How to generate vertical images？ i.e. height=1376, width=760
#9
by
LOOKCC
- opened
I found that when i want to generate vertical images, the composition of the picture is strange, and the characters are out of the frame.
prompt = "city center, boy leaning on the railing of a skyscraper's rooftop, casual attire, shirt underneath."
image = pipe(prompt, height=1376, width=760, num_inference_steps=50, guidance_scale=3.0).images[0]
Is it because only a 1:1 aspect ratio was used during training?
Yes, multi-AR version will be released soon.
ehsanakh
changed discussion status to
closed