AiPlaygroundWM

 Generate with stable diffusion WebUI A1111 Go to Google Gemini Nano Banana and enter the image and prompt Making fake magazine cover. the word is " Fake Ai magazine " Looks just fine. Go to stable diffusion WebUI A1111 img 2 img upscale Image quality drop / Broken text Hm....It seems like decorating it in Photoshop is a better way to go

1024 *1024 as the training resolution But  A test that started with the question: How high can you go with generated resolution? LobotomyMix (I guess merged with ilxl v0.1 base) ILXL 1.0 trained with a resolution of 1500 *1500 Assume the limit of the models merged with ILXL 0.1 is 1500 *1500. The resolution implemented was 1400 * 1752 (4:5 ratio) 1304 * 1624 Conclusion : There are a lot of NSFW images, so I can't post them all on the blog,  To summarize the observed patterns, Additional human body generation, human body distortion, stretched proportions (similar to what was created in SD1.5 when I increased the base resolution) Limitations: 1. only one checkpoint was tested                   2. the subject is a monster and I applied the oversized hands prompt. P.S.   All image has exif info. You can find me on  instagram  ,  deviantart  

In the middle of generating veriety version of the vampirella. parameters (full body:1.2) photo of Vampirella, solo focus, look at viewer, very long hair, micro bikini, muscular, curvy, (huge breasts:1.35), abs, narrow_waist, huge pelvic, (thick thighs:1.35), high_heels, boots, (dynamic pose, fighting stance:1.2), blood_puddle, partially submerged, body half-submerged, (full moon), forest, Blood everywhere, (best quality, realistic) <lora:VampirellaV2:0.4> v4mp1r3ll4, <lora:angel_wings_v10:0.6> (angel_wings) Negative prompt: (breasts out, nude:1.4), extra hand, extra foot, close up, (multiple view), missing finger, extra fingers, fused fingers, cropped, (crossed eyes, worst quality,bad quality:1.4),monochrome, greyscale, (text, signature, watermark:1.4), mutated Steps: 15, Sampler: DPM++ 2M, Schedule type: Karras, CFG scale: 7, Seed: 1489340367, Size: 512x640, Model hash: b89ba7f884, Model: meichidarkV38_pruned, Denoising strength: 0.55, Clip skip: 2, This one As you can se...

 Check Point model :  protovisionXLHighFidelity3D 816 * 1024  Hires 1.5 Denoising strength : 0.45 Fairy tale , fantasy style, 1Girl,  a human  embracing posing with a giantfusion of unicorn, Magical,  storybook style, enchanting, highly detailed, (realistic photograph, best quality, finely detailed, absurdres) negative prompt : mutated, extra arms, (worst quality, low quality, glitch, artistic error), (Watermark, Tex,signature), blurry, 15 40 Same as Stable diffusion 1.5. More steps not any guarantee better quality. (At least, in my settings) Changing check point model :  RealVisXL Changing prompt : (realistic photograph),  red and greenish color theme,  1Girl,  fashion model posing with a giantfusion of horse,  in nature,  enchanting, highly detailed, (best quality, finely detailed, absurdres) There is no detail both. That hexagon artifact, every surface.... check point :  realcartoonXL_v6 Same results. Hexagon ...

 New process test base  Regenerate hires 2.5 x i2i upscale Hands get worse. Inpaint done except hand. Edit right hand Run i2i inpaint, using cotrolnet(canny, mlsd, nomal and etc....most of controlnet model) Control net image is reference.(Edit position with photoshop) Failed . Not even close. Paste reference Run again with control net. About 300 images. This is not much.  But still most of test results are so much fuxxed up. And generate and reboot repeat. (cause of vram or memory issues.) Choose one. Looks ok. Edit and paste the inpaint result.  (The left hand just inpainting and edit.) Edit detail and final image Conclusion : Complex process and the edited hands parts are still noticeable .