Ideogramv3 (fal)

If you've ever asked a diffusion model to put actual readable words on a sign and gotten back cursed alphabet soup, Ideogramv3 is the node for that specific pain. Ideogram is the Toronto lab founded by the Google Brain researchers behind DDPM and Imagen, and for three closed generations it was the benchmark for accurate text inside images. This node calls Ideogram v3 through fal - closed model, hosted, no weights on your disk - and returns a native ComfyUI IMAGE.

Reach for it when the image has to say something: posters, packaging mockups, memes with legible captions, logos with a tagline. For plain aesthetic generation your local models are fine; text is where Ideogram pulls ahead. The node ships in gokayfem's ComfyUI-fal-API pack.

How it works

It's an API passthrough - your prompt and settings go to fal, fal calls Ideogram's v3 endpoint, and the result comes back as an IMAGE. The trick to good text is entirely in the prompt: put the exact words you want in quotes. Something like a vintage travel poster that says "GREETINGS FROM MARS" will render that phrase far more reliably than describing it.

The inputs and outputs that matter

For this node the two that genuinely move things are:

• prompt - where you specify both the scene and, in quotes, any text to render.
• image_size - presets from square_hd through landscape_16_9, plus custom. Pick the shape that suits a poster or a banner.

The node also exposes width / height (512–1440, step 32 - used with custom), num_inference_steps (28), guidance_scale (3.5), num_images (1–10), safety_tolerance (1 strict → 6 permissive), and an optional seed (-1 random). On a closed API model most of those are best left at their defaults; the prompt is doing the real work. Output is a single IMAGE.

How to install it

Via ComfyUI Manager: search ComfyUI-fal-API, install, restart. Manually:

cd ComfyUI/custom_nodes
git clone https://github.com/gokayfem/ComfyUI-fal-API.git
pip install -r ComfyUI-fal-API/requirements.txt

No models to download. Add a fal API key to config.ini under [API] (copy config.ini.example) or export FAL_KEY=your_key, restart, and find it under FAL.

Common issues & troubleshooting

• Auth error. FAL_KEY missing or wrong. Set it, restart.
• The text still comes out garbled. Quote the exact string, keep it short, and don't ask for a paragraph - even Ideogram degrades with long text. Re-roll the seed; text rendering is a bit of a slot machine.
• A prompt gets refused. Ideogram bakes safety into the model itself, so some prompts return a refusal regardless of settings - that's the model, not the node. The pack surfaces fal's actual error text, so you'll see what happened.
• Cost. It's a paid per-image API call. The pack's result cache serves repeat identical calls free (force_rerun to bypass) and the spend guard caps a session.
• Windows Portable. Missing fal-client? python_embeded\python.exe -m pip install fal-client.

Worth knowing the landscape: Ideogram later opened weights for a newer generation, but the community largely moved to other models over licensing friction. None of that changes what this node is good at today - if you need words in the picture and want them right, this is a solid tool to keep around.