How to Use GPT Image 2: Step-by-Step Guide with Prompts
GPT Image 2 excels at text in images, complex compositions, and editing photos you upload. This guide walks through the actual steps, gives you copy-paste prompts, and flags the settings that quietly double your cost.
August 6, 2026 · 6 min read
Contents
Quick answer
Open the GPT Image 2 editor, type a prompt, and hit Generate. Guests get 2 free images a day via Shotari Basic. To run GPT Image 2 specifically, sign up and add paid credits (5 per image at 1K). You can also attach up to 3 reference photos and describe the edit you want — the model edits rather than generating from scratch.
Step-by-step
1. Go to the GPT Image 2 page
Visit [shotari.com/gpt-image-2](/gpt-image-2). No download, no install — it runs in your browser.
2. Write your prompt
Be specific. 'A red ceramic mug on a marble counter, soft morning light from the left, shallow depth of field' beats 'a mug'. Name the material, lighting direction, and camera style.
3. Attach reference images (optional)
Click the image icon and upload up to 3 photos. Use one for the subject, one for the lighting/mood, and one for the background style. The model combines them with your text prompt.
4. Pick resolution and aspect ratio
Start at 1K and square (1:1). Only switch to 2K/4K or 16:9/9:16 for the final keeper — resolution and non-square ratios multiply the credit cost. If you need a different aspect ratio after generating, [expand the canvas with AI outpainting](/uncrop) instead of regenerating — see the [uncrop guide](/blog/how-to-uncrop-an-image-with-ai) for per-ratio costs.
5. Hit Generate
Generation takes 15-40 seconds. The result appears inline. If you are a guest, the image is generated with Shotari Basic (free) — sign in and add paid credits to use GPT Image 2 specifically.
6. Download
Click download. Logged-in users get the full-resolution image. Guests need to sign up first (free, one click with Google).
Prompt examples you can copy
GPT Image 2 handles text rendering and complex instructions better than most models. These prompts exploit that strength — each one asks for something that trips up simpler models. For more starting points, browse the prompt templates page.
Product shot with text overlay
A minimalist product photo of a glass perfume bottle on a black stone slab. Soft studio lighting. Add the text 'NOIR' in thin white serif font centered below the bottle.
Before/after comparison
Split-screen image: left side shows a cluttered desk with papers and coffee stains, right side shows the same desk organized and clean. Label the left 'BEFORE' and the right 'AFTER' in bold sans-serif.
Infographic-style
A vertical infographic titled '5 MORNING HABITS' with numbered items, each with a small icon: 1. Hydrate (water glass), 2. Stretch (yoga pose), 3. Journal (notebook), 4. Walk (footprints), 5. Plan (calendar). Clean flat design, blue and white palette.
Photo edit — change background
Keep the person exactly as they are. Replace the background with a warm sunset over the ocean. Match the lighting on the subject to the new background.
What it costs
GPT Image 2 costs 5 credits per image at standard resolution. Resolution and aspect ratio multiply that base. Here is the real cost table so you can budget before you start.
| Resolution | Aspect ratio | Credits |
|---|---|---|
| 1K (Standard) | Square / 4:3 / 3:4 | 5 |
| 1K | 16:9 or 9:16 | 6 |
| 2K | Square / 4:3 / 3:4 | 8 |
| 2K | 16:9 or 9:16 | 9 |
| 4K | Square / 4:3 / 3:4 | 10 |
| 4K | 16:9 or 9:16 | 11 |
Free credits (20 on signup + 10 daily) only work on Shotari Basic and Pro. GPT Image 2 needs paid credits — the smallest pack is $29.90 for 1,500 credits (300 images at 1K). See is GPT Image 2 free? for the full breakdown.
Common mistakes that waste credits
- ⚠
Generating at 4K before you have the right composition
A 4K generation costs 10 credits instead of 5. If you are still experimenting with prompt wording, you are paying double for drafts. Draft at 1K, finalize at 4K.
- ⚠
Using 16:9 when you do not need it
The widescreen multiplier is 1.1x. On a $29.90 pack that is 50 fewer images over 1,500 credits. Only use non-square ratios when the output genuinely needs it (YouTube thumbnails, banners).
- ⚠
Vague prompts on a 5-credit model
GPT Image 2 is not the place to explore. Draft your composition on Shotari Basic (free) first. Once you know the right prompt, switch to GPT Image 2 for the final render.
- ⚠
Expecting free credits to work
Your signup and daily check-in credits are for Shotari Basic and Pro only. Selecting GPT Image 2 with a free-credit balance triggers 'insufficient credits' even though the number is above 5. You need paid credits.
Editing an existing photo
GPT Image 2 is not just a generator — it edits photos you upload. Attach up to 3 reference images and describe what to change. The model preserves what you did not mention and modifies what you did.
The key to good edits
Name what to keep, not just what to change. 'Keep the person exactly as they are, change the background to a studio setting' works. 'Change the background' alone may alter the subject too.
FAQ
Is GPT Image 2 the same as DALL-E?
No. GPT Image 2 is OpenAI's newer image model with significantly better text rendering and instruction following. DALL-E 3 is the previous generation.
Can GPT Image 2 generate text in images?
Yes — this is its strongest advantage over other models. It renders readable text in specified fonts, positions, and sizes. Use quotes around the exact text you want.
How many images can I generate per day?
Guests: 2 per day (Shotari Basic, free). Logged in with paid credits: unlimited, as long as your balance covers 5 credits per image. No daily cap on paid generations.
Why is my generation failing?
Most failures are content moderation (the model refuses certain prompts), credit issues (free credits do not work on GPT Image 2), or temporary overload. See [GPT Image 2 not working](/blog/gpt-image-2-not-working) for the full troubleshooting guide.
Try GPT Image 2 now
Draft free on Shotari Basic. Switch to GPT Image 2 when your prompt is ready.
Open GPT Image 2 editor