GPT Image 1 Mini

GPT Image 1 Mini is the budget member of OpenAI's image family, and the Runware pack wraps it in exactly the same node shape as its bigger siblings. Same required inputs, same IMAGE output, same wiring. What changes is the price and the ceiling: Mini is dramatically cheaper per image and correspondingly less ambitious. You use it when you want GPT's reliable text rendering and clean composition for a quick asset - a mockup, a thumbnail, an inpainting pass - without paying full-fat rates.

The honest framing from the KB's closed-source-models essay: these closed models win on things like legible text, and you trade away open weights, control, and (with OpenAI's guardrails) some prompts. Mini is the version of that trade where at least the bill is small.

What you actually set

• positivePrompt - required. Mini handles straightforward instructions fine; don't ask it for a 17-element scene it has to cram in.
• width / height - default 1024×1024.
• referenceImages - IMAGE socket, so it still does image-to-image and natively multimodal edits.
• providerSettings.openai.quality - auto, high, medium, low. Set high if you're using Mini for something that actually needs to be good.
• providerSettings.openai.background - auto, opaque, transparent.
• providerSettings.openai.moderation - auto or low.
• numberResults - up to 4 variations.

Output: one image socket, type IMAGE, into Preview or Save. Two-node workflow, same as the other OpenAI image nodes.

Install

The pack-wide routine, one time: ComfyUI Manager → search Runware → install → restart, or

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

Then a Runware API key into ComfyUI Settings → Runware API key (or RUNWARE_API_KEY). Nothing else installs; the models live on Runware's side.

Where people get burned

• "Mini" is a quality promise, not just a price tag. If you're used to GPT Image 1 and drop Mini in, the detail gap on complex scenes will surprise you. Use it for what it's for - cheap, decent images - not for hero shots.
• Moderation still applies. moderation: auto will refuse prompts the same way its pricier sibling does. The filter doesn't get cheaper.
• Default JPG. If you enabled background: transparent, flip outputFormat to PNG or your alpha silently disappears.

The takeaway: this is the node to reach for when the image is a means to an end - a placeholder, a draft, a comp - and you'd rather not spend full-size money on it. When the image is the deliverable, step up.