In this tutorial, we’ll delve into using various LoRAs (Low-Rank Adaptations) to bring the artistic flair of Midjourney to images generated by Stable Diffusion. By combining these LoRAs, you can achieve a variety of artistic effects. We’ll build workflows in ComfyUI to combine these LoRAs, but you can also implement them in A1111 or Fooocus….
The recent release of Stable Diffusion 3 (SD3) has sparked diverse reactions within the community. While some have praised its advanced text generation capabilities, others have been dismayed by its shortcomings, particularly in rendering human bodies accurately. To thoroughly assess SD3’s performance, I spent considerable time testing and generating numerous images, comparing it to other…
Hello everyone! Today, I’m excited to introduce a newly built workflow designed to retouch faces using ComfyUI. This workflow showcases the remarkable contrast between before and after retouching: not only does it allow you to draw eyeliner and eyeshadow and apply lipstick, but it also smooths the skin while maintaining a realistic texture. Importantly, the…
Today, we will compare three AI face-swapping technologies: PuLID, InstantID, and IP-Adapter’s FaceID-V2, using a ComfyUI workflow. These technologies are built on a face analysis system called InsightFace, which is a deep face analysis library designed for face recognition, face detection, and face alignment. Note that InsightFace requires a commercial license. IP-Adapter FaceID was the…
Hello everyone! Today, I’m excited to show you how to use Stable Diffusion to implement a generative fill feature similar to Photoshop’s. Photoshop’s generative fill is quite powerful. For instance, it uses AI to automatically extend new content on the left and right sides of a landscape photo. Just select the area to be filled,…
Today, we’re diving into the innovative IP-Adapter V2 and ComfyUI integration, focusing on effortlessly swapping outfits in portraits. This tutorial simplifies the entire process, requiring just two images: one for the outfit and one featuring a person. By utilizing ComfyUI’s node operations, not only is the outfit swapped, but any minor discrepancies are resolved with…
A common hurdle encountered with ComfyUI’s InstantID for face swapping lies in its tendency to maintain the composition of the original reference image, irrespective of discrepancies with the user’s input. For instance, if a user uploads a headshot while requesting a full body depiction, the output frustratingly remains a mere headshot, ignoring the intended composition….
In a previous piece, I explored how to use the Adetailer extension and your own trained LoRA to swap faces for a person in images in A1111. Building on that knowledge, this article will delve into achieving a similar task in ComfyUI, enhancing the realism and fidelity of face swaps. For those who love diving…
In a previous post, Training Face LoRA, I detailed the use of XYZplot in A1111 to examine a batch of trained LoRA. While XYZplot in A1111 proved quite resourceful, its processing speed in handling a large number of images was notably slower than ComfyUI, often exceeding half an hour. This article aims to guide you…
ComfyUI’s LoRA workflow is well-known among users. The developers offer an array of built-in workflows that utilize default node functionality, demonstrating how to effectively implement LoRA. Applying a single LoRA can be quite straightforward. Simply integrate a LoRALoader node into your existing workflow. However, if you need to incorporate multiple LoRAs, you would typically add…