stable diffusion Practice & Testing
zoom in
Original image high resolution enlargement
Original image
When not satisfied with the quality of the picture
Secondary processing using stable diffusion
Choose a model that suits the style of the picture and again write a prompt based on the picture
Adjust the redraw magnitude after inputting the original image at 1024 size.
Samplerautomatic would choose karras here.
Zoom in on the original mutation
a dog,orange_overalls,hard hat,white dog,take spade,in mountain,rain,gloves,
Change cat to dog and increase the redraw margin to 0.4
If the redraw amplitude is too high, sd will generate images directly according to the prompt words without referring to the user's images, which will be beyond imagination. It is recommended that HD zoom repainting amplitude control at 0.35
If you want bad cue words you can use GPTS to generate them
stylistic shift
Anime to Live Action
Real life without Anime6B is better
Live Action Anime
coloring in
Different styles for different models