I’m not a programmer, but I just built my own custom node and you can too.

A user with no prior coding or GitHub experience has documented a successful method for creating custom ComfyUI nodes using ChatGPT. The process began with a highly specific prompt outlining the exact functionality needed: a node that receives images, saves them to a folder, displays them in a scrollable thumbnail gallery with a size slider, includes a clear button, and allows clicking for a larger view. This specificity, combined with defining a minimal "Version 1" feature set and a real-world use case (testing LoRAs), provided ChatGPT with the necessary constraints to generate workable code.

The initial attempt produced a downloadable zip file that ComfyUI recognized, but it contained bugs that prevented images from displaying correctly. The user then entered a cycle of iterative debugging: copying terminal error messages, pasting them back into ChatGPT, and receiving revised code files. After approximately four revision cycles focused on UI behavior and image expansion logic, the node became fully functional. The user then used ChatGPT to guide them through creating a GitHub repository, completing the entire development and sharing pipeline without writing a single line of code manually. The key insight is that breaking down the request into concrete steps—concept, v0.1 features, use case, and iterative error-fixing—unlocks ChatGPT's ability to act as a development partner for non-technical creators.