LoRA trained on Aoba Suzukaze from New Game!!
requested by: @RiderBond1949
default: suzukaze aoba, 1girl, hair ornament, suit, twintails, bow, purple hair, purple bow
Cardigan Outfit: suzukaze aoba, 1girl, hair ornament, school uniform, cardigan, twintails, purple hair, black bowtie
gym outfit: suzukaze aoba, 1girl, gym shirt, buruma, hair ornament, twintails, purple hair
models used
anime: YesMix - v2.0, YabaL_Mix - V5, StingerMix - v3.1, Sweet-mix - v2.0-flat, MeinaPastel - V6, CottonCandy Pastel - v1.0
2,5D: Animesh - Animesh-Pruned V1.5
negative embeddings: bad_prompt, negative_hand, ng_deepnegative_v1_75t, bad-artist, EasyNegative
Recommended settings:
Make sure you are aware on the usage instructions of LORA
VAE: vae-ft-mse-840000-ema-pruned (for realistic models & 2,5D), kl-f8-anime2, blessed2.vae (for anime models)
Sampler: DPM++ SDE Karras or DPM++ 2M Karras(euler a for img2img upscaling) from time to time im also using DPM++ 2M alt Karras
Steps: 20 to 35
Clipskip: 2
Upscaler : GitHub - pkuliyi2015/multidiffusion-upscaler-for-automatic1111 4x-UltraSharp(for 2,5D & realistic), 4xAnimesharp(for anime)
CFG Scale : 5 to 9 (mostly 7)
LORA weight for txt2img: 0.5-1 (for the examples i mostly used 0.8 and 1)
Denoise strength upscaling: 0.3 to 0.55
If you want to autocomplete the tags & trigger im using, I recommend using Booru style tag autocompletion for AUTOMATIC1111's Stable Diffusion web UI
For full body shots i recommend using: ComfyUI Impact Pack
or the a1111 alternative GitHub - Bing-su/adetailer
Description
FAQ
Comments (10)
thank you very much, I will test later ;)
I would have some more requirements for LORAs (at least 2) , but I don't know if I would not burden you with that 😏..... do you follow any specific tutorial when creating LORAs available somewhere on the internet?
Sure, as long as it's not already available on this platform or is a LoRA that needs to be remade but hasn't been, I'll do it. I don't usually follow tutorials, but this guide is pretty good for starting out. Additionally, if you have a weak GPU, you can refer to this Google Colab guide
@AhriMain OK, so tomorrow I'll post some tips for LORAs :)....
I have a gtx 1080 ti for now (I think i'll older sooner, before it generates anything 😅), but I'm planning to upgrade, which graphics would be the most suitable, more powerful (for faster generation) or with more than 16GB VRAM (for generating higher resolutions) ?
@RiderBond1949 For generation, you don't really need that much VRAM anymore since Tiled Diffusion exists. For training, it would be really useful to have more VRAM though. The only important thing is to buy a new NVIDIA GPU (30x & 40x) for AI tasks, not AMD
@AhriMain Thanks for the tip... I heard about Tiled Diffusion, but there was a lot of text, so I skipped it 😅... but I should read it properly 😏
@RiderBond1949 For upscaling, you can skip most of it. You only need the img2img upscaling tab & Tiled VAE (The Default settings are optimized for your GPU, so you can simply use the default settings). Stable Diffusion – Microsoft Edge (gyazo.com)
So, I have a request for some more LORAs, I couldn't find them anywhere and as for the materials, there aren't many of them (it's probably a good idea to use the anime materials, there's a lot of them 😁, easily downloadable on nyaa.si), so I don't know if you'll be able to work it out, but if you do, I'm sure the result will be worth it! 🤩 These are the characters: "Bloodstone Fino", "Do Aximemor Lamdimia" and "Ortinet Airi" from the anime "Yuusha ni Narenakatta Ore wa Shibushibu Shuushoku wo Ketsui Shimashita".
I don't know if I could fit it into one LoRA, if not, I'll just split it into three separate ones.
@AhriMain No, I didn't mean together... Ideally you should make them separately, like this LORA, hopefully you won't have any problems in realization and the resemblance to the original material will be very good.
So I hope you can work it out, and if you have a patreon page or something, I could support you, you're doing a great job 😃👌💪.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.