It's a merged model, that merged some realistic models as realize for anime characters.I make it as hobby, so personal use for it. not for commercial.Thanks for creators of the original models that use in this model.hugging face page for more details, see <a target="_blank" rel="ugc" href="https://huggingface.co/fcski/real_model_L">https://huggingface.co/fcski/real_model_L</a>if you use webui, I recommend to use 1.5.x, not 1.6.The reason is that the implementation of the samplers has changed and the reproducibility seems to have decreased. (I try hello asuka test for traditional samplers, between 1.5.2 and 1.6.0. Especialy, DPM++ SDE, DPM++ 2M SDE, DPM++ SDE Karras, DPM++ 2M SDE Karras are changed.)There may be some option to fix this, but I haven't found it yet.samples created by webui 1.6.0(untill cat), or 1.5.2(yuina senpai, astolfo), 1.3.2(yukikaze).samples use follow LoRAs:- shirakawa yuina 白河ユイナ(heaven burns red): <a target="_blank" rel="ugc" href="https://civarchive.com/models/125224/shirakawa-yuina-heaven-burns-red">https://civarchive.com/models/125224/shirakawa-yuina-heaven-burns-red</a>- Astolfo (Saber) | Fate/Grand Order: <a target="_blank" rel="ugc" href="https://civarchive.com/models/125302/astolfo-saber-or-fategrand-order">https://civarchive.com/models/125302/astolfo-saber-or-fategrand-order</a>- Mizuki Yukikaze (Taimanin games): <a target="_blank" rel="ugc" href="https://civarchive.com/models/89332/mizuki-yukikaze-taimanin-games">https://civarchive.com/models/89332/mizuki-yukikaze-taimanin-games</a>(Fixed different referenced LoRA)I'm not a native speaker, so sorry for my poor English.

real_model_N

its a first release.update is not planned.

realModelN_realModelN.safetensors

22 years old korean female soldier, unbuttoned uniform, korean_commentary , korean_text ,magazine

ASIANSAKURA-PC


NEGATIVE: (worst quality, low quality:2)
⛔ no need to add really really long prompts like you're writing on thesis, or you'll got worst results 
Sampler: DPM++ 2M Karras
Step: 25
CFG scale: 7
Clip skip: 1
VAE: mse-840000
Hires fix Upscale: 1.5-2.0x
Upscaler: Latent, 4x-UltraSharp
Denoise: 0.4-0.45
*you can also use clip skip 2 to create faces (and some minors) variation.

RECOMENDED SETTINGS FOR BEST RESULT

Don't forget to use hires fix using settings above.
For face (optional), you can use ADetailer by choosing face model and set inpaint denoising strength to 0.3.

Files with SHA256: c2fb34f2e1980cc3a624e955bcda5d0cc583782c63a68ff4adbaf86d07a83e1c