New Version Supir upscaler with Workflow | ComfyUI SUPIR wrapper node UPDATE a better way to do

[Music] don't forget to subscribe and hit the notification Bell so let's go hey everyone here with another MK video and today we're going to be making the workflow based on the super upscale method and the new one so as you know suier was once node and they had made it all a lot better by splitting up into several nodes and there's a whole Suite of nodes now you can download it and it gives us a lot better results closer to like they were showing us in the demos so you would have to go to the link of a GitHub below we will post a link in the description you would just get clone that into your custom nodes folder or you can also go to the manager and comfy UI and install it that way so once you have it installed you would go open up Comey UI so we're going to need a model loader so we start with a right click super and model loader right there and adjust the position year again right click superpier and choose to noiser and adjust the position here so we do the same for a superpier encoder for VA and also a superar conditioner all right we just add a super your sampler no it's a lot of nose but we we will keep it simple okay so we're going to start it's not that hard when going to take the Vee from the first stage hook it up to the Super Vee in the encode and then we will make the super sampler right here go latent to [Music] latent and then from the first stage denoiser we're going to hook up the denoised image to the image in this Superior in code we're going to hook up the Deno latent the light in here and we're going to go to positive from the conditioner to the sampler positive and the negative with negative and then hook up the model from this super model so it's okay this is the first couple nodes that we'll meet so that we can start now or then we're going to do a load image by search then do a resize I do it by search [Music] then we are going to go here and to go to image to the first stage right here and then a right click to load in superar decoder put it beside the sampler and connect them like me and then we're going to do a color match okay then we're going to do or connect image to image Target here for the color match and I'm also going to add another node to save image and connect it to image [Music] ref we can also add an image comparer it's be useful to compare the images before and after so we connect it to the load image output put it beside core match node and connect it I have to [Music] redo so it's good I will connect the and then it's all right we can start doing super model I use super v0f I will resize to 1024 extension [Music] 1024 and here in the color match node I use hm method and in superp your sampler I make same changes on parameters I put 1.5 on scale start and the same on scale end I put resor on five and choose a sampler let's put States on 50 and now Dragon low resolution image for upscaling you have your conditioner which will have the text front so here I have high quality detailed and I add photograph of female and you have your negative I forgot the model so I connected it's all right now we can keep propped let's go and so it will give you the final [Music] image so here's the final image and this will be the comparison node so what you have so we're going to look at the comparison node here is the before and the after as you can see they did a good job we started with a photo that it's not the best quality and see what it does to clean up this image it's a Well promt he kept the same photo with very good confidence and there you go I hope you had a good time with this workflow and that this will be useful to you so before leaving you I would like to share with you another test with another photo on another browser that's it you can clearly see the difference in the eyes here in eyebrows also on mouth i' put the workflow Link in the description so keep in touch and see you [Music] soon don't forget to subscribe and hit the notification Bell so keep in touch

Channel: MunKaw

Views: 1,502

Keywords: stable diffusion animation, ComfyUI, Comfyui animation, Animatediff, comfyui Animatediff, comfyui ip adapter, IP adapter, controlnet, DWPose, Zoe depth maps, comfyui animation, comfyui video, comfyui vid2vid, Animatediff comfyui, Animatediff controlnet, Animatediff IP adapter, lora, Animatediff evolved, controlnet animation, comfyui Animatediff controlnet, Unsampler, KSampler, KSampler Advanced, Noise diffusion, Upscaler, Supir, New Supir, Supir Workflow, Download workflow, AI video

Id: d_uoc88L7K0

Length: 8min 53sec (533 seconds)

Published: Tue Mar 26 2024