Draw Things Prompt

Most of the Draw Things prompt nodes are just text boxes - DrawThingsPositive and DrawThingsNegative exist to carry strings into the sampler. DrawThingsPrompt is the clever one: it's a prompt box plus a textual inversion picker, and it writes the inversion's keyword into your prompt for you.

A textual inversion (a "textual embedding" in Draw Things terms) teaches the text encoder a new word without touching the model - a 10–200 MB file that defines a concept or style, referenced in the prompt with a special tag. In this node you pick one from the server's model browser, and the frontend automatically inserts <keyword> at the start of your prompt text. You never have to remember the trigger tag or mistype it.

Like every node in this pack, it sits in a deprecated repo - the README tells you to migrate to the official draw-things-comfyui extension. The node still functions.

How it works

The backend node is almost empty: its function just returns your prompt string as a STRING. The actual magic lives in the frontend extension, which watches the insert_textual_inversion dropdown. Pick an inversion and it finds the prompt widget and prepends <keyword> - select it again and it removes the tag. The Draw Things server then resolves that tag against its textual inversions during the request. The tooltip on the prompt field is the standard one: "The conditioning describing the attributes you want to include in the image."

The PROMPT output is a plain STRING, so it wires into either the sampler's positive or negative input - which is the neat trick for using a negative embedding (like EasyNegative-style files) in a Draw Things workflow.

The inputs

• insert_textual_inversion - a dropdown of textual inversions from Draw Things' model browser (server must have Model Browser enabled or this stays empty).
• prompt - multiline, defaults to a lovely cat. Selecting an inversion prepends the tag; you can also type <keyword> manually if you know it.

Installing it

ComfyUI Manager (search ComfyUI-DrawThings-gRPC) or:

cd ComfyUI/custom_nodes
git clone https://github.com/Jokimbe/ComfyUI-DrawThings-gRPC

Restart ComfyUI. Server side needs the usual setup - Draw Things app with API Server on, gRPC, TLS enabled, Model Browser enabled, Response Compression disabled (or gRPCServerCLI --no-response-compression --model-browser).

Troubleshooting

Two things bite people. First, the dropdown is empty: the Model Browser is off on the server, or the textual inversion file isn't installed in Draw Things - it won't be pulled from ComfyUI's embeddings folder. Second, the inversion "does nothing": check the tag is actually in the prompt text (the auto-insert only fires when you use the dropdown, not when you reload a workflow), and remember the inversion was trained against a specific model family - an inversion trained for SD 1.5 won't mean much on a Flux-era checkpoint. And if the prompt feels crowded, note that the <keyword> tag is a token like any other; you can leave it in place or delete it freely.