- Jan 20, 2022
- 208
- 2,043
I do something similar, but instead of using refiner, I do the checkpoint switch on highres.fix, i'll do a comparative of wich one works better, stay tune.
I've installed both the
For those not familiar it almost turns SD into DALI with the way it handles drawing objects, less turning clouds into hair.
Here is a
Here is a
You must be registered to see the links
and the
You must be registered to see the links
for
You must be registered to see the links
normally I like building my own pipelines but if anyone has a good SEGS setup or can link to one.For those not familiar it almost turns SD into DALI with the way it handles drawing objects, less turning clouds into hair.
Here is a
You must be registered to see the links
by someone with way more experience then I currently have using it, he also created a UI manger but I haven't gone that path.Here is a
You must be registered to see the links
showing a form of auto mask generation for inpainting, I have seen combinations of SEGS with Automasking and auto prompting from image detection but have never found a shared workflow.
Last edited:
- May 18, 2018
- 199
- 2,028
- Jan 20, 2022
- 208
- 2,043
- Jan 24, 2020
- 1,401
- 3,794
Here's an image with some useful photography terms that describes the composition. Instead of "cowboyshot" you can try "medium full shot". I guess you could also try to describe what should be included in the image.
"knees thighs torso and head in image" or similar phrasing.
Alternatively simply use the DeepBooru style of tags: knees, thighs, torso, head (face).
- Dec 31, 2016
- 316
- 708
Hmmm, forgot about the "medium full"...should fix all the cowboy dressup issues, question is if the AI will know it and if it'll split it up in many cases.
So far it seems to be treating it as just a medium shot, with the occasional full.
This could obviously be down to model and/or seed
So far it seems to be treating it as just a medium shot, with the occasional full.
This could obviously be down to model and/or seed
You must be registered to see the links
Trained for native 1k generation and up scaling to 2k (No Highrez fix)
I thought with the BF16 and ability to train beyond 1k, why not try to teach SD to put details in the right spots at high resolutions. (Even at .5 normal sampler img to img at 2k)
Nude female image training, non pornographic
anyone have a navel penetration lora or prompt? i have managed to create the scenes with controlnet but it takes too much time doing it manually since stable diffusion really doesn't understand the concept, understandably xD.
Putting both at the same time will combine the two, you can in-paint the face or body using masking or adetailer.
Or simply run the image twice with one lora then the other and combine the images in Gimp.
Testing how the
In this case
Native Generation in SD at 768x1024
2k Upscale
While it looses some of the body detail, considering 2048 is more the double the expected render resolution, I would be curious to see if you could train XL to generate at 2k and upscale natively at 4k, I don't have the PC stats for that though.
You must be registered to see the links
works with other lora's.In this case
You must be registered to see the links
Native Generation in SD at 768x1024
2k Upscale
While it looses some of the body detail, considering 2048 is more the double the expected render resolution, I would be curious to see if you could train XL to generate at 2k and upscale natively at 4k, I don't have the PC stats for that though.
- Dec 31, 2016
- 316
- 708
You should be able to do both 4 and 8k natively just fine, both with and without upscaling, at very little to no loss.
I can't remember what it's for XL as that is far beyond what i could even hope to work with so i've not been digging for info about it.
Also interested in seeing how Würstchen will work and do.
On a sidenote, that lora seem to not care about likeness, but considering they don't seem to have cropped out the text in the training data it's probably too big an ask to make it look like the actual person.
I can't remember what it's for XL as that is far beyond what i could even hope to work with so i've not been digging for info about it.
Also interested in seeing how Würstchen will work and do.
On a sidenote, that lora seem to not care about likeness, but considering they don't seem to have cropped out the text in the training data it's probably too big an ask to make it look like the actual person.
- Dec 31, 2016
- 316
- 708
I know what they were trained at, not what i meant.
Anyway with Würstchen i was more thinking about their claim of being much faster than XL and what that means for memory usage and compatibility with older system.
Same with AITemplate i guess, but i can't set that up because ppl don't bother buildiing things well, unfortunately.
Anyway with Würstchen i was more thinking about their claim of being much faster than XL and what that means for memory usage and compatibility with older system.
Same with AITemplate i guess, but i can't set that up because ppl don't bother buildiing things well, unfortunately.
Read both papers and draw your own conclusion. Both models are limited to about 2x the trained latent image generation size even when finetuned.
You must be registered to see the links
You must be registered to see the links
Almost all the models self evaluate or AI evaluate against recreation of the FID-COCO-30k set of images
Last edited:
- Jan 31, 2022
- 212
- 634
Random test I did using the same prompt with different models. The pose is kept the same using controlnet.
The pictures are named for the model used.
The pictures are named for the model used.
You don't have permission to view the spoiler content.
Log in or register now.
- Jan 10, 2019
- 268
- 1,440
After a week mucking about with ComfyUI and not getting any good results with faces, went back to SD and already much better:
- Mar 3, 2019
- 232
- 746
Thought I could share a tip here for anyone looking to create consistent-looking characters without having to rely on LoRAs, using just some openly available checkpoint and prompting: if you ask for a mix of a few known celebrities, SD will create a person that merges the facial features of all of them into a "new" person, but this way you can pretty consistently output that same person.
For example, on every image below I had
For example, on every image below I had
mix of Sarah Shahi and Vanessa Hudgens and Nina Dobrev as part of the prompt, with a few differences on the rest of the prompt for each picture to describe clothing, visual style, etc. Hope this helps (and that it wasn't already explored on the previous 100-something pages).
- Jan 24, 2020
- 1,401
- 3,794
You can also do this either with tag mixing or keyword weighting.
Tag mixing:
As you can see it's essentially like using a "refiner". The first person is the main checkpoint and the second is the "refiner" but the weight is reversed. This means that if you want the "refiner" to have more impact you need to increase the number. If you want the checkpoint to have more power the opposite is true. Don't use "[ ]", use normal brackets "( )" instead. It tend to give you and error with SD if you use "[ ]" .
Keyword weighting:
This can be used for blending more than 2 faces.
(Emma Watson:0.5), (Tara Reid:0.9), (Ana de Armas:1.2)
Source:
You must be registered to see the links
Last edited:
- Jan 24, 2020
- 1,401
- 3,794
Bonus tips..
You can generate a portrait with either SD1.5 or XL and let the SD only focus on the face regardless of being a mix or not. This will give the face more detail. Then you simply use this face with the extension "roop" when you are generating your character. You can also use this method with img2img to make "photobashing". Meaning taking a photo and using "roop" on top of the photo to create fakes.
When using img2img roop will not be part of the generative process, it will only paste the face on top of the existing bone structure. This will obviously not give a good result in all scenarios. For best result use this method with txt2img when generating a new image.
Face:
Character using roop:
Using a photoshoped photo of Angelina with blonde hair in roop:
Without roop:
You can generate a portrait with either SD1.5 or XL and let the SD only focus on the face regardless of being a mix or not. This will give the face more detail. Then you simply use this face with the extension "roop" when you are generating your character. You can also use this method with img2img to make "photobashing". Meaning taking a photo and using "roop" on top of the photo to create fakes.
When using img2img roop will not be part of the generative process, it will only paste the face on top of the existing bone structure. This will obviously not give a good result in all scenarios. For best result use this method with txt2img when generating a new image.
Face:
You don't have permission to view the spoiler content.
Log in or register now.
Character using roop:
Using a photoshoped photo of Angelina with blonde hair in roop:
You don't have permission to view the spoiler content.
Log in or register now.
Without roop:
