Not working, even when enabled
beped opened this issue · 16 comments
Im having a weird issue. I have installed the extension on my main 1111 ambient, the extension didnt work at all. I assumed tha was my installation so i installed on a brand new 1111 ambient, only with the Latent Couple extension installed. Same results. Somehow, looks the extension its all set and working, but actually its not doing anything with the image.
No erros on the CMD prompt either.
The info of the image:
flower garden background
AND beautiful cute woman, long blond hair
AND a man, short brown hair
Negative prompt: (deformed, distorted, disfigured:1.3), poorly drawn, bad anatomy, wrong anatomy, extra limb, missing limb, floating limbs, (mutated hands and fingers:1.4), disconnected limbs, mutation, mutated, ugly, disgusting, blurry, amputation, watermark, text, disabled body, paintings, sketches, (worst quality:2), (low quality:2), (normal quality:2), lowres, normal quality, ((monochrome)), ((grayscale)), skin spots, acnes, skin blemishes, age spot, glans,wrong amount of fingers,
Steps: 20, Sampler: Euler a, CFG scale: 7, Seed: 2834497771, Size: 512x512, Model hash: 1dceefec07, Model: dreamshaper_331BakedVae, Latent Couple: "divisions=1:1,1:2,1:2 positions=0:0,0:0,0:1 weights=0.2,0.8,0.8 end at step=20"
I am noticing two hair colors, which seem to correspond to the prompt. I think it's just decided to only create one person with both qualities showing. Sometimes even this extension does do that; you could maybe add "man and a woman" into the overall prompt or something to guide it toward actually creating two distinct people.
it still didn't work, or I'm the luckiest person on the planet. I tested with about 15 seeds. When i put higher resolutions, the results are a little better, but not even close to the prompt
Prompt:
(8k, RAW photo, best quality, masterpiece:1.2), (realistic, photo-realistic:1.4), best quality, ultra high res, (photorealistic:1.4), city background, 2girls
AND (8k, RAW photo, best quality, masterpiece:1.2), (realistic, photo-realistic:1.4), best quality, ultra high res, (photorealistic:1.4), 2girls , (smiling at camera:1), lora:tifa:0.5, lora:dosmix:0.6, lora:koreanDollLikeness_v10:0.2 ,lora:japaneseDollLikeness_v10:0.3, (black jacket), slim body, abs, slim legs, ((black jeans)), normal hand, large breast, ((brown hair)), ((white shirt))
AND (8k, RAW photo, best quality, masterpiece:1.2), (realistic, photo-realistic:1.4), best quality, ultra high res, (photorealistic:1.4), 2 girls , (smiling at camera:1), lora:aerith:1, lora:dosmix2:0.6 ,lora:koreanDollLikeness2_v10:0.2, lora:japaneseDollLikeness2_v10:0.3, normal hand, ((blue eyes)), (((flat chest))), (brown hair), ((red long dress))
Some of the prompt keyword are not triggered.
i have the same issue. Can't access the mask area in the Extension in A1111. Under Latent Couple its the same view as stated above in #25 (comment)
When trying to run the patch i got following error:
git apply --ignore-whitespace extensions/stable-diffusion-webui-two-shot/cfg_denoised_callback-ea9bd9fc.patch error: patch failed: modules/script_callbacks.py:68 error: modules/script_callbacks.py: patch does not apply error: patch failed: modules/sd_samplers_kdiffusion.py:8 error: modules/sd_samplers_kdiffusion.py: patch does not apply
i have the same issue. Can't access the mask area in the Extension in A1111. Under Latent Couple its the same view as stated above in #25 (comment)
When trying to run the patch i got following error:
git apply --ignore-whitespace extensions/stable-diffusion-webui-two-shot/cfg_denoised_callback-ea9bd9fc.patch error: patch failed: modules/script_callbacks.py:68 error: modules/script_callbacks.py: patch does not apply error: patch failed: modules/sd_samplers_kdiffusion.py:8 error: modules/sd_samplers_kdiffusion.py: patch does not apply
You don't need to apply the patch if you are using the latest version of A1111 web UI.
Im having a weird issue. I have installed the extension on my main 1111 ambient, the extension didnt work at all. I assumed tha was my installation so i installed on a brand new 1111 ambient, only with the Latent Couple extension installed. Same results. Somehow, looks the extension its all set and working, but actually its not doing anything with the image.
No erros on the CMD prompt either.
You need to use correct prompt.
((ultra-detailed)), ((illustration)), (a man and a girl) AND ((ultra-detailed)), ((illustration)), (a man and a girl) a girl with (black hair) AND ((ultra-detailed)), ((illustration)), (a man and a girl) a man with (white hair)
more than 2 girls are bad.
You need to use correct prompt.
((ultra-detailed)), ((illustration)), (a man and a girl) AND ((ultra-detailed)), ((illustration)), (a man and a girl) a girl with (black hair) AND ((ultra-detailed)), ((illustration)), (a man and a girl) a man with (white hair)
This doesn't work. There are two subjects, but they don't correspond with the divisions at all.
I am also having the same error. If you can find a solution, I would love to know!
error: patch failed: modules/script_callbacks.py:68
error: modules/script_callbacks.py: patch does not apply
error: patch failed: modules/sd_samplers_kdiffusion.py:8
error: modules/sd_samplers_kdiffusion.py: patch does not apply
I'm getting the same thing. It used to work, and sometimes it still will, but usually it merges my two characters and puts both of them (now one person) on the left or right side. If I disable the extension the results look completely different, so it's doing something, but it doesn't work anymore.
EDIT: Oh shit I might have just figured out the problem. It looks like if there's too much in your negative prompt, the thing freaks out. I removed a few words from mine and it started working again. Removed a few more, and it started working better. I suspect the problem is the negative prompt.
I don't know if something changed, but I have updated everything and reinstalled the extension and latent coupling is working for me now. If you guys get it working, here's a helpful site to set up your parameters: https://badnoise.net/latentcoupleregionmapper/
Ok, so I was having this exact issue and through painful elimination realised it was due to not giving the canvas enough space.
a city in background, golden hour, morning AND
a young nordic woman wearing a long coat AND
an old jamaican man with a walking stick512x512
1024x512
Hopefully this helps someone! 🏞️😁
Thanks for the nice comment.
I made a mistake in my setup and it worked fine.
Thank you so much.
(I curse myself for being so stupid)
additionally, I noticed if you use a ControlNet, you can reliably get it to work on square (512x512) aspect ratios fwiw.
additionally, I noticed if you use a ControlNet, you can reliably get it to work on square (512x512) aspect ratios fwiw.
Can you provide more details about how you would need to go for this to happen?