- Up to 4K
Max resolution
- Up to 16
Reference images
- 1:3 to 3:1
Aspect ratios
- Low–High
Quality modes
About GPT Image 2
GPT Image 2 is OpenAI's latest flagship image model, released in April 2026 as the third generation of the GPT Image line. It inherits the core architectural advantage that defines the series: a native multimodal system where the same model that understands language also generates the image, rather than a separate encoder-decoder pair. The practical payoff is unusually tight prompt adherence. GPT Image 2 tracks multi-part instructions, respects spatial relationships, and renders in-image text with a level of fidelity that rival models still struggle to match. For business creative teams, this means fewer iteration cycles on briefs that have specific layout or copy requirements, and a reliable editing workflow where targeted changes to one element don't cause the rest of the image to drift.
Beyond generation, GPT Image 2 supports genuine non-destructive editing with up to 16 reference images per call, so you can swap a background, adjust a product color, or refine a detail in an approved hero shot without starting over. Outputs generated through the OpenAI API (and through Masonry, which uses that API) are fully cleared for commercial use under OpenAI's standard usage policies, so you can move from generation directly to campaign deployment without additional licensing steps. At up to 3840×2160 resolution across a flexible 1:3 to 3:1 aspect-ratio range, GPT Image 2 delivers assets ready for digital, print, and out-of-home without an upscaling pass.
Prompts behind these GPT Image 2 images
Actual prompts from GPT Image 2 renders by the Masonry community. Copy any prompt, then remix it into your own creation.
A candid street-style lookbook photograph set in a graffiti-covered urban alley during golden hour — the kind of location found in Brooklyn, Shoreditch, Berlin Mitte, or Melbourne's Fitzroy district. Subject: a young woman in her early 20s with naturally textured loose hair, wearing an oversized vintage acid-wash denim jacket with hand-painted patches, a cropped white baby tee, high-waisted cargo pants in olive green, and chunky platform sneakers. Y2K accessories: butterfly clips, a mini hoop earring set, and a silver chain belt. Golden side-backlighting creates a warm halo effect around her silhouette, with the graffiti wall softly lit behind her in orange and amber tones. Shot on a 35mm film camera with Kodak Portra 400 emulation: slightly overexposed, warm tones, natural grain, soft shadows. Subject mid-stride, looking off-camera with effortless casual confidence. The overall mood is youthful, free-spirited, and universally urban. Vertical portrait orientation.
A photorealistic square 1:1 portrait of a street musician in his late 40s, seated on a worn wooden stool on a cobblestone pedestrian street in a European city on a bright autumn afternoon. He plays a battered acoustic guitar, eyes closed, completely lost in the music. He wears a faded olive corduroy jacket, dark jeans, and worn leather boots. Warm golden autumn light falls from the right side, catching the weathered wood grain of the guitar body, the silver fret markers, and the deep lines of his face simultaneously. A open guitar case lies in front of him on the cobblestones with a few scattered coins visible inside. Fallen amber and rust colored leaves dot the cobblestones around him. Shallow depth of field, the musician sharp against a beautifully blurred street background of soft bokeh shopfronts. Shot on an 85mm lens at f/1.4. Color grade: warm amber and rust autumn tones, rich shadows, gentle contrast. Mood: soulful, timeless, beautifully ordinary.
A tall 9:16 vertical portrait of a female face breaking the surface of completely black water, emerging from below, the face angled upward, mouth open taking the first breath. The water surface parts around the forehead, nose, and chin in sheets of black water that catch the single overhead spotlight as brilliant silver curtains falling away from the face. The eyes are open and looking directly upward into the lens — the first thing seen after breaking the surface is the camera. Water streams from every surface of the face in multiple rivulets, each one catching the overhead light. The hair fans out in the black water on both sides of the face at the surface level. Below the water surface the neck and shoulders are visible as dark shapes in the black depths. Shot on a 50mm lens at f/2.8. Color grade: absolute black water, the face brilliantly lit by a single overhead spot, silver water sheets cascading away. Mood: rebirth, the primal act of surfacing, the face as the entire world.
Extreme close-up macro photograph of a ferrofluid spike formation at 4:1 magnification, a single perfect conical spike of magnetically organized black ferrofluid rising from a flat liquid surface, frozen in its impossible form by the magnetic field below. The spike surface is not smooth — at this magnification the ferrofluid reveals a complex texture of secondary micro-spikes and surface ripples running down from the primary spike tip to the base, each tiny structure following the same magnetic field line geometry. The tip of the spike is razor sharp, the finest point physically possible for a liquid surface. The flat ferrofluid surface surrounding the spike base is mirror-perfect, reflecting the spike and the studio light as a perfect inverted image below. Background: pure white seamless. Single overhead softbox. Shot on a 100mm macro lens at f/8, focus-stacked. Color grade: absolute matte black ferrofluid against pure white background, the spike reflection in the flat surface as a perfect symmetrical composition. Mood: physics as sculpture, the invisible force field made tangible.
A wide 16:9 cinematic portrait of two strangers on opposite sides of a train window at night. Inside the train: a young woman, sharp and warm, lit by the soft amber carriage lighting, looking out the window at something beyond the frame. Outside in the darkness: the faint ghostly reflection of a man standing on the platform, visible only as a transparent overlay on the window glass — his image superimposed over her face and the carriage interior behind her, present but completely unreachable. The outside world is pure black except for his reflection. The condensation on the glass adds a slight soft diffusion to his reflected image. Neither person is looking at the other. Shot on a 50mm lens at f/2.0. Color grade: warm amber interior light on the woman, the man as a cool blue-gray reflection ghost, the window glass as the barrier between two separate worlds. Mood: the profound loneliness and accidental poetry of strangers sharing the same space without connection.
A photorealistic 9:16 vertical portrait of a young female florist, mid 20s, standing in her small flower shop surrounded by towering buckets of fresh blooms in every color. She holds a half-finished bouquet of white peonies and eucalyptus in both hands, eyes looking downward at her work with total concentration. She wears a moss green linen apron over a white long-sleeve shirt, a few flower stems tucked into the apron pocket. Soft natural light floods in from a large shop window behind her, creating a beautiful backlit rim of light around her hair and shoulders. Petals and scattered leaves litter the wooden workbench in front of her. Shot on a 50mm lens at f/1.8. Color grade: warm natural daylight, lush greens and soft petal colors, clean whites. Mood: craft, calm, quietly beautiful.
A wide 16:9 cinematic jewelry fashion editorial photograph of a female model's hands center frame, holding a single long stemmed black rose. She wears the Bulgari Serpenti Viper ring in 18k rose gold with full pavé diamonds coiling around her index finger, the snake head set with two emerald eyes catching the light at the highest point of the coil. Dramatic single-source side lighting from the left creates deep shadow on the right side of both hands, making the diamonds on the ring blaze against the darkness. The black rose petals frame the ring from above and below. Background: seamless deep black. Shot on an 85mm lens at f/2.8. Color grade: rose gold warmth, absolute black background, emerald green eye accents, the diamond pavé blazing white against the darkness. Mood: dark luxury, dramatic, editorial high jewelry campaign.
A wide 16:9 extreme macro photograph of a modern smartphone circuit board at 15:1 magnification, the microscopic components filling the frame as an alien city. Microchip packages rise like glass and black buildings from the green PCB substrate, their surfaces covered in microscopic bond wires — hair-thin gold wires arcing between chip pads and circuit board contacts, each one a perfect parabolic curve. The circuit traces running across the PCB are visible as copper highways, their surfaces showing the crystalline grain structure of the metal under the solder mask. Tiny ceramic capacitors and resistors sit in perfect rows like city blocks. A single solder joint is visible in macro detail — a perfect convex dome of tin-silver alloy with its crystalline surface structure revealed by the raking light. Single oblique fiber optic light from the left at 10 degrees. Color grade: deep green PCB, copper trace gold, black chip packages, the gold bond wires as the most brilliant elements. Mood: the hidden city inside every device, human ingenuity at its most miniaturized.
A square 1:1 dramatic close-up portrait of a face, cropped from chin to forehead, lit by a single venetian blind throwing a perfect ladder of light and shadow stripes across the entire face. The horizontal shadow bars divide the face into alternating bands of brilliant warm light and deep cool shadow, cutting across the nose, lips, cheekbones, and forehead with geometric precision. The eyes fall exactly in a light bar — both irises brilliantly illuminated, the pupil catchlights sharp. The lips fall in a shadow bar — completely dark. The entire portrait reads simultaneously as a human face and as a pure abstract graphic composition of alternating horizontal bands. The subject is completely still, expression neutral, letting the light do all the work. Shot on an 85mm lens at f/2.0. Color grade: warm afternoon sunlight bars, deep cool shadow bars, the contrast between them as sharp as a razor. Mood: noir, graphic, a portrait that is equally a piece of abstract art.
A square 1:1 extreme close-up portrait of a sleeping newborn baby, shot from above, the tiny face filling the entire frame. The baby sleeps in absolute peace — rosebud lips slightly parted, impossibly smooth skin catching soft directional light from the left, the faintest blue veins visible beneath the translucent skin of the eyelids. Downy fine hair across the forehead and temples catches the light as a soft golden halo. The ear is visible in perfect detail — the tiny helix, tragus, and earlobe forming a miniature architectural structure. A single adult thumb rests beside the cheek for scale, the size contrast between newborn skin and adult skin telling the entire emotional story. Shot on a 100mm macro lens at f/2.8. Lighting: one large diffused softbox from the left, a small silver reflector on the right for gentle fill. Color grade: warm ivory skin tones, soft golden hair, cool white background. Mood: tender, miraculous, the overwhelming fragility of new life.
A photorealistic tall 9:16 vertical portrait of an elderly male watchmaker in his late 60s, seated at his workbench in a small cluttered repair shop, shot from waist to just above the forehead. He holds a partially disassembled pocket watch movement in his left hand, a watchmaker's loupe screwed into his right eye, leaning forward in total concentration. His hands are the real subject — deeply lined, impossibly steady, holding the delicate movement with complete confidence built from decades of practice. The workbench surface around him is covered in tiny gears, springs, and tools arranged in a precise private order only he understands. A single warm desk lamp from the right illuminates his hands and the watch movement in a tight pool of amber light, leaving the edges of the frame in cool shadow. Shot on an 85mm lens at f/1.8, the watch movement and his eye razor sharp. Color grade: warm amber bench light, cool surrounding shadow, deep tool steel tones. Mood: craft, precision, a lifetime of accumulated skill visible in a single moment.
A photorealistic 16:9 wide shot inside a small independent specialty coffee shop at 7am on a weekday morning. The cafe is nearly empty — just one customer sitting alone at a window seat, hands wrapped around a ceramic mug, face softly lit by the pale blue morning light coming through the glass. The barista behind the counter is mid-pour on a latte, stream of milk caught frozen in motion above the cup. The entire interior glows with a warm contrast of the amber Edison bulb lights above the counter against the cold blue daylight flooding through the front window. Exposed brick walls, handwritten chalk menu boards, a glass pastry case with croissants. Steam rises from the espresso machine. Shot on a 24mm lens at f/4, everything sharp from foreground counter to the window seat. Color grade: warm amber interior light fighting cool morning blue window light. Mood: quiet, unhurried, deeply atmospheric slice of everyday life.
A tall 9:16 vertical luxury product photograph of a Hermès Birkin 30 in Togo leather, Etain gray colorway, placed upright on a smooth cream travertine stone shelf mounted against a raw white plaster wall. The structured bag stands perfectly centered in the frame, its palladium hardware — turn-lock, clochette, and lock — each catching a soft directional light from the upper left at slightly different angles, creating individual highlights on every metal element. The iconic double stitching along every seam is visible at full zoom. The Hermès Paris stamp on the front hardware is clearly legible. A single dried pampas grass stem leans casually against the right side of the bag. Shot on an 85mm lens at f/4. Color grade: warm travertine stone, cool plaster wall, the gray leather as the tonal hero. Mood: old money, architectural, gallery-level luxury product photography.
A wide cinematic 16:9 fashion campaign photograph shot inside a grand empty train station with vaulted iron and glass ceiling. A female model in her late 20s walks alone across the vast marble floor directly toward camera, small against the enormous architecture. She wears a floor-length Saint Laurent black velvet evening gown with a deep V neckline and long sleeves, trailing slightly behind her. Her hair is slicked back severely. She carries nothing. The station is empty and completely still. Dramatic low winter sunlight streams through the glass ceiling panels above, creating long shafts of light and shadow across the marble floor. Shot on a 24mm wide lens at f/8. Color grade: cold steel and marble tones, the black velvet as the darkest element. Mood: power, solitude, cinematic fashion.
A photorealistic 1:1 square portrait of a weathered deep-sea fisherman in his late 50s, sitting on the edge of a worn wooden dock at dawn. Thick calloused hands rest on his knees, rope burns and sun damage visible across every knuckle. He wears a faded navy fisherman knit sweater and worn rubber waders. Soft pale blue pre-dawn light comes entirely from the horizon behind him, silhouetting his broad shoulders and illuminating the edges of his beard in cold silver light. Fishing nets hang loosely behind him. Expression: distant, tired, deeply content. Shot on a 85mm f/1.4 lens, razor-thin depth of field, face sharp against blurred dock and sea behind. Color grade: cold blue dawn light, desaturated, high detail in the skin and knit texture. Mood: raw, timeless, deeply human.
Why teams choose GPT Image 2
GPT Image 2 is the model to reach for when accuracy is the brief: the right text in the right place, an element changed without the rest of the image shifting, a complex layout that follows every line of the spec. Its native multimodal architecture means prompt adherence is structurally better than models that treat language and image generation as separate steps. For marketing and brand teams running high-stakes campaigns where a misrendered headline or off-brand crop has real cost, that reliability is the differentiator. Inside Masonry, GPT Image 2 sits alongside 50+ other models. Use it where precision matters, then hand off to others where style or speed are the priority.
What GPT Image 2 can do
The capabilities that set GPT Image 2 apart and earn its place in a brief
Strong Instruction Following
Parses complex, multi-part prompts and places each element where you asked. Layouts, spatial relationships, and per-element styling hold reliably across generations.
Accurate In-Image Text
Renders headlines, labels, packaging copy, and signage with high legibility. Dense text and small lettering remain measurably stronger than most competing models.
Precise Non-Destructive Editing
Change a specific element (background, product color, headline) without the rest of the image drifting. Accepts up to 16 reference images per call for complex multi-source compositions.
Native Multimodal Architecture
One model handles both language understanding and image generation, which is why prompt adherence is consistently tighter than systems that bolt a language encoder onto a separate image generator.
Flexible Resolution and Aspect Ratios
Outputs from square to cinematic wide (1:3 to 3:1) at up to 3840×2160, ready for digital, print, OOH, and social without a separate upscaling step.
Commercial-Ready Outputs
Images generated via the OpenAI API (including through Masonry) are cleared for commercial use under OpenAI's standard usage policy. No additional licensing steps before campaign deployment.
Where teams reach for GPT Image 2
- Ad creatives and social posts with on-image copy that must be spelled and positioned correctly
- Product and packaging mockups where label text and placement are part of the brief
- Iterative, non-destructive edits on approved hero shots that let you swap backgrounds, recolor products, and refine details without re-rolling
- Multi-element compositions that reference several brand assets in a single call
- E-commerce imagery where consistent product presentation across SKUs matters
- Copy-heavy promotional banners and retail assets where text accuracy is non-negotiable
- Brand campaign concepting where a detailed written brief needs to translate faithfully into visuals
- Print-ready assets at high resolution without a separate upscaling workflow
What sets GPT Image 2 apart
The strengths teams reach for, shown on real renders.
Precise Instruction Following
Parses complex, multi-part briefs and places every element exactly where specified, with headlines in position, product in frame, and backgrounds on cue. Ideal for layout-driven ad creative and branded content.
Accurate In-Image Text
Renders headlines, labels, and packaging copy with crisp legibility, a consistent strength of the GPT Image line and essential for copy-heavy marketing assets.
Non-Destructive Editing with Up to 16 References
Swap backgrounds, refine details, and iterate on approved concepts without regenerating from scratch. Supports up to 16 reference images per edit for precise, consistent revisions.
Explore related categories
Browse adjacent categories and creative directions teams are exploring
Frequently asked questions
What teams need to know about creating with GPT Image 2 in Masonry
Can I use GPT Image 2 outputs commercially?
Yes. Images generated through the OpenAI API (including through Masonry) are cleared for commercial use under OpenAI's standard usage policies. You can move from generation directly to ad trafficking, print production, or digital publishing without additional licensing. If distributing in the EU, you may need to surface the embedded C2PA metadata at the point of publication, but that is a regulatory requirement, not an OpenAI restriction.
What resolution does GPT Image 2 output?
GPT Image 2 supports up to 3840×2160 (roughly 8 megapixels) across a flexible aspect-ratio range from 1:3 to 3:1. You can also specify exact pixel dimensions such as 1536×1024 for a banner or social post. This makes it usable for digital, print, and out-of-home assets without a separate upscaling step in most workflows.
How does GPT Image 2 handle in-image text compared to other models?
It is consistently one of the strongest models for in-image text. The native multimodal architecture means it "understands" text as part of the image generation process rather than treating it as a post-hoc overlay. Dense text, small labels, and multi-word headlines render with high legibility. In comparative tests, GPT Image 2 outperforms FLUX and Midjourney on dense text and complex typographic layouts, though FLUX.2 [flex] closes the gap on structured typography.
How many reference images can I provide for editing?
Up to 16 reference images per call. This makes GPT Image 2 one of the most flexible models for multi-source compositions. You can supply a product shot, a background reference, a style board, and additional brand assets all in a single request and ask the model to synthesize them into one coherent image.
How does GPT Image 2 compare to Midjourney for marketing creative?
The two have different strengths. Midjourney v8 leads on raw aesthetic quality and stylized output; GPT Image 2 leads on prompt accuracy, text rendering, and editing. If your brief is "make something beautiful in a loosely defined aesthetic," Midjourney is often faster. If your brief is "this text, in this position, on this product, with this background change," GPT Image 2 is more reliable and has a proper API that integrates into automated pipelines.
How does GPT Image 2 compare to FLUX models?
GPT Image 2 and FLUX models have complementary strengths. GPT Image 2 is stronger on prompt adherence, text rendering, and non-destructive editing. FLUX.2 Pro and Max lead on photorealism and film-quality aesthetics. FLUX.2 Dev gives you open weights for self-hosting and fine-tuning. In Masonry, you can use GPT Image 2 for layout-precise creative and FLUX models where photorealism or style is the priority. You are not locked to one.
Does GPT Image 2 support inpainting and targeted edits?
Yes. GPT Image 2 supports targeted editing. You can describe what to change and the model will modify that element while preserving the rest of the image. This is more reliable than "remix" style editing found in some other tools, where changing one element causes unpredictable drift in surrounding areas. For asset-intensive workflows, generating a clean base image and editing in variants is often faster than re-generating from scratch each time.
What output formats does GPT Image 2 produce?
GPT Image 2 outputs standard raster images (PNG and JPEG) suitable for direct use in ad platforms, CMS uploads, and print workflows. Because Masonry connects to the OpenAI API, outputs flow directly into your Masonry workspace for further editing, annotation, or handoff to the rest of your creative pipeline.
Is GPT Image 2 good for product photography and e-commerce imagery?
Yes, particularly for structured product shots where placement, lighting direction, and background are specified in the prompt. It handles multi-SKU workflows well. Generate a clean product base and use editing to spin out background or colorway variants rather than running a full generation for each. For pure photorealistic product photography with complex surface materials, FLUX.2 Flex or Pro may produce sharper detail, but GPT Image 2's editing precision often makes it faster end-to-end.
How long does GPT Image 2 take to generate an image?
GPT Image 2 takes longer than faster models like FLUX Schnell or Nano Banana 2. The model does additional reasoning before generating, which contributes to better prompt accuracy but adds a few seconds of latency. For high-volume batch workflows where speed is the priority, faster models are a better fit. For considered, high-stakes creative where a few extra seconds in exchange for higher accuracy saves rework time, GPT Image 2's generation speed is reasonable.
Can GPT Image 2 generate images with multiple distinct products or brand elements in one shot?
Yes, and this is one of its clearest differentiators. Its long, detailed prompt handling and multi-reference support mean you can describe complex scenes with several products, props, and environmental elements and have each placed correctly in the frame. Teams running lifestyle or flat-lay creative with multiple SKUs in a single shot find this particularly useful.
What is GPT Image 2?
GPT Image 2 is an AI image generation model from OpenAI, available inside Masonry, the AI creative agent teams use to produce marketing, product, and brand images.
How does my team use GPT Image 2 in Masonry?
Open a Masonry canvas, pick GPT Image 2 from the model selector, and describe the image you need: a product shot, an ad creative, a social post. Masonry generates it, then you refine, edit, and combine GPT Image 2 with other models in one workspace.
Is GPT Image 2 free to try?
Yes, you can start generating images with GPT Image 2 on Masonry's free tier, then scale up with higher limits and priority processing as your team grows.
How do I write good prompts for GPT Image 2?
GPT Image 2 follows detailed instructions well, so be explicit. Describe each element, where it sits, and the exact text you want, then use editing to refine rather than re-rolling from scratch. See the prompt gallery on this page for real GPT Image 2 prompts you can copy and adapt.
Who makes GPT Image 2?
GPT Image 2 is built by OpenAI. Inside Masonry it runs alongside 50+ image and video models, so your team can pick the right one for each brief without switching tools.
Can I see examples made with GPT Image 2?
Yes, the prompt gallery on this page shows real images teams have generated with GPT Image 2 in Masonry, each paired with the exact prompt you can copy and adapt for your own brand.
Start creating with GPT Image 2
Generate, edit, and compare across 50+ models in one workspace.
Guides for GPT Image 2
Prompt walkthroughs and examples from the Masonry blog
Explore more AI models
Compare GPT Image 2 with other models teams run in Masonry