PhotoLoop
AI Text to Image Generator
Describe an image and generate it with GPT Image 2 — two credits each, in the same place you make your videos.
Describe the layout, not just the subject
The single biggest difference between a prompt that works and one that does not is whether it says anything about composition. Where the subject sits in the frame, how much empty space surrounds it, what the light is doing. A prompt naming only the subject returns the statistical average of that subject; a prompt naming the frame returns something you can use.
Two credits, no subscription
GPT Image 2 costs two credits an image. Nano Banana 2 Lite is also two, and the full Nano Banana 2 is four. A credit is one cent of supplier cost, so these are the real numbers rather than a marked-up currency. Nothing recurring, nothing expiring monthly.
Text inside images actually works here
Most image models mangle lettering. GPT Image 2 is noticeably better at it, which is why it is our default when the output is a poster, a card or anything carrying a headline. It is still best-effort: keep required copy short, and expect to fix details in an editor when the text has to be exact.
Fast enough to iterate
Ten to twenty seconds an image, against two to four minutes for video. That gap changes how you work: with images it is worth generating several variations of a prompt and picking, where with video each attempt is a real commitment of time.
- GPT Image 2 at 2 credits an image
- 1:1, 16:9, 9:16, 4:3 and 3:4
- Strong text rendering for posters and cards
- 10–20 seconds, not minutes
Questions
What does one image cost?
Two credits with GPT Image 2, which is one cent of supplier cost rounded up. Nano Banana 2 Lite is also two; the full Nano Banana 2 is four. There is no subscription and no per-month minimum — you spend credits per image.
How is this different from writing a video prompt?
Images reward describing layout rather than motion. Say where things sit in the frame, how much empty space you want, what the light is doing. Camera paths and timing, which matter for video, do nothing here.
Can it render text inside the image?
GPT Image 2 is unusually good at it compared to most image models, which is why it is our default for posters and anything with a headline. Treat it as best-effort anyway: keep required text short and plan to fix it in an editor if it matters.
What aspect ratios are available?
1:1, 16:9, 9:16, 4:3 and 3:4. These go to the API as ratio strings, not pixel dimensions — a detail that trips up a lot of integrations, since passing something like 1024x1024 is rejected outright.
Do I need an account?
Yes for image generation. The no-account free tier covers image-to-video only. New accounts start with credits, so you can generate a number of images before deciding whether to buy more.
How long does an image take?
Ten to twenty seconds, an order of magnitude faster than video. Images are a synchronous-feeling wait; video is a coffee break.
What makes a prompt produce a deliberate-looking image?
Constrain the palette, name the lighting setup, and specify the camera and lens if you want photographic output. 'A mug on a table' averages every mug photo. 'A single ceramic mug on pale linen, 50mm at f/2, soft window light from the left, muted warm palette, generous space above for a headline' produces a shot you could actually use.