I'm very happy to release its first update, which is not just a minor fix. Although I did fix quite a few bugs, that's not the main point. This time, I've brought a feature that I've always loved. To be honest, at first, I thought it would be very difficult to implement, but when I completed it, what came to mind was "Is that it? I did it?" Haha... AI is truly amazing! Alright, no more suspense. Now, let me introduce the star of this update.

NO8D-Prompt-plus

As you can see, it is an intelligent prompt node. I know you might think, ! Another LLM." In fact, it is much lighter. It is an API node, and you can connect to any API you like to generate prompts. Moreover, it is a versatile player. You can connect both text and images, and it can generate prompts for both.

So how does it work? First, you need to have an API key, which can be for any model. If you just want to implement prompt expansion, then a large language model is sufficient, but if you want the ability to reverse-engineer prompts from images, it's best to choose a large model with vision capabilities. Then open the settings of ComfyUI, and you'll see this page.

Open the API manager, fill in your key, and if the monitoring passes, you can start having fun. It's that simple.

You may have noticed that there is also a Rule Manager button, which is used to set prompt rules, which determine the quality of the prompts output by the large model you use. I know that writing rules is always a hassle, so I have built-in excellent prompt rules. Usually, you don't need to do anything and can just keep the default settings. Of course, I have also prepared JSON prompt rules for supporters who like ideogram4. These rules are refined from the official ideogram4 prompts and are very useful.

OK, that's all the settings you need to do. Now your NO8D-Prompt-view is fully charged, LOL.

So far, it's just an ordinary prompt generation node. Maybe you can also find good nodes to implement it, but for the following features, I've never seen anyone do it this way.

It's this little switch called auto. Although prompt generators are convenient, often we don't want to avoid writing prompts; instead, we hope to have a basic template and then enrich it with our own ideas. Unfortunately, most generators I've used don't allow this, so I created this node. When you turn off auto, you'll find that the generated prompt will only appear in the window below and won't continue with the subsequent process.

You can modify the content of the prompt freely in the window until you are satisfied, and then click Send below for the prompt to enter the subsequent process.

In NO8D-Prompt-plus, you can choose Natural language, JSON structure, and Style

If you already have some ideas, you can add new rules in extra rules, such as red hair, pink round glasses...

The results of Natural language are very good, and it is applicable to both Klein and the latest Krea2.

OK, this is the biggest update this time. Of course, there are also some minor fixes, but in comparison, those are not worth mentioning. I really like this node and hope you will too. For this reason, I specifically built a new workflow. This is also a really cool workflow, but I still need to do some debugging. If everything goes smoothly, I will release the workflow tomorrow.

Update on GitHub