For V.2 read "About this Version" The realistic images that i post i make them with CyberRealistic CyberIllustrious , for that i use the Highres-Fix Script (comfyUI) and use dual checkpoints, you can use any checkpoint you want...it's real nice:DBest result : Sampler euler_A , simple , cfg, 3-5 , stepes 35 , (try different ones , test it and it may produce some nice results) merger of stable models, i have no idea how it ended this way (for Highres-fix-script from efficiency nodes 1.25 upscale , 14 steps , 1-2 iterations and 0.36 denoise) (for face detailer from impact-pack 8 steps with 0.36 denoise) if you use comfyui just load the image you like in it and run the workflow...use artists names likeartist:wamudraws:1), (artist:slugbox:1), (artist:zankuro:1)- (artist:mazjojo:1)- (artist:kotteri:1), (artist:pigeon666:1) (artist:makoto-shinkai:1) (artist:zawar379:1) (artist:yoneyama mai:1) etc,bad with backgrounds and detailsLora use 0.5 -1 strength for 1-2 and 0.2-0.5 for multiple (i think)My images have the workflow embedded in them , use this link to load workflow to see what it hasMy go to positive prompt : - embedding:lazypos, (masterpiece, best quality, amazing quality, ultra-HD, high detail, very awa, newest, very aesthetic, ultra-detailed, absurdres, 8k), it has 1 embeddingNegative prompt : - embedding:lazyhand, embedding:lazyreal, Bad Quality, worst Quality, low Quality, photorealistic, (photorealistic , embedding:lazyreal - this ones you can use in positive for a more "realistic" feel )
