Upscaling workflow comfyui. RunComfy: Premier cloud-based Comfyui for stable diffusion. I made this using the following workflow with two images as a starting point from the ComfyUI IPAdapter node repository. I’m very happy to share my LoRA Caption workflow for ComfyUI that will let you run a batch of images through two different captioning methods using Florence 2 model or using Clip Interrogator. Download it and place it in your input folder. ex: upscaling, color restoration, generating images with 2 characters, etc. You can find the . Workflow Considerations: Automatic 1111 follows a destructive workflow, which means changes are final unless the entire process is restarted. Aug 26, 2024 路 FLUX is a new image generation model developed by . Here's how you set up the workflow; Link the image and model in ComfyUI. 1_0) Video2Video Upscaler It's a Video to Video Upscaling workflow ideal for 360p to 720p videos, which are under 1 min of duration. This option is offered as it's very inexpensive computationally and it can be useful if you don't need the final image to be 100% of the initial generation. inputs. I wanted a very simple but efficient & flexible workflow. x and SDXL; Asynchronous Queue system; Many optimizations: Only re-executes the parts of the workflow that changes between executions. Relaunch ComfyUI to test installation. Launch ComfyUI again to verify all nodes are now available and you can select your checkpoint(s) Usage Instructions. ThinkDiffusion Merge_2 AuraSR v1 (model) is ultra sensitive to ANY kind of image compression and when given such image the output will probably be terrible. Goto Install Models. I am not saying this is bad, but personally i prefer upscaling in latent. You signed in with another tab or window. 55+). When use Ultimate SD upscale? Performance and Speed: In terms of performance, ComfyUI has shown speed than Automatic 1111 in speed evaluations leading to processing times, for different image resolutions. Thanks for all your comments. In the most simple form, a ComfyUI upscale workflow would just use an Load Upscale Model node, and might look something like; My actual workflow file is a little messed up at the moment, I don't like sharing workflow files that people can't understand; my process is a bit particular to my needs and the whole power of ComfyUI is for you to create something that fits your needs. x and SDXL latents. Apr 24, 2024 路 ComfyUI workflows for upscaling. NO PROMPT NEEDED - It just works!! How to use this workflow Simply drag and drop the image you want to upscale into the BASIC SETTINGS group box, select your favourite SD 1. Short version uses a special node from Impact pack. ThinkDiffusion Merge_2_Images. Changelog: Converted the scheduler inputs back to widget. Then the upscaling workflow can read the seed from the filename and reconstruct the prompt and all other settings from that. I was just using Sytan’s workflow with a few changes to some of the settings, and I replaced the last part of his workflow with a 2-steps upscale using the refiner model via Ultimate SD upscale like you mentioned. You can also easily upload & share your own ComfyUI workflows, so that others can build on top of them! :) Why I built this: I just started learning ComfyUI, and really like how it saves the workflow info within each image it generates. This is ur workflow copied, but with a second sampler and a NNLatentUpscale (1. Actually, I included a krea file in the zip on comfy workflow as an easter egg heheh 馃榿 This one was for zero-change in the image. 1 Dev Flux. pkuliyi2015 & Kahsolt's TIled Noise Inversion for better upscaling. Most of the workflows focus on upscaling with Ultimate SD Upscale or just plainly upscaling with model. 2) Batch Upscaling Workflow: Only use this if you intend to upscale many images at once. You can Upscale Videos 2x,4x or even 8x times. You can find the Flux Schnell diffusion model weights here this file should go in your: ComfyUI/models/unet/ folder. Upscaling latents with a model is not as sharp as upscaling pixels but a lot better than using a standard upscale algorithm. SUPIR upscaling wrapper for ComfyUI. Versions. 3 days ago 路 --v2. ComfyUI breaks down a workflow into rearrangeable elements so you can easily make your own. An All-in-One FluxDev workflow in ComfyUI that combines various techniques for generating images with the FluxDev model, including img-to-img and text-to-img. com/comfyanonymous That said, I prefer Ultimate SD Upscale: ssitu/ComfyUI_UltimateSDUpscale: ComfyUI nodes for the Ultimate Stable Diffusion Upscale script by Coyote-A. In this example we will be using this image. At the end, when you open and zoom on your image, it’s quite noticeable that your upscale generated visible seams between the upscales tiles. The FLUX models are preloaded on RunComfy, named flux/flux-schnell and flux/flux-dev. Mines a dual model workflow, but its highly adaptive to whatever project im working on, includes img2img and prediffusion, includes multi-step upscaling - face fixes at multiple steps, tiled upscaling and all that. What is ComfyUI? Upscaling with ComfyUI. ai's upscaling services. 1 Pro Flux. Learn how to use ComfyUI to create stunning images and animations with Stable Diffusion. I have a custom image resizer that ensures the input image matches the output dimensions. ComfyUI provides a powerful yet intuitive way to harness Stable Diffusion through a flowchart interface. Click to see full-size comfyUI upscaling workflow ( download the picture of this workflow, drag&drop it on your comfyui, and install missing nodes if needed ) To do this we will use the Manager. This workflow can use LoRAs, ControlNets, enabling negative prompting with Ksampler, dynamic thresholding, inpainting, and more. , ImageUpscaleWithModel -> ImageScale -> UltimateSDUpscaleNoUpscale). Here is an example of how the esrgan upscaler can be used for the upscaling step. Hopefully there will be new solutions soon. It's an extension of the basic Refiner workflow. Note that in ComfyUI txt2img and img2img are the same node. 25x) before it, latent upscales do tend to copy things, like the cat here, but regardless more detail. What is the recommended sequence for these tasks? Hand Fix > Face Fix > Upscaling Face Fix > Hand Fix > Upscaling Upscaling > Hand Fix > Face Fix Or something else? Welcome to the unofficial ComfyUI subreddit. Pressing the letter or number associated with each Bookmark node will take you to the corresponding section of the workflow. Some workflows alternatively require you to git clone the repository to your ComfyUI/custom_nodes folder, and restart ComfyUI. Workflow is available here, you can download. Nothing fancy. been using this for a while. Yes, i have magnific/krea workflow tutorial video (complete with the workflow, the files and the prompts of course!) in the work for that specific use case of inventing detail. Category: image/upscaling; Output node: False; This node is designed for upscaling images using a specified upscale model. Aug 26, 2024 路 What is the ComfyUI FLUX Img2Img? The ComfyUI FLUX Img2Img workflow allows you to transform existing images using textual prompts. It’s insane how good it is as you don’t lose any details from the image. I share many results and many ask to share. Users can drag and drop nodes to design advanced AI art pipelines, and also take advantage of libraries of existing workflows. You can also look into the custom node, "Ultimate SD Upscaler", and youtube tutorial for it. Put them in the models/upscale_models folder then use the UpscaleModelLoader node to load them and the ImageUpscaleWithModel node to use them. The Upscale Image node can be used to resize pixel images. . Apr 1, 2024 路 This tutorial focuses on SUPIR for ComfyUI, some core concepts and upscaling techniques used with SUPIR. Aug 25, 2024 路 ComfyUI LoRA Caption Workflow. You can then load or drag the following image in ComfyUI to get the workflow: Flux Schnell. Feb 7, 2024 路 Why Use ComfyUI for SDXL. Please keep posted images SFW. Mar 22, 2024 路 As you can see, in the interface we have the following: Upscaler: This can be in the latent space or as an upscaling model; Upscale By: Basically, how much we want to enlarge the image; Hires Both work on the same principle of Upscaling using a Upscaling Model. pth Click on the Install button. image. Aug 16, 2024 路 Open ComfyUI Manager. Pinto: About SUPIR (Scaling-UP Image Restoration), a groundbreaking image restoration method that harnesses generative prior and the power of model scaling up. Created by: Gambz: If the upscaled image looks "noisy" try changing the upscaling model to something like 4x-UltraSharp, it will lower overall micro details but will make a cleaner image. This workflow uses lightning for latent creation and refiner for upscaling. Feb 28, 2024 路 This guide caters to those new to the ecosystem, simplifying the learning curve for text-to-image, image-to-image, SDXL workflows, inpainting, LoRA usage, ComfyUI Manager for custom node management, and the all-important Impact Pack, which is a compendium of pivotal nodes augmenting ComfyUI’s utility. I hope this clarifies things a bit! Aug 29, 2024 路 Inpaint Examples. Load the 4x UltraSharp upscaling model as your Feb 1, 2024 路 The UltraUpscale is the best ComfyUI upscaling workflow I’ve ever used and it can upscale your images to over 12K. If the final image can be rather different from Apr 16, 2024 路 I saw article about upscaling in ComfyUi and though, i have not really seen much info about latent upscaling. You can then load or drag the following image in ComfyUI to get the workflow: I'm a beginner working with ComfyUI and I have a GPU with 6GB of VRAM. If you want to specify an exact width and height, use the "No Upscale" version of the node and perform the upscaling separately (e. Basic Refiner Workflow: Upscaling In Comfy UI Done Easy!In this video we show you how to upscale your images in Comfy UI in two ways. Here is an example of how to use upscale models like ESRGAN. Simply drag or load a workflow image into ComfyUI! Simply drag or load a workflow image into ComfyUI! See the "troubleshooting" section if your local install is giving errors :) We would like to show you a description here but the site won’t allow us. With AuraFlow, users can experience the cutting-edge capabilities of GAN-based upscaling, delivering high-quality images at Share, run, and discover workflows that are meant for a specific task. 34 per hour) and discovered this workflow by @plasm0 that runs locally and support upscaling as well. I often reduce the size of the video and the frames per second to speed up the process. The node itself is the same, but I no longer use the Eye Detection Models. Feature/Version Flux. ComfyUI . Contribute to kijai/ComfyUI-SUPIR development by creating an account on GitHub. To speed up your navigation, a number of bright yellow Bookmark nodes have been placed in strategic locations. My SUPIR upscaling comfy workflow and others Just wanted to share a relatively manageable setup for SUPIR I've been using to restore old family photos. I'm revising the workflow below to include a non-latent option. Like many XL users out there, I’m also new to ComfyUI and very much just a beginner in this regard. In this workflow we are using the NNLatentUpscaler extension that is compatible both with SD1. This workflow empowers users to fine-tune yellow-highlighted nodes, providing flexibility and control over the image generation process. Since ESRGAN Created by: Mad4BBQ: What this workflow does Extremely EASY to use upscaler/detailer that uses lighting fast LCM and produces highly detailed results that remain faithful to the original image. 15 KB. Flux Schnell is a distilled 4 step model. g. TLDR In this tutorial, Seth introduces ComfyUI's Flux workflow, a powerful tool for AI image generation that simplifies the process of upscaling images up to 5. Reply reply Hello! I am hoping to find find a ComfyUI workflow that allows me to use Tiled Diffusion + Controlnet Tile for upscaling images~ can anyone point me toward a comfy workflow that does a good job of this? May 23, 2023 路 Here is a non cherry picked example with a simple “hiresfix” or 2 pass type workflow with a 2x upscale with nearest-exact, bislerp and bilinear in that order (if you want to see the exact workflows you can open these images in ComfyUI: You’ll notice that the image in the middle seems more alive. So I ran up my local instance on my computer of ComfyUI with Flux and started to see some incredible results. As a pivotal catalyst within SUPIR, model scaling dramatically enhances Thanks for sharing your workflow! I've been trying to upscale SVD results with little success. com/drive/folders/1-g8GjBdoJYlOs This extension enables large image drawing & upscaling with limited VRAM via the following techniques: Two SOTA diffusion tiling algorithms: Mixture of Diffusers and MultiDiffusion; pkuliyi2015 & Kahsolt's Tiled VAE algorithm. I tried to keep the noodles under control and organized so that extending the workflow isn't a pain. This is under construction Upscale Image node. json file. I think I have a reasonable workflow, that allows you to test your prompts and settings and then "flip a switch", put in the image numbers you want to upscale and rerun the workflow. It is highly recommended that you feed it images straight out of SD (prior to any saving) - unlike the example above - which shows some of the common artifacts introduced on compressed images. By adding it into your current work flow or by Welcome to the unofficial ComfyUI subreddit. 5 Lora with SDXL, Upscaling Future tutorials planned: Prompting practices, post processing images, batch trickery, networking comfyUI in your home network, Masking and clipseg awesomeness, many more. To upscale images using AI see the Upscale Image Using Model node. com/watch?v=GV_syPyGSDYComfyUIhttps://github. Oct 21, 2023 路 From a little research I made, I found out about three different methods of upscaling: Latent, Non-latent, and ControlNet-assisted Latent. By combining the visual elements of a reference image with the creative instructions provided in the prompt, the FLUX Img2Img workflow creates stunning results. It handles the upscaling process by adjusting the image to the appropriate device, managing memory efficiently, and applying the upscale model in a tiled manner to accommodate for potential out-of-memory errors. More to come. I spent some time fine-tuning it and really like it. Contribute to greenzorro/comfyui-workflow-upscaler development by creating an account on GitHub. json file in the attachments. google. Use the Models List below to install each of the missing models. That means that the second pass after the upscale has to be done at a very high denoise (0. It can still be done, though. Some commonly used blocks are Loading a Checkpoint Model, entering a prompt, specifying a sampler, etc. Support for SD 1. Here is the txt2img part: As a result, I get this non-upscaled 512x1024 image: Jul 12, 2024 路 Here's a step-by-step guide on setting up a ComfyUI workflow that upscale images on your local machine. 5 days ago 路 31 Aug 2024 76:17. x, SD2. Created by: Reed Simpson: SVD Upscaling and Face Restoration (Good for Medium Wide Shots). I have gotten more Aug 5, 2024 路 I was working on exploring and putting together my guide on running Flux on Runpod ($0. Apr 27, 2024 路 The workflow has Upscale resolution from 1024 x 1024 and metadata compatible with the Civitai website (upload) after saving the image. Preparation. Video tutorial on how to use ComfyUI, a powerful and modular Stable Diffusion GUI and backend, is here. Welcome to the unofficial ComfyUI subreddit. Then I created two more sets of nodes, from Load Images to the IPAdapters, and adjusted the masks so that they would be part of a specific section in the whole image. To load a workflow, simply click the Load button on the right sidebar, and select the workflow . Upscale Image¶. Mar 8, 2024 路 In this easy ComfyUI Tutorial, you'll learn step-by-step how to upscale in ComfyUI. 5 model, and then enter your Starting workflow. com) See also: ComfyUI - Ultimate SD Upscaler Tutorial. ComfyUI Examples Installing ComfyUI Features Nodes/graph/flowchart interface to experiment and create complex Stable Diffusion workflows without needing to code anything. Any issues or questions, I will be more than happy to attempt to help when I am free to do so 馃檪 This is a Demo of the Upscaling workflow using ai upscaling models inside comfyui1) Download workflow: https://drive. 0 Updates - Revised the presentation of the Image Generation Workflow and Added a Batch Upscale Workflow Process--Workflow (Download): 1) Text-To-Image Generation Workflow: Use this for your primary image generation. Here’s a simple workflow in ComfyUI to do this with basic latent upscaling: Non latent Upscaling. I added the node clip skip -2 (as recommended by the model), remembering that in ComfyUI the value -2 is equal to 2 (positive) in other generators (Civitai, Tensorart, etc). So I'm happy to announce today: my tutorial and workflow are available. (Ersgan, Sformer etc), No ksamplers are used in this workflow. This is under construction Nov 25, 2023 路 Upscaling ComfyUI workflow. Click on “Manager” > Install Models > Search > “ESRGAN x4” Choose the one whose filename is: ESRGAN_4x. The initial image KSampler was changed to the KSampler from the Inspire Pack to support the newer samplers/schedulers. Just use an upscale node. youtube. Upcoming tutorial - SDXL Lora + using 1. Ending Workflow. Close ComfyUI and kill the terminal process running it. Jan 8, 2024 路 This involves creating a workflow in ComfyUI, where you link the image to the model and load a model. Aug 9, 2024 路 Inpaint (using Model) (INPAINT_InpaintWithModel): Perform image inpainting using pre-trained model for seamless results, restoration, and object removal with optional upscaling. json file as well as a png that you can simply drop into your ComfyUI workspace to load everything. I needed a workflow to upscale and interpolate the frames to improve the quality of the video. Download & drop any image from the website into ComfyUI, and ComfyUI will load that image's entire workflow. But let me know if you need help replicating some of the concepts in my process. Created by: John Qiao: Model Stable Diffusion 3 Medium is a Multimodal Diffusion Transformer (MMDiT) text-to-image model that features greatly improved performance in image quality, typography, complex prompt understanding, and resource-efficiency. 1 Schnell; Overview: Cutting-edge performance in image generation with top-notch prompt following, visual quality, image detail, and output diversity. Hello, It’s always nice to have new tips being shared and thanks for that but from what I see I think you still need to work on your workflow. Created by: C. Download 10 cool workflows for upscaling, merging, inpainting, animating and more. Learn how to use the Ultimate SD Upscaler in ComfyUI, a powerful tool to enhance any image from stable diffusion, midjourney, or photo with scottdetweiler. You can construct an image generation workflow by chaining different blocks (called nodes) together. Both these are paired with WD14 Tagger node which generates some additional tags/keywords for the caption. This workflow is a partial adaptation to ComfyUI, therefore the results might be different from those that you can expect on Runtime44 - Mage Preamble Both this workflow, and Mage , aims to generate the highest quality image, whilst remaining faithful to the original image. Unlike other Stable Diffusion tools that have basic text fields where you enter values and information for generating an image, a node-based interface is different in the sense that you’d have to create nodes to build a workflow to generate images. I liked the ability in MJ, to choose an image from the batch and upscale just that image. Jan 5, 2024 路 I have been experimenting with AI videos lately. Input types Thanks. Current Feature: I'd never ignore a post I saw asking for help :D So when I refer to denoising it, I am referring to the fact that the lower resolution faces caused by using Reactor need to be denoised if you want to add more resolution, this requires passing through a sampler with denoising, the higher the denoising is on this sampler the more it will change and mess the face back up again. Txt2Img is achieved by passing an empty image to the sampler node with maximum denoise. Here's a video to get you started if you have never used ComfyUI before 馃憞https://www. This FLUX NF4 model is created by lllyasviel, please visit for more information. This image has had part of it erased to alpha with gimp, the alpha channel is what we will be using as a mask for the inpainting. I use this youtube video workflow, and he uses a basic one. Created by: OpenArt: The latent can be upscaled directly but the loss in quality is important. Created by: CongDC: Workflow AuraFlow We are excited to introduce our newly created ComfyUI workflow utilizing the AuraFlow model. Image restoration, enhancement, and some mixed techn Dec 19, 2023 路 Here's a list of example workflows in the official ComfyUI repo. Jan 8, 2024 路 The upscale workflow provided by ComfyUI offers a compelling, cost-free alternative to Magnific. We would like to show you a description here but the site won’t allow us. Feb 26, 2024 路 Explore the newest features, models, and node updates in ComfyUI and how they can be applied to your digital creations. You switched accounts on another tab or window. In this article, I am going to use the ComfyUI workflow I made. Feb 24, 2024 路 ComfyUI is a node-based interface to use Stable Diffusion which was created by comfyanonymous in 2023. The issue with this workflow and the ones I've tried is that you lose consistency when upscaling. In ComfyUI the image IS the workflow. Leveraging multi-modal techniques and advanced generative prior, SUPIR marks a significant advance in intelligent and realistic image restoration. It becomes more challenging when you generate images with randomized prompts, like I usually do. Empowers AI Art creation with high-speed GPUs & efficient workflows, no tech setup needed. Jun 3, 2024 路 Image of the workflow (download workflow below) Infinite Zoom Workflow (171 downloads ) The custom node bundle includes a lot of other fun nodes which are useful for generating some cool results that can be explored further in future posts but this post was only focused on infinite zoom loop that you can generate using a single image in ComfyUI. Upscale Model Examples. Less is more approach. Since SDXL requires you to use both a base and a refiner model, you’ll have to switch models during the image generation process. Masked latents are now handled correctly; however, iterative mixing is not a good fit for using the VAEEncodeForInpaint node because it erases the masked part, leaving nothing for the iterative mixer to blend with. Please share your tips, tricks, and workflows for using this software to create your AI art. Today we will use ComfyUI to upscale stable diffusion images to any resolution we want, and even add details along the way using an iterative workflow! This Aug 26, 2024 路 The ComfyUI FLUX Latent Upscaler workflow also includes various mathematical operations using the SimpleMath+, SimpleMathFloat+, SimpleMathInt+, and SimpleMathPercent+ nodes to calculate dimensions, ratios, and other parameters for the upscaling process. What it's great for: If you want to upscale your images with ComfyUI then look no further! The above image shows upscaling by 2 times to enhance the quality of your image. ComfyUI provides some of the most flexible upscaling options, with literally hundreds of workflows and nodes dedicated to image upscaling. (github. I've never had good luck with latent upscaling in the past, which is "Upscale Latent By" and then re-sampling. Latent Upscaling – When not Upscaling during testing, make sure to bypass every upscaling group and the very latent upscale video combine node. Although not the same ComfyUI delivers notch upscaled images allowing users to customize settings according to their preferences making it a compelling choice, for image enlargement projects. ThinkDiffusion_Upscaling. Older versions are not better or worse, but they are long and expanded. After borrowing many ideas, and learning ComfyUI. The pixel images to be upscaled. -- Below is my XL Turbo workflow, which includes a lot of toggles and focuses on latent upscaling. x, SDXL, LoRA, and upscaling makes ComfyUI flexible. Reload to refresh your session. Created by: akihungac: SUPIR upscaler is the highest quality, but it is very slow, requires high hardware/configuration and has a few exceptions that cause the image to be blurry than the original (like the gem image). Jul 6, 2024 路 What is ComfyUI? ComfyUI is a node-based GUI for Stable Diffusion. x, 2. Fully supports SD1. Merging 2 Images together. These will have to be set manually now. You signed out in another tab or window. The video demonstrates how to integrate a large language model (LLM) for creative image results without adapters or control nets. inputs¶ image. I created this workflow to do just that. AP Workflow is a large ComfyUI workflow and moving across its functions can be time-consuming. For more technical details, please refer to the Research paper . Here is details on the workflow I created: This is an img2img method where I use Blip Model Loader from WAS to set the positive caption. The iterative mixing sampler code has been extensively reworked. json. Be prepared to download a lot of Nodes via the ComfyUI manager. I need a good workflow for fixing hands, fixing faces, and upscaling images. Eye Detailer is now Detailer. You get to know different ComfyUI Upscaler, get exclusive access to my Co Created by: Michael Hagge: Updated on Jul 9 2024 . Super simple yet powerful upscaler node that delivers a detail added upscale to any image! Feb 13, 2024 路 It's a workflow to upscale image several times, gradually changing scale and parameters. The workflow uses SAF (Self-Attention-Guidance) and is based on Ultimate SD Upscale. Please note: this model is released under the Stability Non-Commercial Research Aug 16, 2024 路 Open ComfyUI Manager. ↑ Node setup 1: Generates image and then upscales it with USDU (Save portrait to your PC and then drag and drop it into you ComfyUI interface and replace prompt with your's, press "Queue Prompt") For a dozen days, I've been working on a simple but efficient workflow for upscale. 4x using consumer-grade hardware. About FLUX. A 2nd ControlNet pass during Latent Upscaling – Best practice is to match the same ControlNets you used in first pass with the same strength & weight Apr 26, 2024 路 Workflow. In this tutorial we're using a 4x UltraSharp upscaling model known for its ability to significantly improve image quality. SDXL works with other Stable Diffusion interfaces such as Automatic1111 but the workflow for it isn’t as straightforward. Mar 25, 2024 路 The zip file includes both a workflow . The best way is to use a single seed for the entire workflow and use that seed as the filename. ggnosayobqtxpiknktohrjqrcatniqzvdjijervdrrnjxmf