AI Prompt Pixaroma

Runs a language model you already have, on your own machine, using an instruction you save on the node. Wire in a picture, a video, some audio or text, type an idea, and it hands back text. The instruction is called the formula and it lives on this node, so every copy carries its own. That is what makes these chainable: put one after another and each does a different job, like describe this photo, then rewrite it in another style, then shorten it. With no model chosen it simply passes its text straight through, so you can drop it into a working graph and set it up afterwards without breaking anything. The banner on the node always says which it is about to do. Wire text in and it is joined with your idea, the same way Prompt Pixaroma joins its text input, and you choose which of the two comes first. The idea box understands the same @tags as Prompt Pixaroma, out of the same library: @name drops in a saved phrase, *Category picks a random one from a group, and #name picks a random line. They are swapped for real words before the model is asked anything. Press Tags on the node to open the library. Nothing is sent anywhere. It needs a model in your text_encoders folder, and a vision model if it has to look at a picture. Find it by searching for ai, prompt, llm, caption, or rewrite.