Itrytoblenderrender @ Itrytoblenderrender @lemmy.world

Posts

19

Comments

177

Joined

2 yr. ago

The screenshot is from the comfyui interface. Here is a full view of the workflow (the workflow can be imported in your local comfyui instance if you drag the original picture of my post into the comfyui interface [You may have to install some custom nodes]) :

The idea is: A cozy room with magical windows showing distant exotic lands.

Made with comfyui. The Workflow should be embedded in the image.

This is a bit more complicated regarding the generation process.

• Model: realisiticVisionV51VAE
• Sampler for all: uni_pc
• Steps: for the Base Image 20, everything else 60

Base Image with Multi Area Conditioning:

Negative Prompt: ugly

Base Prompt: cozy medieval living room with 2 windows

Multi Area 1 and 2 Prompt: Train window

This generates a base image which is the base for the further processing.

Window Image 1:

• Positive: Magical, mystical Landscape
• Negative: Person, blurry, boring, close-up, dark (optional), details are low, distorted details, eerie, foggy, gloomy , grains, grainy, grayscale, homogenous, low contrast, low quality, lowres, macro, monochrome (optional), multiple angles, multiple views, opaque, overexposed, oversaturated, plain, plain background, portrait, simple background, standard, surreal, unattractive, uncreative, underexposed

Window Image 2:

• Positive: Raw Photo an alien landscape with alien tentacly purple plants, shot on canon eos
• Negative: Person, blurry, boring, close-up, dark (optional), details are low, distorted details, eerie, foggy, gloomy , grains, grainy, grayscale, homogenous, low contrast, low quality, lowres, macro, monochrome (optional), multiple angles, multiple views, opaque, overexposed, oversaturated, plain, plain background, portrait, simple background, standard, surreal, unattractive, uncreative, underexposed

I take the latent from the base image and combine the latents of the window images with the base image.

After that i do a final sampling to blend all together and 2 Upscales.

Thank you for the friendly feedback! Looking forward for the next contest!

Lies of P

I´ve used Multiarea conditioning for this one. The comfyui Workflow should be embedded in the image.

• AI: Comfyui
• Model: lyriel_v14
• Sampler: uni_pc
• scheduler: Normal
• steps: 40
• cfg: 7
• Upscale Model: 4x Ultra sharp

General Prompt:

RAW Photo of a birds eye view of a medieval city , shot on a canon eos

Left Side area prompt:

daytime

Right Side area prompt:

nighttime, burning buildings

Negative Prompt:

blurry, low quality

Another seed.

The line . They are already building arcologies.

I don't know in which region in Germany you live, but self checkout is quite common in my region and I love it.

Another Seed which came out quite nice:

I am quite confident that your Art has not been used for training in the particular model. But of course I cannot be 100% sure as a stable diffusion model is not a collection of images.

Technically it would be possible to train an existing model on your Art by creating a Lora model which functions like a patch on top of an existing model which forces the output in the direction of your art style.

Recently it would also be possible via a model called Clipvision to feed images from you into the generation process to get a result that is similar to your artwork.

Then we have image to image which would use a existing image from you as base in the generation process.

Lastly controlnet which would read among more things character poses from your images.

All of this is not h the case here. As charming as your character is, I had it not in my mind when generating this picture.

Also I definitely see myself not as an artist. You are an artist doing creative work. This is just a technical hobby for me to heat my office room with my graphic card .

Hi there! I really like your cartoon and am always happy when I see a new issue in my feed.

Any resemblance with your character is purely coincidental. Well I like DnD and currently playing Baldurs Gate 3. So I dusted of my old core rulebook III and started to create creatures.

If you are in any way offended by this, I will happily remove all my posts with this picture.

The Tusks do not want to come out..

You seem to have a special ... taste. Here is a bonus swamp witch for you.

Forest Fairy:

Edgy Elon enters the Board via hyperloop on h4

There is also runpod.io. you can rent quite powerful machines on a hourly base which gives you the possibility to run the large models. Also they have templates so the machine will be set up ready to go in minutes. All you have to do is to load the model you like to try via the oogaboga web interface of your machine.