comfyUI: Create Your Own Node in Seconds Using AI! No Programming Needed!

In my latest video, I'll explore how we can use Large Language Models like GPT-4 or LLAMA to help us code a ComfyUI Node. There are already a lot of useful nodes out there that can be used in many ways, but there's always that one nagging feature you wish existed to make your work so much easier!

This is where this video will blow your mind! Use my simple step-by-step instructions to actually write a program for your very own node. Just 3 simple steps, and you are done! All prompts and steps are explained in the video and in the descriptions below.

Enjoy this next level of learning new things in AI and use it to your advantage. Don't forget to like, subscribe, and hit the bell icon for more tutorials!

Here is the PROMPTS I used to create the comfyUI NODE:
Prompts for llama32k online

1.) FIRST PRompt you ask your LLM of choice:
Do you know comfyUI? I would need help with creating a custom node.

2.) Second Prompt you give:
Ok here is the example/template of a comfyUI Node. Please analyze, I will add later a working example as well.

3.)Third Prompt
Here is a working example of a LoadImage code snippet please analyze

4.) the final Prompt to hopefully get a Program out of your LLM
I need to create a comfyUI Node that loads images similar to the LoadImage node, however it should respect subfolders as an option (default off). The node needs to output the total images found as text so I can know how much images are found in total, it also needs an output for the image name without the file extension and the IMAGE for further processing with another node. Please include all necessary image and other Import directives so that it's a fully executable node code.
Also I want to read all image formats supported. Ensure the Return names are properly declared for the outputs. Also ensure Return Types are properly declared according to return definition.
Make sure that the IMAGE is converted into a proper comfyUI Tensor format before returned as IMAGE for processing with another node. add a return true for the Is_Changed class. Also make sure to use all proper IMPORT directieves for the all image processing you use.

The name of the node should be UltraImageLoader
There should be the following inputs:
ImageID (number)
FolderPath (string)
RespectSubfolders ON/Off

There should be the following outputs:
TotalImages (string) this will output the total images found
ImageFilename Output (string) without file extension this will be used by another node to do further processing
Image (IMAGE) output to connect to other nodes

That's it folks, let me know if you coudl follow and what nodes you created!
#ComfyUI #AI #Programming #Tutorial #GPT4 #LLAMA

I was able to successfully create my own custom node following your methodology. Not without a little back n forth bug fixing with the LLM, but I eventually pushed something to git in one day. thank you.

[DanielSchweinert]

Just found your video! Thanks! AI is a Ferrari but most are using it like a Rollator. LOL. I had this in mind for some time now but I want AI to create a complete workflow. I want it to understand all available nodes and create a workflow based on my inputs (Text, Images, etc.)