Adler, how get output only, in 1024 #1

by trysem - opened

what modification has to do to get output image only, also in 1024, instead of stitch of both..

The source is available at and you can get what you wish by commenting out these lines:
aligned_img = array_to_image(aligned_img).resize((output.width, output.height))
output = img_concat_h(aligned_img, output)

Norod78 changed discussion status to closed

hello Adler,
how get good resolution results like below, without 512 crop
mostly, results are blurry, cropped, and low res..
it will be very useful if get at least 1024 without noise, is there any other pretrained?
anyway thanks Adler..

widen_1840x0 (1).jpeg
widen_1840x0 (2).jpeg

widen_1840x0 (2).jpeg
widen_1840x0 (3).jpeg

widen_1840x0 (5).jpeg

The model I've provided here is 512x512. If you wish, you can use the dataset of FFHQ aligned vintage faces to train a model which suits your needs:

Sign up or log in to comment