#img2img

subjects = ["the void", "a shadow", "the echo"] verbs = ["swallows", "haunts", "follows"] objects = ["your dreams", "the silence", "the light"]

print(f"{random.choice(subjects)} {random.choice(verbs)} {random.choice(objects)}")

Coming Into That Old Money. ---

Oppenheimer if he a girl and nuclear weapons development is conducive to stress-eating~

Expector

Co-creation with @morphovision ✊ (source image)

Unexpected collaborative process: Work born from an online but indirect dialogue, using mostly images as language.

Both, the humans and the machines involved.

Both, the organic and the artificial intelligence.

Rice Shower

From there it's a matter of tuning the denoising strength, cfg strength, prompting and weighting the lora styles to create all sorts of faces.

Giving it all of the creative freedom by turning up the denoise high and CFG low can yield some very weird or cool results.

So I think the sweet spot to create a unique looking character face with this kind of input is usually going to be a high denoise value and a moderate CFG value.

Keeping the denoising level low can render a facial onto that dummy face, making for a great source of simple references.

For more variety, why not turn clip skipping up from 1 to 2?

Thanks for scrolling this far; I hope you enjoyed looking at my results and got some ideas from it.

I will be posting more as I continue to learn so if you have questions or comments, feel free to get in touch.