Model: "HAL": Hellocomic (25) + Addictivefuture (50) + LeviR36 (25)Sampling Method: PLMSdenoising Strength: 0.75512x800 &amp; 2x Hires fix = 1024x1600Use with or without ControlNetCFG Scale: 11Sampling Steps: 40Upscaler: 8X_NMKD-Superscale_150000VAE: <a target="_blank" rel="ugc" href="http://vae-ft-mse-840000-ema-pruned.wae.pt">vae-ft-mse-840000-ema-pruned.wae.pt</a>Clip Skip: 2****After extensive testing, I have found that images created in ComfyUI have occasional facial defects and minor skin abnormalities. However, you can take an image from ComfyUI and put it in A1111 img2img. Select "Just resize (latent upscale)", sampling method "Euler a", sampling steps: 20, restore faces: true, width: same as input (1024),height: same as input (1600),****CFG Scale:1****,Denoising strength:0.15 and leave the prompts blank then A1111 will correct the defects in the face and skin, and also adjust the coloring. It sharpens the face but smooths out other details a little. That being said, this is the best damn model I've seen yet!Check out more sample images: <a target="_blank" rel="ugc" href="https://www.deviantart.com/naraphim">https://www.deviantart.com/naraphim</a>

Files with SHA256: 9d669b25e8eee35e40e2e57613a575caf5bb40567da9835545331759cae09285