The MAI Image 2.5 prompt generator gives you 20 free, copy-ready prompts for Microsoft's newest AI image model — announced June 2, 2026. Covers text-to-image generation, localized image editing, typography, portraits, product photography, and more.
The MAI Image 2.5 prompt generator on this page provides 20 free, professionally written prompts for MAI-Image-2.5 — Microsoft's latest AI image model, unveiled at Microsoft Build on June 2, 2026. MAI Image 2.5 represents a significant leap over its predecessor: it debuts at #3 on the Arena text-to-image leaderboard (score 1,254 — up 72 points from MAI Image 2.0) and reaches #2 on the image editing leaderboard, a capability entirely new to this model line.
The defining upgrade in MAI Image 2.5 is precise, localized image editing. Unlike text-to-image generation, editing means you provide an existing image and describe targeted changes — replace an object, update text on a label, change a wall colour, or shift the time of day — and the model makes those changes while preserving everything else. MAI Image 2.5 understands scene structure, lighting, scale, and spatial relationships, so edits fit naturally into the image context. It also preserves facial identity across edits, making it reliable for portrait retouching and headshot batching.
On text-to-image, MAI Image 2.5 improves on MAI Image 2.0's already strong photorealism with better prompt adherence on complex, multi-element compositions and more accurate in-image text rendering. It is available as both the full MAI-Image-2.5 model and a lighter MAI-Image-2.5-Flash variant for high-volume API use, integrated into Microsoft PowerPoint for presentation visuals, and rolling out to Bing Image Creator and Copilot.
MAI Image 2.5 handles two distinct prompt types — generation and editing. Each requires a different approach:
Click any prompt to copy — paste into Microsoft AI Foundry, Bing Image Creator, Copilot, or PowerPoint
Edit this portrait: remove the cluttered shelving behind the subject and replace it with a clean, softly blurred neutral grey studio background. Preserve the subject's face, hair, clothing, and the existing window light falling on their left cheek. Do not alter skin tone, expression, or facial features. The result should look like a professional headshot session, not a composite.
Edit the label on the glass bottle in this image: replace the existing text with 'PRESTIGE BOTANICS — Cold Press Argan Oil — 30ml'. Keep the same label shape, colour palette, and typeface style. The new text must be fully legible and correctly spelled. Preserve the bottle shape, lighting, and background exactly as they are.
A photorealistic portrait of a woman in her early 30s, sitting by a large north-facing window on an overcast day. Flat, even, naturally diffused light across her face — no shadows. She wears a cream knit sweater, her expression is calm and direct. Shot at 85mm equivalent, shallow depth of field, background blurred to creamy neutrals. Photorealistic, editorial portraiture.
A traditional Parisian bakery storefront, narrow frontage on a cobblestone street. The shop fascia reads 'BOULANGERIE LECONTE — Artisan depuis 1932' in hand-lettered gilt serif on dark green. Below the awning: a chalkboard sign reads 'Pain au levain — 3,50€'. Morning light, slightly misty. All text is fully legible. Photorealistic street photography.
Edit this architectural exterior photograph: change the sky and ambient light from midday to dusk — warm amber tones at the horizon, deep blue above, interior lights glowing through windows. Ensure the building's materials, shadows, and reflections are updated consistently with a dusk light scenario. Keep the building structure, composition, and foreground landscaping unchanged.
A luxury wristwatch lying on a piece of brushed anthracite steel, photographed at a 30-degree angle. A single narrow LED strip light from above creates a razor-thin reflection across the case and bracelet links. The dial reads '12:10' and the text 'AUTOMATIQUE — SWISS MADE' is clearly legible at the dial's lower section. Black background fading to dark grey. Luxury commercial photography. Photorealistic.
A music festival poster, portrait orientation, dark blue background. Large condensed sans-serif headline: 'COASTLINE FESTIVAL 2026' in white, all caps, centred at the top. Below: 'July 18–20 · Lisbon, Portugal' in a medium weight. A minimal graphic of a cresting wave in pale cyan sits in the centre. Bottom section: 'coastlinefest.com · Tickets from €45' in small text. All text fully legible. Clean graphic design.
Edit this interior photograph: repaint all the walls from their current beige to a deep forest green (similar to Farrow & Ball 'Calke Green'). Preserve the furniture, flooring, trim colour, artwork, lighting, and shadows exactly as they appear. The green should look like genuine painted walls in the existing light conditions — not a flat overlay. Photorealistic interior design render.
A high-contrast fashion editorial shot: a male model, late 20s, wearing a structured black leather jacket over a white t-shirt. Single hard key light from 45 degrees above left — crisp shadows on the right side of his face and neck. The background is pure black. His posture is sharp, slightly turned. The image has the hard contrast of a Helmut Newton editorial. Photorealistic.
Fine dining plated dish from directly overhead on a matte black ceramic plate: duck breast sliced in a fan, five slices arranged over a swipe of parsnip purée, three dots of cherry jus, micro herbs and edible flowers scattered. The plate sits on a dark oak table. Light source: single soft overhead diffused box, slightly warm. Food editorial photography style. Photorealistic.
Edit this portrait: change the subject's expression from neutral and serious to a natural, relaxed smile — teeth slightly visible, eyes crinkling slightly at the corners. Preserve the subject's exact facial identity, hair, skin tone, clothing, lighting, and background. The smile must look natural and unforced, as if captured mid-laugh. Do not alter anything else in the image.
An exterior real estate twilight photograph of a contemporary two-storey house: warm interior lighting glowing through floor-to-ceiling windows, exterior up-lights illuminating the textured stone facade, a deep blue sky above with the last traces of sunset at the horizon. The pool in the foreground reflects the building. No people. Wide-angle architectural photography, photorealistic.
A clean, professional presentation slide on white background for PowerPoint. Title: 'AI Adoption in Enterprise 2026' in bold dark sans-serif at the top. Three illustrated stat cards below in a row — '87% of companies use AI', '4.2× productivity gain', '$2.1T annual value'. Each card has a simple icon above the number and a short descriptor below. Blue and teal colour palette. Clear, legible, business presentation style.
A macro photograph of a hoverfly perched on the centre of a white daisy. Shot at 1:1 macro ratio — the compound eyes, wing venation, and leg hairs are rendered in sharp detail. The white petals surround the fly, softly blurred. Overcast diffused light, no harsh shadows. Scientific natural history photography aesthetic. Photorealistic.
Tokyo Metro station platform at 8 AM: a woman in a dark office coat stands reading her phone, blurred commuters rushing past her in both directions at 1/15s. Fluorescent overhead lighting, the station name board in Japanese characters visible behind her. Slight motion blur on commuters, subject sharp. Documentary street photography. Photorealistic.
Edit this product photograph: remove the white studio background and replace it with a natural marble surface from directly below and a soft, slightly warm grey gradient behind. Keep all shadows under the product consistent with the new surface — add soft natural contact shadows. Preserve the product shape, colour, and all labelling exactly as photographed. Photorealistic beauty product shot.
A children's book illustration in a warm, painterly style: a small fox wearing a red scarf sits on a snow-covered log in a birch forest at dusk. The fox looks up at a lantern hanging from a branch, glowing warm amber. Snowflakes fall gently. The style is reminiscent of Beatrix Potter meets Studio Ghibli — soft, textured, inviting. The scene is peaceful and magical. No text.
A professional corporate headshot, shot with consistent parameters to match an existing team: the subject is a man in his late 30s wearing a mid-blue shirt. He faces the camera directly with a confident, approachable expression. Studio background: soft light grey, evenly lit. Key light from camera-left, gentle fill from camera-right. 85mm equivalent. Photorealistic corporate photography.
A luxury hotel exterior at night: a classic European Belle Époque building on a wide boulevard, the facade illuminated by warm external up-lights, the entrance canopy with gold lettering reading 'GRAND HÔTEL RIVIERA'. A black car is parked at the entrance, a doorman in dark livery stands at the door. The street is wet from recent rain. Wide-angle architectural nighttime photography. Photorealistic.
A square social media advertisement image for a skincare brand. Background: a clean, pale sage green with a subtle linen texture. Centre: a single glass dropper bottle, contents a pale gold serum, label reads 'BOTANICAL RADIANCE SERUM'. Below the bottle in clean serif typography: 'Your skin. Restored.' A small sprig of dried chamomile is placed to the left of the bottle. The text is fully legible. Photorealistic commercial photography, beauty brand aesthetic.
How Microsoft's MAI Image 2.5 compares to the current top-ranked image generators and editors:
| Model | T2I Rank | Editing | Best For |
|---|---|---|---|
| GPT Image 2 (OpenAI) | #1 | Limited | Product photography, versatility, UI mockups |
| MAI Image 2.5 (Microsoft) ★ | #3 (1,254) | #2 editing | Editing, portraits, multi-element prompts, PowerPoint |
| Nano Banana Pro (Google) | Top 5 | Good | Typography, 4K resolution, multi-reference images |
| Imagen 4 (Google) | Top 5 | Moderate | Photorealism, architectural visualization, product |
| MAI Image 2.0 (Microsoft) | #3 (1,182) | None | Bing/Copilot baseline, portrait photorealism |
| FLUX.2 (Black Forest Labs) | Top 10 | Via API | Open-weight, 4MP native, 10-reference composition |
★ MAI Image 2.5 uniquely combines strong generation (#3) with best-in-class image editing (#2). All outputs include C2PA Content Credentials. Scores from Arena.ai leaderboard, June 2026.
The MAI Image 2.5 prompt generator on this page gives you 20 free, copy-paste prompts for MAI-Image-2.5 — Microsoft's latest AI image model, announced at Microsoft Build on June 2, 2026. MAI Image 2.5 ranks #3 on the Arena text-to-image leaderboard (score 1,254 — 72 points higher than MAI Image 2.0) and debuted at #2 on the image editing leaderboard. It is available via the Microsoft AI Foundry API and a lighter MAI-Image-2.5-Flash variant for high-volume applications.
MAI Image 2.5 introduces three major upgrades over MAI Image 2.0: (1) Image editing — 2.5 supports precise, localized edits on existing images, including object removal/replacement, text updates, lighting changes, and expression changes, while preserving facial identity and scene context. This was not available in 2.0. (2) Improved text rendering — MAI Image 2.5 handles complex, multi-element typographic prompts with higher accuracy and legibility than 2.0. (3) Better prompt adherence on multi-element compositions — when you specify precise spatial relationships, styles, and detail layers, 2.5 follows them more accurately. The Arena score jumped 72 points (+6.1%) from 2.0 to 2.5.
For editing prompts, MAI Image 2.5 needs three things: (1) Specify what to change precisely — name the object, region, or property to edit, not the whole image. 'Change the wall colour to forest green' works better than 'edit this photo'. (2) Specify what to preserve — explicitly state what should not change: 'Keep the furniture, lighting, and flooring exactly as they are'. (3) Describe the result in terms of real-world reference — 'the wall should look like freshly painted walls in the existing light conditions' gives better results than 'make it look good'. MAI Image 2.5 understands scene structure and will make contextually consistent edits when you describe them in concrete terms.
MAI Image 2.5 is available in three main ways: (1) Microsoft AI Foundry — the primary API access for developers, available as both MAI-Image-2.5 (full quality) and MAI-Image-2.5-Flash (faster, cheaper); (2) Microsoft PowerPoint — MAI Image 2.5 powers the 'Generate visuals from prompts' feature for creating presentation graphics and slides directly in PowerPoint; (3) Microsoft Copilot and Bing Image Creator — MAI Image 2.5 is expected to roll out as the backend for Copilot image generation. Compared to MAI Image 2.0, which was already in Bing and Copilot, the 2.5 update adds the editing capabilities and improved prompt adherence.
As of June 2026, the three top AI image models each lead in different areas: GPT Image 2 (OpenAI) ranks #1 overall and leads on versatility, product photography, and complex scene composition. Nano Banana Pro (Google) leads on in-image text accuracy and native 4K resolution. MAI Image 2.5 (Microsoft) ranks #3 on generation and #2 on editing, and uniquely leads on image editing capabilities — the ability to make localized, context-aware changes to existing images. MAI Image 2.5 is the strongest choice for workflows that involve both generating and editing images in the same session.
MAI-Image-2.5-Flash is a smaller, optimized variant of MAI Image 2.5 released simultaneously at Build 2026. It delivers faster generation and lower API costs — comparable to the relationship between MAI Image 2.0 and MAI Image 2.0-Efficient. Flash is ideal for high-volume workflows, rapid prototyping, and applications where speed matters more than peak visual quality. For final deliverables, professional use cases, and complex image editing, the full MAI-Image-2.5 model produces stronger results.
Microsoft's previous model — photorealistic portraits and natural light photography
OpenAI's #1 Arena-ranked image model — product photos and UI mockups
Google's Gemini 3 Pro image model — 4K, best text rendering, 14 reference images
Google's photorealism specialist — product photography and architecture
Black Forest Labs — 4MP native output, open-weight, 10 reference images
Generate prompts for any AI image model — aesthetic, fantasy, photography