Krea2 Style Studio

The long version: This workflow can be manually installed by placing the supplied .js file into '/Comfy-Desktop/ComfyUI-Installs/ComfyUI/ComfyUI/custom_nodes/rgthree-comfy/web/comfyui' but there is also a supplied python installer and setup instructions written by ChatGPT.

The workflow takes the guesswork out of creating complex and interesting images by automating the photo style, the character's pose, and lighting. There is logic behind the scenes, in subgraphs, so that you cannot produce an incoherent image unless you specifically prompt it to do something unrealistic. So the only thing you really need to write in the main subject prompt are the specifics of your character and what must be shown. Everything else is automated by the logic in the style controls .js file provided.

So if you didn't write anything at all in the main prompt box, the logic would always produce a coherent image. But if you do write a prompt in the main subject box, anything you include which would lead to an incoherent image is replaced, or the context is changed to match Krea2's expected syntax.

The only set of nodes which will over-write your prompt are contained in the Pose It frame. Pose it can be set to random, or you can choose from hundreds of different poses manually. If you choose random and you queue, say, six generations, each generation will choose a random pose from several different categories: standing, sitting, laying down, reaching up... and so on. If you use this in combination with a pose style in your main subject prompt the logic will attempt to combine them both in the final prompt - which you can preview before you send it to the queue.

You can also auto-prompt face types, glasses, hairstyles, skin type and ethnicity using the Face It frame. You can also copy the Face It node prompts to your main subject prompt to create combinations of different characteristics with a single button. Of course if you find something you like, you can also copy and paste it from the prompt preview and modify it accordingly.

Each group of nodes can be bypassed to simply use the main middle section of the workflow to prompt and generate Krea2 as normal. But with all nodes enabled you will always get a preview of the changes made to your prompt before you send it to the queue if you click the Refresh Prompt Preview button first. This will parse the prompt through the logic subgraph so you can make any changes before sending it to the image generator subgraph.

The short version: Download the required files and follow the instructions. Once you load the workflow into ComfyUI, click Run without writing any prompt, and the rest will become clear from there.

HAVE FUN!