This is my personal basic workflow for QI-21 single-image editing. Since many people have been asking for it, I've decided to share it today, completely free of charge. You need to install the custom nodes I developed to use it. The link is below

https://github.com/no8d/ComfyUI-NO8D-controls

Applicable to all QI-21 editing LoRAs released by me

✨Remix for QI-21✨private version

✨HighQuality for QI-21✨private version

✨MoreReal for QI-21✨private version

Now, let's take a look at the full workflow. Here it is.

The upper part is the main functional area

The lower part is for the puzzle function and can be deleted if not needed.

The main functional area is divided into four parts.

Image Composer (for remix) can combine multiple images and send the result to 3 for computation. When the composer is expanded as shown below, you can adjust image size/angle, perform matting, and add strokes to images. These are essential functions for image compositing, which should be familiar to anyone who has used Photoshop.

It is the image library, used to load the images you want to edit. Like the Image Composer, you can choose either one to connect to 3 for computation.

The core of single-image editing takes the output from 1/2 for computation. You can configure LoRA, weights, prompts, sampler types and random seeds. For example, if you want it to process composite images, select LoRA: Remix, and you can generally leave other parameters at their default values. Leaving the prompt blank works fine without trigger words. Of course, you may also write prompts if you want additional edits to the image; however, prompts in QI-21 carry less weight than the image itself.

If you want to learn more about how it works, you can expand the subgraph, and here is its internal structure, The upper part is for adjusting the size of the imported image. After multiple tests, a long side of 2048 is a suitable dimension.

The lower part is the basic image editing workflow of QI-21. The base model used here is the turbo-int8 version, and the VAE is the texture_fix version, which can eliminate grid textures and has been proven effective in practical tests.

Another thing to note is the ManualSigmas node. The values inside it are officially recommended and are not recommended to be modified.

Nothing else to note. It is basically identical to the official QI-21 image editing workflow. The image below is a composite generated using this workflow.

OK! That's all! I hope you like it!

If you like my sharing and want to get more content and resources related to AI image generation, welcome to join my private community.👇👇👇

NO8D-AI-Workshop

🫡May the AI-power be with you, see you soon!🫡

Disclaimer: My model or workflow is solely for learning and communication purposes and shall not be used for any commercial activities. All responsibilities for the generated content, including its legality, ethical implications, or potential misuse, are borne entirely by the user. By using this model or workflow, you agree to this principle. Please use it responsibly within the scope of local laws.