GPT Image 2 · Text to Image

Seven inputs and a reputation. GPT Image 2 is the model family that got people past "AI text in images looks like alien runes", and this node is the plain text-to-image version of it - no source images, no mask, just a prompt and a batch of results. Starting at about 0.27 credits ($0.0027) per image.

What it is

A PopNow node wrapping OpenAI's GPT Image 2 text-to-image endpoint. Cloud inference, per-image billing, nothing downloaded. It's one of roughly 248 model nodes in the pack, which is worth a sentence of explanation if this is your first: the pack doesn't run models, it drives PopNow's API, so every node in it works on a laptop, a Mac, or a CPU-only box, and none of them need a checkpoint folder.

Where it sits in the lineup: below the 2.5 models in freshness, and at the same starting rate. Pick it when you want the older model's look - a bit less clinical than Sunburst - or when you want a cheap, reliable render and don't need 4K.

How it works

Prompt plus settings go out; PopNow quotes the cost, holds at most the quote, renders, and the node decodes the returned files into an IMAGE batch. Batch outputs go to Save Image, which writes a file per image rather than overwriting, so a num_outputs of 4 gives you four files.

There's nothing cached and no seed field, which has two consequences: results aren't reproducible by number, and every queue is a purchase. Save the images you like as you go rather than planning to regenerate them later.

The inputs

• prompt - describe the image you want. Detailed prompts win; this model reads a paragraph and holds most of it, which is not true of a lot of generators. For text in the image, quote the exact words and say where they sit.
• aspect_ratio - 1:1 default, with the wide spread including 16:9, 9:16, 4:5, 2:3 and the banner extremes.
• resolution - 1k, 2k, 4k.
• quality - medium by default on this model, with low and high available. Note the default differs from the 2.5 nodes, which start at high - so a like-for-like comparison between them needs the quality matched, not left alone.
• num_outputs - 1 to 10, each billed separately.
• output_format, output_quality - png default; JPEG/WebP at a quality you pick from 0 to 100, or -1 to let the model decide.

Outputs: images and task id.

Where it fits in a real graph

Its best role is the front half of an edit pipeline. Generate a base image here where re-rolling is cheap, then hand the one you like to an edit node - in this pack or a local one - for the changes that need to be surgical. Generate-then-edit has been the community's consistent answer to consistency and revision since late 2025, and text-to-image is the generate half.

The other role is straight asset production: social crops in a batch, thumbnails, illustrated headers. Cheap, fast, and boring in the way you want production assets to be boring.

Install

Manager → search PopNow → Install, or:

cd ComfyUI/custom_nodes
git clone https://github.com/popnow-ai/popnow-comfyui

Restart, paste a key from popnow.ai/api/console/keys into Settings > PopNow, restart once more so the node list is built from your account's catalog. No dependencies beyond ComfyUI itself - the pack's install list is empty.

Errors and expectations

No PopNow API key is set is a missing key; PopNow 401 is a bad or revoked one. PopNow 40201 is credits, PopNow content_rejected is the provider filter and gets refunded. If the node list looks short or a model is missing, restart - the catalog is fetched from your account at startup and the bundled snapshot is only the fallback.

And the honest caveat that comes with every closed model: your prompt leaves the machine, the filter applies, and there is no local workaround. If any of that is a dealbreaker, this whole pack is the wrong tool - but for a legible poster by tonight, it's very hard to beat on price.