Finetuned on 100k artworks/images relating to riot games.
https://huggingface.co/Totsukawaii/RiotDiffusion
Please do not post AI images generated from this model online. Please use it only for yourself, or only share it within a closed group. Training images were used without consent.
Usage
Make sure you have at least 12gb vram.
Images were captioned with the format: {aesthetic caption if any}, {manual caption}, {WD14 caption}, {llava llama caption}
Make sure to use "best aesthetic" in the positive prompt
Notes
Anyone can recreate sample images for free using Amazon SageMaker or Google Colab - guide in my discord server
Model still undertrained
Model may be inclined towards anime styled imagery as at least half of the dataset are anime styled images
Links
Join my discord server, for more image samples, dataset tags/captions, free image generation, Q&A/Help, regular updates, and more.
Credits
Stabilty AI - Stable Diffusion XL base model
Riot Games - majority of the artwork
Kohya ss sd scripts - trainer
Description
added Arcane dataset
added images from booru boards, pixiv, artstation, tumblr
FAQ
Comments (20)
I think you have less train data about SETT or forgot him. I can't gen Him. no result about him. Thanks you for your model.
Indeed there is only 15 images of sett in the main dataset whereas there are over 200 images of Ahri. It can't be helped. The model is still in training. I will release a further trained model in about 5 days.
All dataset tags are in my discord server if you are looking for prompts to use.
Some examples I made:
https://cdn.discordapp.com/attachments/1141740755027755030/1162715964006223922/5192b922-b982-46a1-b1a9-230b155b153f.png?ex=653cf2a0&is=652a7da0&hm=64a8de3f819680593e91f0b7d7a3b1b723f071bd1848e5306bad8af222735a1e&
https://cdn.discordapp.com/attachments/1141740755027755030/1162715962735345705/2b88e5b9-8533-4dd9-a3cb-1815a36e5e7e.png?ex=653cf29f&is=652a7d9f&hm=47ad22dce4ec3adb7f0abec49403c66f9446e6a3bf19abd7e14055ee02a6e639&
https://cdn.discordapp.com/attachments/1141740755027755030/1162704382102614106/8da785a9-f909-46e2-8de5-80ebad0c69e2.png?ex=653ce7d6&is=652a72d6&hm=42002a4acebeeee6d2b622e623a981016f160d7d92be47153ce6544bee26fc2f&
https://cdn.discordapp.com/attachments/1141740755027755030/1162704382102614106/8da785a9-f909-46e2-8de5-80ebad0c69e2.png?ex=653ce7d6&is=652a72d6&hm=42002a4acebeeee6d2b622e623a981016f160d7d92be47153ce6544bee26fc2f&
prompt:
best aesthetic, soul fighter, sett, sett \(league of legends\), 1boy, solo, smile, blonde hair, animal ears, male focus, teeth, belt, pants, muscular, abs, pectorals, muscular male, a man with a wolf-like appearance
@xxxxxxx Oh Thank you.
@xxxxxxx Oh Thank you.
@xxxxxxx Oh Thank you.
I have a lora coming up that will have sett in him, There's WAY too many champions to stick in XD -- plus i have heartsteel coming up soon
On your discord do you share how you created it? Tools, training params etc?
yes sir. just ask
WOW！ 使用的1.0版本，虽然手部特别难处理，但总体已经蛮厉害了，非常期待2.0的发挥，一定返图支持您！！！！！！
Hello, I see you use stable diffusion and comfyUI two ways, using comfyUI I successfully restore your example picture, but using stable diffusion I did not restore your example picture, what is going on, do you have other unpublished parameters?
hi, im using telsa t4 gpu with 15gb vram. depending on your hardware you will create a different image. if you want to recreate the same image you can look at my guide to set up sagemaker in my discord server. https://discord.gg/2RBWXFuVV6
@xxxxxxx I was lucky to join your discord channel, but there are so many options, how can I find tutorials on creating the same images as you, and the sagemaker guide you said, how can I find them
@luxiansheng it's in the #guides channel
古德man
The generated image is very poor, different from the one used for returning the image. May I ask what VAE you are using?
The VAE is already in the model
Why is the generated image pure black? If VAE is included in the model, it should be set to automatically recognize VAE, right
can you share the dataset ? I want to try and create a pix art sigma model with this
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.