Furkan Gözükara
patreon
1k Resolution Class Images & Direct SDXL 0.9 and 1.0 Download Links
🕑 Added 2023-07-12 11:08:35 +0000 UTC27 July : Added SDXL 1.0 official links
Join discord and tell me your discord username to get a special rank : SECourses Discord
LoRA SDXL training tutorial : First Ever SDXL Training With Kohya LoRA - Stable Diffusion XL Training Will Replace Older Models
New 1024x1024 Unsplash Real man class images : https://huggingface.co/MonsterMMORPG/SECourses/resolve/main/1024x1024_2734_imgs.zip
Can be used for SDXL training
Class images are addition to my this post : https://www.patreon.com/posts/4k-2700-real-84053021
I didn't add password this time just download and extract
For extracting in RunPod execute below commands
- apt update
- yes | apt-get install p7zip-full
- 7z x 1024x1024_2734_imgs.zip
SDXL 1.0 Base safentensors > https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/resolve/main/sd_xl_base_1.0.safetensors
SDXL 1.0 Refiner safetensors > https://huggingface.co/stabilityai/stable-diffusion-xl-refiner-1.0/resolve/main/sd_xl_refiner_1.0.safetensors
SDXL 0.9 Base safentensors > https://OwlMaster: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@huggingface.co/stabilityai/stable-diffusion-xl-base-0.9/resolve/main/sd_xl_base_0.9.safetensors
SDXL 0.9 Refiner safentensors > https://OwlMaster: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@huggingface.co/stabilityai/stable-diffusion-xl-refiner-0.9/resolve/main/sd_xl_refiner_0.9.safetensors
You can download above links with wget in RunPod
wget URL
Examples
- wget https://OwlMaster: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@huggingface.co/stabilityai/stable-diffusion-xl-base-0.9/resolve/main/sd_xl_base_0.9.safetensors
- wget https://huggingface.co/MonsterMMORPG/SECourses/resolve/main/1024x1024_2734_imgs.zip
- wget https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/resolve/main/sd_xl_base_1.0.safetensors
Comments
Walt
hello and thanks for the supir-installer unfortunately I currently have the problem that from the 49.v2 and currently also the v3 does not start, I get this message:
Furkan Gözükara
it is released have you seen it?
Furkan Gözükara
hello. i plan to release them hopefully this weekend
Khoi Dam
Looking for the other class too
Furkan Gözükara
Thank you so much. I plan to work on this hopefully this week
James R
Hope you had a good vacation? I will be keeping an eye on this when you return! The other class types are definitely high on the wishlist here!
Furkan Gözükara
sure just private messaged you
JustDrwIt
I would love to get a copy of that scraper too if you wouldn't mind. I tried a couple I found on github, but I couldn't get them to work.
Furkan Gözükara
comfyui don't have training options. you need to train with the Kohya script and move training checkpoint into the comfyui folder as shown in tutorial video. however i would suggest you to use automatic1111 for using loras.
Furkan Gözükara
hello. it is on my todo list. hopefully i will release even bigger dataset for both man and woman
Dallin Mackay
I would love a female reg image database
Rasika Singal
where will the training be available for comfy ui?, thanks
Furkan Gözükara
ok then using high quality real images will be certainly much better. i have all the preprocessing tools. now i will send you unsplash scraper as well from private message
Oinksauce
Definitely realism at this point - Love making images of the people in my life.
Furkan Gözükara
it depends on purpose. if you are aiming realism, unsplash like real images will perform better. but if your aim is styling, then generating images from the model itself will keep the model style. if you are aiming a certain styling, then using such styling class images will be even better. i think i can prepare more class images hopefully for those classes as well. but i am on vacation atm. thank you for support
Oinksauce
Yes! Which is awesome, but was hoping for more classes other than just "Man." I'm looking to make some SDXL LORAs of my kids 9-14 (the class "Person" worked well for Dreambooth) as well as my wife "Woman." Not sure if people get best results by using real images, or if I should just generate a bunch using SDXL itself.
Furkan Gözükara
If you need anything else please let me know. i am on vacation but i will try to prepare them hopefully once i returned back
Furkan Gözükara
I also have some more class images here : https://www.patreon.com/posts/84292083
Furkan Gözükara
hello. more resolutions are posted here did you check them out? : https://www.patreon.com/posts/4k-2700-real-84053021
Oinksauce
I'd love to see more class reference images at all resolutions! it would be awesome to have a direct link to those.
Furkan Gözükara
auto1111 started working on SDXL. hopefully coming soon to auto1111
Furkan Gözükara
not working with auto1111 but working with comfy ui. i am preparing a tutorial for that as well
indievish
Looking forward to LoRA and DreamBooth training on SDXL. Does it still work on Automatic 1111?
