Illustrious Lightweight Realistic Image Generation ModelThis is an optimized version of an illustrious image generation model, with a focus on realism and speed while still maintaining high-quality results. The primary goal was to create a lightweight, fast-running model that can run efficiently even on low VRAM devices.The key modifications include:<ul><li>Integration of the DMD2 LoRA adapter</li><li>Reduction in UNET precision to 8-bit</li><li>Lowered generation steps down to only 4-8</li></ul>In testing, I've found that optimal generation parameters are as follows (tested on an M2 MacBook Air with 8GB RAM):<ul><li>Steps: 4-8</li><li>CFG Scale: 1.0 - 1.2</li><li>Sampler: LCM</li><li>Scheduler: Polyexponential</li><li>Upscaler: Foolhardy_Remacri at 2x scale</li><li>Denoising Strength: 0.4 - 0.8</li></ul>Please note that the final image size will be generated from a 512x512, 512x768, etc, base for 4-8 steps with a 2x upscale to achieve a 1024x1024, 1024x1536 resolution output.This model can be run using Automatic1111 or Forge.For best results, ensure your system meets the minimum requirements:<ul><li>CPU with at least 8GB RAM</li><li>GPU with low 8GB VRAM (this model is optimized to reduce VRAM usage) </li></ul>NOTE: does tend to be NSFW, so beware. <img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/9f431194-42f5-4811-a267-0c68d06c6f95/width=525/9f431194-42f5-4811-a267-0c68d06c6f95.jpeg" />Possible upcoming NatVis merge, still in experimental phase...

ReallyIllustrious

This has no merged LoRA's for speed, no UNET precision reduction. Put here for those that want to tinker. Without the optimisations, it's not as good, IMO, the whole idea was to make this lightweight, realistic and easy enough to deploy on many machines, and not another simple merge. As such, experimentation with this hasn't really been extensive, and the only guidance I can offer is to use DPM++ 2M SDE, Karras/SGM Uniform, 40+ steps at 1024x1024 to get a baseline for further experimentation. Your mileage may vary. Happy experimenting.

Files with SHA256: f95fe9465ba9305102e210b13c6a5ccd82839db153cd49a41caa076fbf9ec5cc