DallERender

DallERender is the most honest node in its pack: it takes a prompt, calls OpenAI's DALL-E 3 API, downloads the image it gets back, and hands you a normal ComfyUI IMAGE tensor. That's the whole job. No local model, no VRAM, no diffusion loop on your machine - the generation happens on OpenAI's side and you pay per image.

So when would you actually reach for it? If you're on a machine that can't run SDXL comfortably, or you want OpenAI's particular look dropped into a bigger graph - say, to feed a DALL-E result into an upscaler or use it as an img2img reference. It's also the fastest way to see a picture for a prompt you're still editing. Just don't expect it to be cheap at volume: DALL-E 3 is billed per image, and the censorship is as well documented as anything in image AI.

How it works

The pack is old-school in the worst and most literal way: the API key is hardcoded in the source file. At import time nodes.py literally runs os.environ["OPENAI_API_KEY"] = "sk-", then builds an OpenAI() client. When you run the node, it calls client.images.generate(model="dall-e-3", ...), grabs the URL OpenAI returns, downloads it, and converts it into the float32 image tensor every other node in ComfyUI speaks. You get exactly one image per run.

That means the README's terse install instructions are not optional garnish - "Fill API keys in code" is the actual setup step.

The inputs that matter

There's only one: prompt, a string input (marked as a forced input, so wire it from a text node). That's it. The IMAGE output plugs into anything that accepts an image - PreviewImage, SaveImage, an upscaler, an img2img sampler.

Installing it

From ComfyUI Manager, search for "Memeplex_DALLE" or the pack title. Or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/shi3z/ComfyUI_Memeplex_DALLE
pip install openai

There's no requirements.txt in the repo, so Manager won't install the OpenAI package for you - that pip install openai is mandatory. The whole pack fails to load without it. Then open ComfyUI/custom_nodes/ComfyUI_Memeplex_DALLE/nodes.py, find the line setting OPENAI_API_KEY to "sk-", and put your real key in there before restarting ComfyUI.

Where people get burned

The big one: setting OPENAI_API_KEY in your shell does nothing. The import line overwrites the environment variable with the placeholder, so the key you edit must be the one in nodes.py. If you forget, you'll get an auth error the moment the node runs, not at startup.

Second, the model name is hardwired to dall-e-3. OpenAI has been moving everyone off the legacy images API toward gpt-image-1 for a while now - this node will keep working until the day it quietly doesn't, and there's no dropdown to change it. If you want a prompt-driven cloud renderer with a future, this isn't it. If you want one image of a cat samurai without touching your GPU, it's genuinely fine.