I wanted to see how much I can get using my simplest “collage sketch” ideas as a basis for image-to-image generation. If i can get away with being simple wit my image guidance images it will speed things up (next I just need to train a LoRa in my style…)
This was a quick idea I had of an image of Abe discovering the a city while walking onto the edge of a cliff.
I used very quick brush strokes and clipped some images from online to use as placeholders. I wanted to see how little detail I could use that would allow the model to re-create the final output without having to rely on Prompt adherence (which is kinda “different” using Flux)
Original Artwork
Re-imagined with Flux
