GPTImage123 FAQ — Prompts, GPT-Image-2, Credits & More

Getting started

What is GPTImage123?

GPTImage123 is an online image-generation platform built around GPT-Image-2, OpenAI's latest image model. It offers text-to-image and image-to-image generation, a PPT generator, and credit-based billing — all in the browser, with no API setup required. New here? Start with the tutorial.

Do I need my own API key?

No. Generation runs through the platform, and you pay with credits. New accounts get a small amount of free credits to try before buying.

How does billing work?

Every generation spends credits. You can buy credits as a monthly subscription (credits refresh each month) or a one-time top-up (valid for a fixed window). See Pricing and your Credits page.

Writing better prompts

Prompt quality is the single biggest factor in your results. These techniques are drawn from official prompting guides published by Google (Gemini / DeepMind) and Microsoft (Azure OpenAI); see Sources.

What makes a good prompt?

Think in layers rather than a single word. A strong prompt usually covers some combination of:

• Subject — what/who is in the image.
• Composition & aspect ratio — framing, close-up vs wide, portrait vs landscape.
• Action or pose — what the subject is doing.
• Setting / location — where the scene takes place.
• Style — e.g. "watercolor," "isometric 3D render," "35mm film photograph."
• Camera & lighting — angle, lens, time of day, soft vs hard light.
• Text — the exact words you want rendered, in quotes.
• Reference inputs — images that anchor a character, product, or look.

You don't need all of these every time. Adding even one or two of them noticeably lifts quality. (Source: Google DeepMind & Google Cloud prompting guides.)

Can I just write a short prompt?

Yes. Modern models produce good output from one or two clear sentences. Add detail only where you need more control — think of specificity as a dial, not a requirement. (Source: Google Gemini image prompting tips.)

An example of leveling up a prompt

• Weak: a dog
• Better: a golden retriever puppy
• Strong: A golden retriever puppy sitting on a wooden porch at sunset, warm backlight, shallow depth of field, 35mm photograph, 3:2 aspect

The strong version specifies subject, setting, lighting, style, and framing — which is why the result is far more predictable.

How do I get accurate text inside an image?

Put the exact words in quotation marks (e.g. a poster that says "Grand Opening"). GPT-Image-2 is strong at rendering text — including many non-Latin scripts — but being explicit about the exact string reduces spelling errors. Keep text short for best results.

How do I keep a character or product consistent across images?

Use Image to Image and upload one or more reference images, then describe the change you want. This is the reliable way to keep a face, product, or brand look consistent across a set. (Source: model documentation for reference-image / editing features.)

Should I iterate or try to nail it in one shot?

Iterate. Treat prompting as a refinement loop: change one variable at a time (lighting, then style, then composition) so you can tell what actually improved the result.

Do prompt tricks transfer between models?

Not cleanly. Different models respond to different phrasing, so a prompt tuned for one model may need adjustment on another. If you switch models, expect to re-test. (Source: provider prompting guides.)

Getting the most out of GPT-Image-2

Why GPT-Image-2?

GPT-Image-2 (released April 2026) is OpenAI's flagship image model. Its notable strengths are strong instruction following, high-quality text rendering across many languages, and a reasoning-style approach that helps with complex, multi-element prompts. (Source: OpenAI / GPT Image documentation.)

What is it especially good at?

• Prompts with multiple elements or specific spatial relationships ("A on the left, B on the right").
• Text in images — signs, posters, labels, UI mockups.
• Editing and restyling existing images via reference inputs.
• Following detailed, structured instructions faithfully.

Any tips specific to GPT-Image-2?

• Describe layout explicitly when position matters ("centered," "top-left," "in the background").
• For edits, upload a clean reference and state the change plainly.
• Keep rendered text short and quoted.
• If a complex scene comes out muddled, break the request into clearer components rather than piling on adjectives.

Can I compare it with other models here?

Yes — the model selector lets you switch models to compare. For a broader landscape of Gemini, Seedream, Midjourney, FLUX, Z-Image and others, read our image model comparison.

Usage, rights, and troubleshooting

Can I use the images commercially?

Paid plans include commercial usage. Review the current terms on the Pricing page and the Terms of Service before publishing, and be mindful of third-party trademarks and likenesses in your prompts.

Why did my generation fail or look wrong?

• Not enough credits — top up on Pricing.
• Upload failed — remove and re-upload the reference image; check format and size.
• Off-target result — add specific style/composition terms, or use Image to Image with a reference.
• Slow generation — peak-time queues; higher tiers get priority.

Where do I find my past generations?

On the History page. It doubles as a personal prompt library — save the prompts that worked and reuse them.

Still stuck?

Start with the tutorial for a full walkthrough of every feature.

Sources

Prompt-engineering guidance above is based on official documentation:

• Google — Gemini image generation prompting tips
• Google DeepMind — Gemini image prompt guide
• Google Cloud — Image generation best practices
• Google — Prompting strategies
• Microsoft — Azure OpenAI vision prompt engineering
• OpenAI — GPT Image documentation and GPT Image overview