OpenAI GPT Image 2 on Pixel Dojo
GPT Image 2 AI Image Generator
GPT Image 2 renders the words you put in quotes cleanly, from poster headlines and bottle labels to Korean shop signs and app buttons.
GPT Image 2 is OpenAI's image model for designs that carry words: event posters, packaging, menus, app screens and infographics, up to a full 3840 x 2160 file. It shines without text too, with lifelike portraits and crisp architectural interiors. We made every image on this page on Pixel Dojo, and each one comes with its exact prompt so you can start from it.
5 credits ($0.31) per image at 1024x1024. Included in every plan, from $10/month.
Already a member? Sign in
By Brian Lovett · Updated
Why designers choose GPT Image 2 for text in images
OpenAI built GPT Image 2, which launched in ChatGPT as ChatGPT Images 2.0, for pictures where the words matter. Every text example on this page shows the quoted copy spelled exactly as written: a six block event poster, a three line bottle label, a bilingual shop sign with a handwritten Korean menu and an app screen with eleven interface strings.
It reads a prompt like a design brief. Name the artifact first (poster, label, infographic, app screen), put every exact string in double quotes, and say what role each string plays: headline, subhead, callout, button. It sizes and stacks the type to fit each role, so none of our prompts needed font sizes.
It is a strong photographer as well. The portrait and the library hall on this page carry no text at all, and they show believable skin, straight architectural verticals and sunlight that falls exactly where the windows are.
Reach for it when you need event posters and flyers, packaging and label mockups, menus and storefront signs in more than one language, app and web screens for pitch decks, infographics and explainer graphics, or polished editorial photography.
GPT Image 2 prompts and examples: what you can create
Six images we made on Pixel Dojo at High quality, one for each kind of job: a street scene with bilingual signage, an event poster, an editorial portrait, a beverage ad, an app mockup and an architectural interior. Tap Use this prompt on any tile to start from it, and set Image size to the size in the caption for the same framing.
Night street photo at 1920x1080. The glowing 밤국수 NIGHT NOODLES sign and the handwritten Korean menu are spelled exactly as prompted, with wok fire, rain and neon reflections around them. Start here for any storefront or signage scene.
Photorealistic street photograph at night, eye level, 35mm lens. A cook in a white cap tosses noodles over roaring wok flames at a tiny alley stall in Seoul, steam and orange sparks lifting into the rain, wet asphalt mirroring the neon. The glowing sign above the stall reads "밤국수" in bold Korean letters with "NIGHT NOODLES" in English beneath it, and a handwritten menu board on the counter reads "해물칼국수 9,000원". Mixed tungsten and magenta neon light, shallow depth of field, fine film grain, candid documentary feel.
Swiss style event poster at 1920x1080. Six blocks of copy, from the giant vermilion headline to three small callouts, each placed and spelled as written. Swap in your own event details and run it.
A bold Swiss style event poster in landscape format, printed on heavy uncoated paper and taped to a raw concrete wall, lit by one hard spotlight from the upper left that throws a crisp shadow off the curled corners. Giant condensed sans serif headline in vermilion: "NIGHT OF FALLING STARS". A large cobalt blue circle overlaps the headline like a rising moon. Subhead in thin black serif: "An open air evening at Halden Ridge Observatory". Date block in heavy black type: "12 AUGUST / 22:00". Three small callouts aligned along the bottom edge: "Telescopes open", "Live string quartet", "Hot cider bar". Fine paper grain, slight ink spread at the letter edges, every word spelled exactly as written.
Editorial portrait at 1024x1536. Natural skin with fine lines and freckles, bold cobalt and red color blocking, and a crisp window shadow across the face. A strong base for fashion and beauty campaigns.
Fashion editorial portrait, medium close up, 85mm lens. A woman in her sixties with a cropped silver pixie cut and bold red lacquer earrings stands against a sunlit ochre plaster wall in a sculptural cobalt wool coat with an oversized collar, gazing just past the camera. Hard late afternoon sun throws a crisp window frame shadow across her cheek and the wall, natural skin texture with fine lines and freckles, warm 35mm color negative film look, magazine quality.
Beverage ad at 1920x1080. SOLANA, Blood Orange Sparkling Tonic and 330 ml sit crisp on the label behind a frozen crown splash. Put your own product name in quotes and keep the lighting.
Studio beverage advertisement photograph, low angle macro. A frosted glass bottle of blood orange soda rises through a crown shaped splash of ice cold water against a deep teal backdrop, halved blood oranges and ice cubes tumbling beside it, condensation beads across the glass. The paper label reads exactly "SOLANA" in tall art deco capitals, with "Blood Orange Sparkling Tonic" in small serif beneath and "330 ml" at the bottom. Hard rim light from behind, soft frontal fill, frozen splash with crisp droplets, glossy commercial color grade.
App screen mockup at 1024x1536. Every interface string from the prompt, from the SWELLCAST header to the tide times and the Set dawn alert button, on a phone that looks photographed. Made for pitch decks and landing pages.
A clean product design mockup: a modern smartphone lying face up on sand colored linen next to a folded wetsuit sleeve and a few smooth pebbles, top down, soft overcast daylight. The screen shows a surf forecast app with legible interface text. Header: "SWELLCAST". Location line: "Porthleven Reef". A large number reads "1.8 m" with the label "Swell height" beneath it. A row of three small cards reads "Period 11 s", "Wind 12 km/h offshore", "Water 16°C". A tide curve chart is labeled "High 06:42", "Low 12:58", "High 19:10". A rounded button at the bottom reads "Set dawn alert". Deep navy interface with coral accents, crisp pixel sharp typography, subtle screen reflection.
Architectural interior at 1920x1080. A brutalist reading hall with true verticals, a deep coffered ceiling, warm light across terrazzo and one reader for scale. Start here for hospitality, real estate or set design boards.
Architectural interior photograph, wide 24mm lens at eye level, perfectly vertical lines. A brutalist concrete library reading hall with a deep coffered ceiling and a long oak table, late afternoon sun pouring through tall slot windows in sharp golden beams through floating dust, one reader in a mustard sweater at the far end for scale. Warm sun against cool concrete shade, polished terrazzo floor reflections, calm and monumental, fine detail.
GPT Image 2 settings and what each one gives you
Besides the prompt, the GPT Image 2 tool has four settings. Here is what each one does and what it costs.
- QualityLow, Medium, High
- Low is 0.5 credits, Medium 1.5 credits and High 5 credits per image at standard sizes. Low already renders clean, correctly spelled text, great for social graphics and big batches. High gives the richest texture and finest line work, and it is what we used for every showcase image.
- Image size1024x768, 1024x1024, 1024x1536, 1920x1080, 2560x1440, 3840x2160
- For full 4K, pick 3840x2160: a true 3840 x 2160 file for 10 credits at High or 3 credits at Medium, ready for print and big screens. 1920x1080 is the everyday widescreen pick and returns a crisp 2048 x 1152 file. 1024x1536 suits portraits and phone screens, and 1024x1024 suits square social posts.
- Number of images1 to 4 per request
- Ask for up to four at once to get a set of concepts in one go. Each image uses its own credits.
- Output formatPNG, JPEG, WebP
- PNG keeps every detail for print and big screens. Pick JPEG or WebP when the image is headed for a web page.
Choose your GPT Image 2 quality: one infographic at every tier, plus 4K
One infographic prompt at Low, Medium and High, so you can see what each tier adds, plus the poster at full 4K and a second SOLANA ad concept from the same prompt. Tap Use this prompt on any tile to make your own version.
Low, 0.5 credits: the title, five part labels and four numbered steps, all spelled correctly on a clean editorial layout.
An editorial infographic poster in portrait format titled "ANATOMY OF A SOURDOUGH LOAF" in large bold serif at the top. A photorealistic half sliced loaf sits in the center on a linen cloth, open glossy crumb visible, with thin leader lines pointing to five labeled parts: "Ear", "Blistered crust", "Open crumb", "Scoring line", "Bottom crust". A strip along the bottom shows four numbered steps with small icons: "1. Mix", "2. Bulk ferment 5 hours", "3. Shape and cold proof overnight", "4. Bake at 250°C". Cream paper background, charcoal and rust accents, clean editorial grid, every label sharp and correctly spelled.
Medium, 1.5 credits: adds a short, correctly spelled explainer under every label and step, for a finished magazine page look.
An editorial infographic poster in portrait format titled "ANATOMY OF A SOURDOUGH LOAF" in large bold serif at the top. A photorealistic half sliced loaf sits in the center on a linen cloth, open glossy crumb visible, with thin leader lines pointing to five labeled parts: "Ear", "Blistered crust", "Open crumb", "Scoring line", "Bottom crust". A strip along the bottom shows four numbered steps with small icons: "1. Mix", "2. Bulk ferment 5 hours", "3. Shape and cold proof overnight", "4. Bake at 250°C". Cream paper background, charcoal and rust accents, clean editorial grid, every label sharp and correctly spelled.
High, 5 credits: the richest crust and crumb texture, a framed layout and finely hatched step icons down to the cast iron pot.
An editorial infographic poster in portrait format titled "ANATOMY OF A SOURDOUGH LOAF" in large bold serif at the top. A photorealistic half sliced loaf sits in the center on a linen cloth, open glossy crumb visible, with thin leader lines pointing to five labeled parts: "Ear", "Blistered crust", "Open crumb", "Scoring line", "Bottom crust". A strip along the bottom shows four numbered steps with small icons: "1. Mix", "2. Bulk ferment 5 hours", "3. Shape and cold proof overnight", "4. Bake at 250°C". Cream paper background, charcoal and rust accents, clean editorial grid, every label sharp and correctly spelled.
The poster prompt at 3840x2160 for 10 credits: a full 4K file with every line of copy intact, in a bold new stacked layout.
A bold Swiss style event poster in landscape format, printed on heavy uncoated paper and taped to a raw concrete wall, lit by one hard spotlight from the upper left that throws a crisp shadow off the curled corners. Giant condensed sans serif headline in vermilion: "NIGHT OF FALLING STARS". A large cobalt blue circle overlaps the headline like a rising moon. Subhead in thin black serif: "An open air evening at Halden Ridge Observatory". Date block in heavy black type: "12 AUGUST / 22:00". Three small callouts aligned along the bottom edge: "Telescopes open", "Live string quartet", "Hot cider bar". Fine paper grain, slight ink spread at the letter edges, every word spelled exactly as written.
A second SOLANA ad concept from the same prompt: a sunburst under the copy, a taller splash and a darker backdrop, with SOLANA, Blood Orange Sparkling Tonic and 330 ml on the label.
Studio beverage advertisement photograph, low angle macro. A frosted glass bottle of blood orange soda rises through a crown shaped splash of ice cold water against a deep teal backdrop, halved blood oranges and ice cubes tumbling beside it, condensation beads across the glass. The paper label reads exactly "SOLANA" in tall art deco capitals, with "Blood Orange Sparkling Tonic" in small serif beneath and "330 ml" at the bottom. Hard rim light from behind, soft frontal fill, frozen splash with crisp droplets, glossy commercial color grade.
- Clean text at every tier. Low spells the title, all five part labels and all four steps correctly for just 0.5 credits, ideal for social graphics and big batches.
- Medium added short, correctly spelled explainers under each label and step on its own, an easy route to a finished editorial look.
- High is where GPT Image 2 shows off: the deepest texture, the finest line work and the most designed layouts. Every showcase image on this page is High quality.
- Need print or big screen resolution? 3840x2160 returns a full 3840 x 2160 file for 10 credits at High, with the same copy as the widescreen version.
- One prompt, many concepts. Both SOLANA ads carry the exact label copy, each with its own label art and splash, so one prompt gives you a range of ad ideas.
- Quote every word that matters and give it a role: headline, subhead, callout or button. We wrote every text prompt on this page that way, and it is the surest route to exact text.
GPT Image 2 pricing on Pixel Dojo
GPT Image 2 is included in every Pixel Dojo plan, and plans start at $10/month. Each image uses credits from your plan: 0.5 credits at Low, 1.5 credits at Medium and 5 credits at High, at every size up to 2560x1440.
At full 4K (3840x2160), Medium is 3 credits and High is 10 credits. Ask for up to four images in one request and each one uses its own credits.
Every showcase image on this page is a 5 credit High image, and the 4K poster is a 10 credit image. Open any of them with Use this prompt and make it yours.
GPT Image 2 questions
What is GPT Image 2?
GPT Image 2 is OpenAI's image generation model, launched in ChatGPT as ChatGPT Images 2.0. It is known for rendering readable text inside images. On Pixel Dojo you can use it in the web tool or through the API with the model id gpt-image-2.
What is GPT Image 2 best at?
Designs that carry words: posters, packaging and labels, menus and storefront signs (Korean included), app screens and infographics. It also makes strong photographic images, like the editorial portrait and the architectural interior on this page.
Can GPT Image 2 put readable text in images?
Yes, and it is the main reason to choose it. Put each exact string in double quotes and name its role, such as headline, label or button. Our poster, soda label, bilingual noodle stall sign and app screen all show the quoted copy spelled exactly as written.
Does GPT Image 2 make 4K images?
Yes. Choose 3840x2160 and you get a full 3840 x 2160 file, ready for print and big screens. You can also pick 1024x768, 1024x1024, 1024x1536, 1920x1080 or 2560x1440 for landscape, square, portrait and widescreen work.
How much does GPT Image 2 cost?
GPT Image 2 is included in every Pixel Dojo plan, from $10/month. An image costs 0.5 credits at Low, 1.5 credits at Medium and 5 credits at High. At 3840x2160, Medium is 3 credits and High is 10 credits.
How do I use GPT Image 2 on Pixel Dojo?
Type your prompt in the box on this page, or tap Use this prompt on any example, then pick a quality tier and an image size and generate. To change an existing picture, use the GPT Image 2 Edit tool. Developers can call the model through the Pixel Dojo API at the same credit prices.
How does GPT Image 2 compare with GPT Image 2.5 and Nano Banana Pro?
All three put readable text in images. GPT Image 2 is OpenAI's ChatGPT Images 2.0 model, with Low, Medium and High tiers and full 4K output at 5 credits per High image. On a shared bakery prompt it spelled the chalkboard exactly and caught fine details like the flour dusted forearms we asked for. GPT Image 2.5 is 2 credits per High image, great for high volume work, and Nano Banana Pro is 4.5 credits per 2K image.
Make your first GPT Image 2 image
Write a prompt above or start from any example on this page. One subscription covers every model on Pixel Dojo.
Start with GPT Image 2