Find all my models, scripts, and beta versions on my Ko-Fi.
☄️ IntoRealism
IntoRealism is a model designed for high-fidelity realism. From lifelike portraits to nuanced lighting and natural textures, this model brings your prompts to life with stunning detail, without sacrificing flexibility or creative control.
📢 Join The Community
We're building a friendly space to chat, share creations, and get support.
👉 Click here to join the Discord
You can visit my Ko-fi page if you appreciate my work for a small donation, orders, etc.
⚡ DMD2 Usage
Sampling method: LCM / Exponential or Karras for V3.5
Sampling steps: 8 to 12
Resolution: 896x1152 / 832x1216
CFG: 1Upscaler
4x-UltraSharp or others
Denoising strength : 0.3
Hires CFG Scale : 1Adetailer
Adetailer model: face_yolov9c.pt
Inpaint denoising strength : 0.25 or 0.30SDXL Usage
Sampling method: DPM++ 2M SDE Karras
Sampling steps: > 25
Resolution: 896x1152 / 832x1216
CFG: 4-5Upscaler
4x_NMKD-Siax_200k or other
Hires steps : 4
Denoising strength : 0.3
Hires CFG Scale : 4Adetailer
Adetailer model: face_yolov9c.pt
Inpaint denoising strength : 0.4 (long view)ZIT Usage
Sampling method: Euler / Beta
Sampling steps: 6/10
Resolution: 896x1152 / 832x1216
CFG: 1Description
Version 3.5 is certainly the most realistic one. The faces may look more "natural." I'm trying something different, so there might be a few occasional issues. The model is also more consistent with NSFW content. For an even more realistic touch, you can try using Karras. Have fun : )
FAQ
Comments (5)
3.5 have baked in DMD or VAE?, i am getting cooked results with DPM ++ SDE Karras which you mention in description to avoid but Also using LCM Karras or Exponential with 8-12 steps is somewhat dull or lack details , previous models 1.0 or 3.4 are perfect working for me , has something changed in 3.5? , thanks would be waiting for some guidance.
3.5 An excellent model, with a complex range that conveys the nuances of plein air well.
the absolute most frustrating thing about this model, i want it to have both individuals fully in the frame, no matter what I do, it won't do it.
use kontext for that
There is an SDXL "zoom" LoRa. @ -1.0/-1.5 it's a fully in frame subject