and easy to use. Sadly its not maintained anymore.
Yes, comfy workflows seem to have the tendency to get messy 🙃 .
Edit: The ComfyUI_Dave_CustomNode is the one i was talking about. Seems to be dead.
Sibling rivalry.
Made with Comfyui. the Workflow Should be embedded in this picture (Catbox).
I´ve tried to understand regional prompting with this one, but i am only a bit less confused about it now. I was able to define a workflow which gave me roughly that what i wanted to see.
Before you adjust the masks, mute the sampler nodes. When you are satisfied with the positions of the masks, enable them again to generate a picture.
Model:
fenrisxl_V16fp16 (CivitAI)
Prompt: Its complicated because of the regional prompting
The base Prompt is:
A boy and a girl in A beautiful childs room. The boy lays flat on the floor, the girl stands menacing over him.
Regional Prompts:
Full body shot of a 15 year old boy, laying flat on the floor in painful agony
Full body shot of a 16 year old european girl, holding a pointy knife in her hand, grinning evil, threathening stance
Full body shot of a 16 year old european girl, holding a pointy knife in her hand, grinning evil, threathening stance, red hair
CFG: 8
Sampler: dpmpp_3m_sde
Scheduler: Karras
Upscaling:
I used a LDSR Upscaler for the final Image. Its extremly slow but seems to give the best results.
In the Workflow is also a Face Detailer Variant. But here is the Problem that the upscaler Applies the mix of Face expression Prompts on every Face.
Also a Iterative Upscale Variant. This one does not work at all and gives you a completly different final Picture.
Oh, is SD3 already available?
I had the feeling that they tried to merge a dark drama and a slapstick comedy into one movie.
The God killer, especially the intro was quite interesting and dark. I could relate with his anger, disappointment and urge for vengeance.