Ideogram 4.0

If you've ever needed an image that contains legible text - a poster, a logo, a sign - you know Ideogram is the model that made it look easy while everything else mangled letters into eldritch runes. Ideogram 4.0 (ideogram:4@0) is the current generation, and this node runs it on Runware's cloud. But here's the thing that will decide whether you love or hate this node: Ideogram 4 doesn't take a plain prompt. It takes a structured prompt, and the node's required input is settings.structuredPrompt.high_level_description - one or two sentences describing the full image, subject and style included.

That's not bureaucracy for its own sake. Ideogram's structured prompt is the model's actual input format, and the reason its text rendering and composition are so good. The node lets you write the high-level description and optionally fill the structured slots: art_style, medium, photo (photographic style), aesthetics, lighting, and compositional_deconstruction.background. The plain positivePrompt field exists as a convenience - the tooltip says it's automatically expanded into a structured prompt before generation. So you can start plain and go structured when you need control.

How it works, and the licensing elephant

Ideogram 4 is open weights, but it shipped with a non-commercial license and a trained-in safety filter that emits a refusal image instead of your prompt. On this node, the copyrightDetection toggle opts into post-generation Hive likeness-and-logo checks. The takeaway: this is a fantastic model for personal work and internal design iteration, and a legal minefield for commercial use - check the license before you put it in a client deliverable.

Inputs that matter

• settings.structuredPrompt.high_level_description (required) - your one-or-two-sentence spec. This is the single most important input on the node.
• settings.structuredPrompt.style_description.art_style / medium / photo / aesthetics / lighting - the structured slots; the more you fill, the more deterministic the output.
• settings.structuredPrompt.compositional_deconstruction.background - environment, framing, lighting context.
• positivePrompt - freeform prompt, auto-expanded into the structured format. Easiest start; the structured fields override when set.
• renderingSpeed - DEFAULT/speed-quality tradeoff dropdown.
• width / height (1024 default), numberResults (1–20), safety gate, outputFormat/outputQuality, and advanced_json for the slots without dedicated widgets (the tooltip mentions compositional_deconstruction.elements and color_palette fields).

The image output is a native IMAGE into Preview or Save, cost on the title bar.

Where people get burned

• The required field is a structured slot, not the prompt box. New users type into positivePrompt, leave high_level_description empty, and get a "required input" wall. The required one is the structured description.
• Licensing, again. Non-commercial. The model's the best text renderer in the pack and it's also the one you can't ship with. That's Ideogram's whole controversy in one sentence.
• The safety filter is trained in. If your prompt trips it, you get a refusal image, not an error - the node can't route around what's baked into the weights.
• Structured vs. freeform - mixing both is fine (structured fields win), but if you're getting results that ignore your positivePrompt, it's because the structured slots are carrying the weight.

Ideogram 4 is the node you reach for when the text has to be right and the license allows it. For everything else, an open Apache-2.0 model will get you most of the way without the fine print.