The Forest Daughter ( text_to image then image_to_image process in Stable Diffusion)
Hello everyone : for this new collection I tried to use a new approach.
Indeed this time still with the help of AI Stable Diffusion I made images from a prompt that I imagined this one: "realistic portrait of an alien, bright eyes, scaly skin, with hair wood branches and tree leaves, godray, beautiful, detailed, octane rendering, by artgerm, elegant, matte, symmetrical realistic facial features, biblical" I got 5 images.
The 5 first images.
Among these 5 images I selected the one I liked best and which was the closest to the result I wanted to obtain.
I then used it as a starting image in a new creation but this time in image_to_image but still using the same prompt.
And the results are quite surprising I let you see for yourself from the starting prompt then the image I chose here are the results of the image_to_image mode for this second treatment.
*Images obtained with the mode image_to_image / Same Prompt: *
Finally still in this image_to_image mode :I modified slightly the prompt by adding :green for the eyes, complex white lace skin instead of the scaly skin in the prompt what gives : realistic portrait of an alien ,bright green eyes, complex white lace skin, with hair wood branches and tree leaves, godray, beautiful, detailed, octane rendering, by artgerm, elegant, matte, symmetrical realistic facial features, biblical .
Then passing the mose_steps from 90 to 150
and here are the last results.
Images obtained with the mode image_to_image / Modified Prompt:
Great work and post here! Thank you for explaining your process with prompts and refinement. I am still trying to wrap my head around how to "evolve" a creation properly
Very cool results!!! In general I like to delete with photoshop all the "logos" and writing it generates, I don't know it bugs me that much ahaha.
I really like the lips on the last face, very cool.