Enhance image details to the pinnacle with [Comfyui workflow]

tell me why there's something [Music] different tell me why it's so broken fall fall I'll take my hello everyone today I will guide you on the method to enhance a photo or video using AI with com fui if you are a person create images or videos sometimes you spend a lot of time creating however the final results are not always satisfactory there are old photos very blurry the color and details are not appealing however not everyone has advanced skills in post-processing with Photoshop however the time it takes to do that is not quick either and the results are not always satisfying so now with AI this task is incredibly easy with just a few nodes and confy hi will help you do this quickly and beautifully first prepare an image for enhancement then call out these nodes I use the efficiency node as the main node I use the efficiency node to streamline everything without using too many input nodes this makes everything less messy with fewer input nodes here I will use the custom prompt set node and the prompt read from the image tagger node I will concatenate them using the concat string Noe the advantage of this is that you don't have to describe the prompt in detail it will do that for you I use this node almost throughout my entire workflow it also helps the generation [Music] process be more reasonable without the necessary details without unnecessary details being added then I will pull in these noes if you don't know why there is no path in the node you call [Music] out right click on it and it will revert to the original for you to visualize I will provide the paths below for you to download right click and select the path you need for input it will automatically switch to input for you to connect to another note I will pull in these noes to have the initial prompt for the Lura part I use turbo and LCM I will provide the pads below for you to download it I use all Lorax plus checkpoint plus other stal models all are stal with turbo it will will reduce the generation time it makes the image creation process faster with fewer steps than previous methods and here are the two Lorax ey used to enhance the input details it greatly supports the process of enhancing details you can use your own [Music] checkpoint since The Lorax turbo can make it faster I also use the IP adapter [Music] method with IP adapter Plus for stal models the first image is upscaled to enhance the initial [Music] details using the ultra sharp upscale [Music] model you can directly download it here Miami after a scaling X4 I will downscale the image again to avoid overloading Ram during [Music] generation you can find a downscale node here [Music] this set of nodes has many other functions such as [Music] display this downscale node has the ability to limit the width or height while maintaining the frame ratio then decode through the vine encoder after that I will upscale [Music] again you can follow my other [Music] settings remember to download and use the correct stal [Music] model [Music] you have to turn this button on or it will report an [Music] error adjust the remaining parameters [Music] accordingly in Sample I will use two more rounds of upscale models they will be used used directly in the generation process I use ultra sharp to enhance bold features and SX to avoid making it to Sharp you can also download sax here if you don't have [Music] it remember its location with step and CFG just as much as needed I tested this Lorax turbo beforehand so these are the parameters I will make a video to guide you on how to find out which parameters work best for your model within a certain range I will make a video to guide you on how to find out which parameters work best for your model within a certain range follow the instructions for the remaining [Music] indices for the upscale part I only limited to 1.5 and this is the [Music] output I will downscale once more then continue with the upscale after that add one last note which is the enhanced detail and here's the result I'll drag it to this side for easy comparison you can see that there is a very big difference the image after being enhanced has been added with a lot of details I will zoom in to add one more detail we will see the difference with the original image I will speed up this part so as not to waste your time this is the result it looks a little more expensive but the skin hair and eyebrows look quite good this depends on the degree you want more or less the result looks quite good I will upload this workflow to patreon if you don't have much time to create it from the beginning it will cost a little money for this after you download the workflow and encounter this error phenomenon it means that you are missing some notes or models need I will guide you to follow this method so that they can work click on manager and select the install missing custom nodes here it will'll show that your nodes are missing you just need to plug them in after you have fully installed update the whole thing to the latest version there is one thing to note for you is that confy has two main repo versions one is the CU 118 version and the cu1 121 version please use the cu1 121 version if not the other nodes will not work not working or making mistakes

Channel: Archilives | Ai | Ue5

Views: 2,793

Keywords: comfyui, LCM lora, IP adapter, stable diffusion, workflow, stable diffusion tutorial, upscale, ai generated art, image enhancer, free, turbo, SDXL, ip adapter, comfyui workflow, comfyui tutorial, comfyui nodes

Id: joemK6nIrp8

Length: 9min 27sec (567 seconds)

Published: Sun Dec 17 2023