Draw Things Positive Prompt

This is the smallest useful node in the pack: one multiline text box, one string output. DrawThingsPositive is how your prompt gets into a Draw Things generation. Type what you want, wire the POSITIVE output into the sampler's positive input, and that's the whole job.

It exists for a structural reason. ComfyUI's normal sampler takes conditioning objects produced by a CLIP text encoder; the Draw Things sampler takes plain strings, because the server does the text encoding. Something has to be the box you type into, and this node is that box. If you've used the pack's DrawThingsPrompt node instead, that one exists for inserting textual-inversion keywords - this one is just the plain positive prompt.

Same honest note as every node here: the pack is deprecated in favor of the official draw-things-comfyui extension. The node still functions.

How it works

Under the hood there's almost nothing to it: the node's function returns your text as a STRING, and the sampler passes it through as the prompt field of the gRPC request. The Draw Things server runs its own CLIP / text-encoder path - whatever model you picked in the sampler decides how the text gets embedded. That means prompt syntax follows what Draw Things supports for that model, not what ComfyUI's CLIP loader happens to understand.

The inputs and output

One input worth any attention:

• positive - multiline, defaults to a lovely cat, tooltip: "The conditioning describing the attributes you want to include in the image." Everything you type here gets sent verbatim.

The single output, POSITIVE (type STRING), plugs into the sampler's positive input. That's the entire wiring diagram.

Installing it

With ComfyUI Manager, search ComfyUI-DrawThings-gRPC, or:

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

Restart ComfyUI. You also need the Draw Things server reachable (API Server on, gRPC, TLS on, Model Browser on, Response Compression off, or the gRPCServerCLI equivalent) - otherwise the sampler you're feeding will fail to connect even though this node is innocent.

Troubleshooting

The usual beginner confusion is the opposite direction: you typed a prompt, it appears in the box, and the image ignores it. That's almost never this node. Check the sampler's CFG first - on guidance-distilled models CFG 1 is the real default and prompt adherence is tuned differently; on SD-family models a too-low CFG makes prompts go fuzzy. And remember the text only takes effect if the POSITIVE output is actually wired into the sampler's positive input. It's easy to leave a node that has no input connected, type something hopeful, and wonder where it went.