AI Models
What each model is good at, where it differs from the others, and what it costs to run.
OpenAI's current image model. Built for a long Visxy library prompt — layout, type, and what to keep — not a five-word vibe.
Credits come with a Generator plan. The exact cost is shown on the Generate button before anything is spent.
GPT Image 2 is OpenAI's current image model — the successor to GPT Image 1 / 1.5, not DALL·E 3 with a new name. On Visxy it is a long image-library prompt: subject, camera, light, and especially words that have to read. OpenAI sells instruction-following and in-image type as the point. There is no other GPT Image row here, so there is nothing to "upgrade from" in the picker.
The full brief. Visxy prompts are already paragraphs. GPT Image 2 is the model that is supposed to keep the fifth constraint instead of dropping it. Spell the headline. Name what sits left and what sits right. "Cinematic, 8k, ultra detailed" is how you get a generic face.
It is also the edit model in this family. Layouts, labels, a product that must stay itself while the set changes — that is still this page with a picture attached.
What it is not: a five-word mood engine, and not ChatGPT's "thinking" studio. Hosts who expose a reasoning pass are describing OpenAI's own app. Visxy does not put that chip on this composer.
There is no GPT Image 1.5 in this catalog. The comparison that matters is GPT Image 2 used as a poster versus used as a vibe. A prompt that is really atmosphere with no type does not need this page's strengths. A prompt that contains a sentence, a UI, or a pack label does.
Do not treat it as a silent upgrade of every other stills model on the site. It is OpenAI's row. Use it when the library brief is a set of instructions.
Paste the library prompt. If it has type, write the string exactly. If it has a count ("three glasses"), keep the number.
9:16 when it came from the image library. auto is only useful when you truly do not care about the crop; a Visxy portrait already knows.
Structure helps when the prompt is long: scene, subject, the one detail that must not move, then what the file is for. OpenAI's own guide is that spine. Flattery at the end ("masterpiece") does not.
No Visxy run is published yet. Generate fewer. Change one constraint.
There is no public GPT Image 2 Edit page. Attach a still to this model. Say what changes and what stays: "replace the headline with VISXY, keep the bottle." A new brief ("luxury ad") reads as start over. Detail that is not in the source will be invented.
No. This catalog row is GPT Image 2. Older OpenAI image names are not listed here.
Yes — especially the long 9:16 briefs that already name layout and type. A five-word mood belongs somewhere else in your own workflow, not as a rewrite of this page.
Not as a Visxy control. If a host documents that pass, it is their product, not this composer.
No. Attach a reference to GPT Image 2 and describe the change. We do not publish a separate article for that shape.
