/

GPT Image 1.5

by OpenAI
  • What it is:GPT Image 1.5 is a production-quality image generation model built into GPT-5 that creates high-fidelity images with precise editing, strong text rendering, and 4x faster generation speeds.
  • Best for:Content marketing teams, Product designers needing mockups, Developers building AI applications
  • Pricing:Starting from Included with ChatGPT Plus/Pro
  • Rating:88/100Very Good
  • Expert's conclusion:The First GPT Image 1.5 Model is the Leading Choice For Production Grade Image Generation and Editing Where Prompt Precision, Speed, and Text Handling Are Important.
Visit website
Reviewed byMaxim Manylov·Web3 Engineer & Serial Founder

What Are GPT Image 1.5's Key Business Metrics?

📊
5-8 seconds per generation (4× faster)
Image Generation Speed
📊
20% lower than predecessor
API Cost Reduction
📊
3 sizes (1024×1024, 1024×1536, 1536×1024)
Image Sizes Supported
📊
5+ edits before quality degradation
Iterative Edit Consistency
📊
December 17, 2025
Release Date
📊
15-45 seconds (varies by complexity)
Generation Speed Range

How Credible and Trustworthy Is GPT Image 1.5?

88/100
Excellent

GPT Image 1.5 is backed by OpenAI which has a high level of technical credibility as an AI company. It demonstrates a number of other technical improvements, it is also production ready with clear documentation and API support.

BREAKDOWN
Product Maturity85/100
Company Stability95/100
Security & Compliance85/100
Transparency90/100
Performance Claims88/100
User Accessibility85/100
TRUST SIGNALS
Backed by OpenAI, industry leader in AIBuilt on proven GPT-5 architectureC2PA content authentication supportPublicly documented technical improvements4× performance improvement independently verifiedReduced API costs (20% savings)

What Are the Key Features of GPT Image 1.5?

✨
Region-Aware Editing
Ability to modify specific parts of an image while maintaining non-modified aspects such as facial features, lighting and background composition.
⚡
4× Faster Generation
Generation of images from input in 5-8 seconds using architectural optimizations such as fewer sampling steps and efficient attention mechanism.
✨
Text Rendering
Ability to create readable text inside images such as small font sizes, dense paragraphs and complex layouts, such as infographics with correct spelling and kerning.
📊
Advanced Instruction Following
Ability to perform complex, multi-step instructions with high levels of systematic accuracy regarding composition, lighting, style and specific elements.
✨
Iterative Context Retention
Consistency across 5+ iterative edit rounds, remembers previous state and applies modification without loss of quality.
✨
Unified Multimodal Architecture
Uses the same neural network to process text and generate images through native multimodal architecture providing improved consistency.
✨
Multiple Output Sizes
Generates images in three aspect ratios; 1024 x 1024 (square) 1024 x 1536 (portrait) and 1536 x 1024 (landscape).
🔗
API Integration & Batch Processing
Can be integrated into systems through OpenAI’s API supporting quality tier options, size selections and batch generation with webhook callback notifications.
✨
C2PA Content Authentication
Images will contain C2PA metadata for authentication of content and verification of AI-generated content.

What Are the Best Use Cases for GPT Image 1.5?

E-commerce and Product Teams
Ability to rapidly generate professional product photography and marketing assets at scale and provide precise control over composition, lighting and background - reducing photography costs and speeding up asset generation for listings.
Content Creators and Designers
Ability to rapidly explore visual concepts using region-aware editing that maintains overall composition while allowing targeted changes - providing rapid creative workflow solutions and reduce revision cycle time.
Marketing and Advertising Teams
Develop a variety of campaign assets, that are created using a similar branding, lighting, and composition from one iteration to another without needing to fully recreate each one, allowing for rapid A/B testing, and optimizing your campaigns.
Business and Technical Documentation Teams
Create diagrams, infographics, presentations, etc., with clear text, proper formatting, and complex layouts that you can't produce manually.
Software Developers and API Integrators
Use image creation through OpenAI's API within an application by using batch processing, quality levels, and web hook calls to generate images automatically on a large scale.
NOT FORReal-time Image Editing Applications
Not Recommended — The time it takes for generation is 5-8 seconds or longer, therefore will never work with real-time interactive workflows which require responses in less than a second.
NOT FORTasks Requiring Photorealistic Historical Accuracy
Not Applicable — While capable of preserving details, it does not guarantee historical accuracy when used for archival, or educational purposes that require verification of factually correct representations.

How Much Does GPT Image 1.5 Cost and What Plans Are Available?

Pricing information with service tiers, costs, and details
☐Service$CostℹDetails🔗Source
Standard Image Generation (ChatGPT)Included with ChatGPT Plus/ProText-to-image and image editing available through ChatGPT interface for subscribers—
API - Standard Quality (1024×1024)$0.080 per imageStandard quality tier with base feature set—
API - Enhanced Quality20% discount vs. predecessorImproved quality output with optimized settings. Exact pricing not disclosed.—
API - Batch ProcessingVolume-based pricingGenerate multiple variations and batch requests with potential volume discounts—
Enterprise Custom QuoteCustom pricingFor high-volume usage, dedicated support, and custom integrations—
Standard Image Generation (ChatGPT)Included with ChatGPT Plus/Pro
Text-to-image and image editing available through ChatGPT interface for subscribers
API - Standard Quality (1024×1024)$0.080 per image
Standard quality tier with base feature set
API - Enhanced Quality20% discount vs. predecessor
Improved quality output with optimized settings. Exact pricing not disclosed.
API - Batch ProcessingVolume-based pricing
Generate multiple variations and batch requests with potential volume discounts
Enterprise Custom QuoteCustom pricing
For high-volume usage, dedicated support, and custom integrations

How Does GPT Image 1.5 Compare to Competitors?

FeatureGPT Image 1.5DALL-E 3MidjourneyStable Diffusion
Generation Speed5-8 seconds15-45 seconds30-60 seconds10-30 seconds
Region-Aware EditingYesLimitedLimitedPartial
Text Rendering QualityHigh (readable)ModerateModeratePoor
Iterative Edit Consistency5+ edits1-2 edits2-3 editsVariable
API AvailabilityYesYesLimited APIYes
Starting Price (API)$0.080/imageSimilar pricingWeb-based subscriptionFree/self-hosted
InterfaceChatGPT + APIChatGPT + APIWeb/DiscordSelf-hosted/third-party
Enterprise SSO/SAMLAvailableAvailableLimited
Content Authentication (C2PA)YesLimitedNoNo
Built on GPT ArchitectureYes (GPT-5)Yes (GPT-4)NoNo
Generation Speed
GPT Image 1.55-8 seconds
DALL-E 315-45 seconds
Midjourney30-60 seconds
Stable Diffusion10-30 seconds
Region-Aware Editing
GPT Image 1.5Yes
DALL-E 3Limited
MidjourneyLimited
Stable DiffusionPartial
Text Rendering Quality
GPT Image 1.5High (readable)
DALL-E 3Moderate
MidjourneyModerate
Stable DiffusionPoor
Iterative Edit Consistency
GPT Image 1.55+ edits
DALL-E 31-2 edits
Midjourney2-3 edits
Stable DiffusionVariable
API Availability
GPT Image 1.5Yes
DALL-E 3Yes
MidjourneyLimited API
Stable DiffusionYes
Starting Price (API)
GPT Image 1.5$0.080/image
DALL-E 3Similar pricing
MidjourneyWeb-based subscription
Stable DiffusionFree/self-hosted
Interface
GPT Image 1.5ChatGPT + API
DALL-E 3ChatGPT + API
MidjourneyWeb/Discord
Stable DiffusionSelf-hosted/third-party
Enterprise SSO/SAML
GPT Image 1.5Available
DALL-E 3Available
MidjourneyLimited
Stable Diffusion—
Content Authentication (C2PA)
GPT Image 1.5Yes
DALL-E 3Limited
MidjourneyNo
Stable DiffusionNo
Built on GPT Architecture
GPT Image 1.5Yes (GPT-5)
DALL-E 3Yes (GPT-4)
MidjourneyNo
Stable DiffusionNo

How Does GPT Image 1.5 Compare to Competitors?

vs Midjourney

XYZEO Analysis: GPT Image 1.5 is designed for use by General Users, Developers, and via ChatGPT, or API; It uses Strong Instruction Following and Editing; Midjourney is designed for Creative Professionals via Discord and has Superior Artistic Style Options. GPT Image 1.5 is faster 5-8 seconds, is Integrated, and has no Community Momentum or Style Diversity compared to Midjourney.

GPT Image 1.5 for exact control over generated text using a prompt; Midjourney for artistic freedom as well as to engage users in an open community.

vs Stable Diffusion (via ComfyUI/Flux)

XYZEO Analysis: GPT Image 1.5 is a Premium Hosted Service that is Easier to Use and More Consistent; Stable Diffusion is an Open Source Platform that allows for Customization and Local Control at a Lower Cost; OpenAI has a much Larger Market Share, but is at Risk of being Locked into their Ecosystem Compared to Stable Diffusions Customization and Growing Momentum.

GPT Image 1.5 for work-related projects that require multiple steps; Stable Diffusion for those who want to customize without the cost.

vs Google Imagen 3

XYZEO Analysis: Both are Premium Enterprise Models; GPT Image 1.5 is Better at Producing Consistency in Iterative Editing; Imagen 3 is Best at Producing Photorealism; OpenAI Has a Much Larger Developer Ecosystem; Google is Targeting Vertex AI Customers who Require Compliance Features in Their Enterprise Solutions.

GPT Image 1.5 for multi-step project development; Imagen 3 for high-compliance requirements related to creating photorealistic images within enterprises.

vs DALL-E 3

Comparison of predecessors - GPT Image 1.5 produces 4 times faster results (5-8 seconds) than GPT Image 1 (20-30 seconds); both versions are equal in terms of the ability to render text; also, GPT Image 1.5 supports area-specific editing. OpenAI is clearly moving toward version 1.5 of GPT Image internally based on performance metrics.

GPT Image 1.5 will replace DALL-E 3 for all possible applications.

What are the strengths and limitations of GPT Image 1.5?

Pros

  • 4 times faster results - 5-8 seconds to generate instead of 20-30 seconds to generate with older versions.
  • Better at following instructions - able to follow complex multi-step instructions correctly.
  • Supports region-aware editing - allows users to edit a portion of an image without changing other parts.
  • Great text rendering - creates readable, proper spelling, and grammar for images.
  • Will preserve identities - maintain lighting and facial expressions through 5+ iterations of an image.
  • Multiple output resolutions - 1024 x 1024 (square), 1024 x 1536 (portrait), 1536 x 1024 (landscape).
  • Includes native integration with ChatGPT - seamless workflow for conversational image creation.
  • Has an API for developers - has all of the necessary features to be used in production environments.

Cons

  • Is proprietary "black box" - does not allow for fine tuning or local installation.
  • Uncertainty about API costs - no information about costs, could lead to large bills from high usage.
  • Not specialized in styles - good at most styles, but lacks some of the unique styles of the best artistic models.
  • Still sequential - can't do real time collaborative editing.
  • Locks into the OpenAI ecosystem - requires an API key and access to their credit system.
  • Restrictive content policies - content safety filters may prevent valid prompts from being sent.
  • Does not support negative prompts - can't specifically tell it what not to include in the image.
  • Dependent on hardware for speed - 5-8 second results are optimal when running on optimized hardware.

Who Is GPT Image 1.5 Best For?

Best For

  • Content marketing teams — Enables fast iterative workflow for making precise changes to a company's brand assets.
  • Product designers needing mockups — The new version of GPT Image has a lot better text rendering and layout options compared to previous versions for presentation purposes.
  • Developers building AI applications — GPT Image 1.5 also has an API ready for production that consistently adheres to the prompt, and is ready to be used.
  • Social media creators — Multiple image sizes are available when you generate images using this tool, which allows for quicker production time.
  • E-commerce teams — When you need to show your products from different angles, this tool makes it easy to create consistent visualizations and variant generation.

Not Suitable For

  • Fine artists seeking unique styles — Because Midjourney's Discord community specializes in creating unique styles, and their user base is more likely to have what you're looking for, I would recommend going there first before trying to find it here.
  • Cost-sensitive hobbyists — If you don't want to pay for the API to use GPT Image, then you might want to look into one of the many open-source alternatives out there such as Stable Diffusion.
  • Real-time interactive design tools — If the 5-8 second delay after generating an image is too long for you, you should probably just use the AI plugins offered by Figma.
  • Highly regulated industries — Since OpenAI does enforce certain content restrictions on all of its services, if these restrictions will conflict with your company's own compliance standards, you might want to consider using Google Imagen.

Are There Usage Limits or Geographic Restrictions for GPT Image 1.5?

Image Sizes
1024×1024, 1024×1536 portrait, 1536×1024 landscape only
Generation Time
5-8 seconds standard, 15-45 seconds high quality
API Model ID
gpt-image-1.5 - pay per image generation
Content Policy
Strict safety filters block violence, nudity, celebrities, trademarks
Edit Consistency
Reliable for 5+ iterations, may degrade beyond
No Fine-tuning
Cannot train on custom datasets
Prompt Length
Complex prompts supported but token-limited by API
Commercial Rights
Allowed per OpenAI terms, attribution not required
Geographic Availability
Global except sanctioned countries (China, Russia, Iran restricted)

What APIs and Integrations Does GPT Image 1.5 Support?

API Type
REST API via OpenAI platform, model ID: gpt-image-1.5
Authentication
OpenAI API key required for all requests
Input Parameters
Text prompt + optional input image, quality/speed tiers, size selection
Output Format
Image URLs or base64 data, supports batch processing
Generation Speed
5-8 seconds standard (4x faster than predecessor), quality/latency tradeoffs
Documentation
OpenAI API reference + gpt-image-1.5 prompting guide with examples
SDK Support
Official OpenAI SDKs: Python, Node.js, compatible with all OpenAI clients
Rate Limits
Standard OpenAI tiered limits based on usage level
Use Cases
Text-to-image generation, inpainting/outpainting, iterative editing, commercial applications

What Are Common Questions About GPT Image 1.5?

GPT Image 1.5 produces images approximately four times faster 5-8 seconds vs 20-30 and has far better region-aware editing capabilities than DALL-E 3 that preserve original areas and provides much better text rendering. With five or more iterations, GPT Image 1.5 tends to maintain consistency, whereas DALL-E 3 tends to fail.

As far as I know, GPT Image currently offers three standard resolutions: 1024 x 1024 square, 1024 x 1536 portrait, and 1536 x 1024 landscape. Currently, no custom resolutions are available.

Yes, GPT Image 1.5 does allow for region-aware editing and inpainting, so you can make modifications to specific parts of an image, but at the same time preserve facial expressions, lighting, composition, and other details in the surrounding area across multiple iterations.

Depending upon the quality level of the output, standard quality will take around 5-8 seconds to produce, high-quality or complex prompts will typically take 15-45 seconds. This is 4x faster than the previous OpenAI image models.

Yes, as model gpt-image-1.5, GPT Image is available through the OpenAI API. In addition to returning the URL of the generated image, it will also return the base64 encoded image data, and will support both batch processing and quality levels.

Yes, GPT Image 1.5 is a significant improvement over the previous versions of GPT Image. It handles very small amounts of text, full paragraphs, proper spelling, correct alignment, and even complex layouts like infographics.

According to OpenAI's website, legitimate commercial use of GPT Image is allowed under two conditions: proper use of language, and proper use of prompting.

For OpenAI Plus users, GPT Image is available in the ChatGPT interface, and for developers, it is available via the OpenAI API as model identifier gpt-image-1.5.

Is GPT Image 1.5 Worth It?

XYZEO Analysis of GPT Image 1.5 The first GPT Image 1.5 Model is a Flagship Text-to-Image Model and an Image Editing Model from OpenAI that has the capability to adhere to prompts extremely well, edit regions precisely (if you want), render readable text, and create photorealistic images. It creates ChatGPT Images, and it can be accessed through an API. The new model is able to generate images 4 times quicker than its predecessors, and it does so at a cost 20% less expensive than its predecessors. Because of this, the model is ready for use in production environments for both structured and unstructured workflows for creatives. While it is a very capable model, it is still a developing model and is therefore subject to the same evolving limitations of AI that exist today such as occasionally generating inconsistent images in complex scenes.

Recommended For

  • Developers & Designers Creating Production Apps That Need Fast, Precise Image Generation & Editing
  • Marketing & Content Teams Creating Infographics, UI Mockups, Posters & Promotional Visuals
  • E-commerce Businesses Requiring Product Mockups, Virtual Try-Ons, Style Transfers
  • Creative Professionals Iterating Photorealistic Images With Multi-Step Refinements

!
Use With Caution

  • Enterprise Users In Regulated Industries — Verify Compliance For Use Of Generated Content
  • High-Volume Applications — Test Latency And Cost At Scale Even Though There Has Been Improvement
  • Users Requiring Ultra-High Resolution Beyond 1536×1024 — Currently Limited Size Options Available

Not Recommended For

  • Budget-Constrained Hobbyists — API Costs Will Add Up Even Though There Are Improvements
  • Teams Requiring Fully Open-Sourced Models — Proprietary OpenAI Technology
  • Applications Demanding Perfect Consistency In Highly Complex Scenes With Multiple Objects
Expert's Conclusion

The First GPT Image 1.5 Model is the Leading Choice For Production Grade Image Generation and Editing Where Prompt Precision, Speed, and Text Handling Are Important.

Best For
Developers & Designers Creating Production Apps That Need Fast, Precise Image Generation & EditingMarketing & Content Teams Creating Infographics, UI Mockups, Posters & Promotional VisualsE-commerce Businesses Requiring Product Mockups, Virtual Try-Ons, Style Transfers

What do expert reviews and research say about GPT Image 1.5?

Key Findings

GPT Image 1.5 is an advanced version of OpenAI's image model that is significantly improved over prior models in respect to: • Prompt following (ability to follow user prompts) • Region aware editing (ability to edit a specific portion of an image) • Legibility of text rendered into images (ability to create readable text) • 4x faster than prior versions of the model for generation speed. The model has been fully integrated into ChatGPT and the OpenAI API and can be utilized in a variety of sizes and styles to produce high levels of photorealism, structured visualization, and iterative workflow processes. The technical architecture utilizes a unified multimodal autoregressive predictive technique to maintain consistency across both text and visual media.

Data Quality

Good - comprehensive details from OpenAI developer docs, API references, and third-party hosting platforms like Replicate and Runware. Official prompting guides and capability breakdowns available. Pricing and exact training data not fully public.

Risk Factors

!
A proprietary model reliant upon the availability of the OpenAI API and subject to the terms of such service.
!
There are potential content policy constraints related to the creation and use of generated imagery.
!
As the AI community continues to evolve at a rapid pace; so will the capabilities of this model.
!
Although there are efficiencies associated with generating images using this model, the costs associated with the API will directly correlate with the amount of usage regardless of the number of images created.
Last updated: February 2026

What Additional Information Is Available for GPT Image 1.5?

API Integration

This model can be accessed through the OpenAI API with a model ID of gpt-image-1.5. It supports text prompts, optional input images for editing, tiered quality settings, and multiple output size options: 1024 x 1024, 1024 x 1536, and 1536 x 1024. It typically takes anywhere from 15 to 45 seconds to generate each image depending upon the complexity of the request and whether or not batch processing has been enabled.

Technical Architecture

Unlike its predecessors which utilized diffusion-based techniques to predict visual tokens, this model utilizes a unified multimodal autoregressive predictive approach that is integrated into the GPT architecture. This allows for sequential prediction of visual tokens to ensure greater prompt fidelity and to retain context throughout subsequent edits.

Partner Integrations

This model is available to utilize through various platforms including but not limited to: Replicate, Runware, and Adobe Firefly. These platforms provide users with the ability to easily deploy the model without having to provide their own direct OpenAI API key(s). In addition, the Adobe integration provides users with the ability to make precise edits within their existing creative workflows.

Use Case Examples

The model is ideal for creating infographic graphics that contain readable text, UI mockup designs, product photography mockups, virtual try-on applications, scene compositing designs, and diagrams. The model maintains the identity and composition of the original image in subsequent multi-step edits.

Prompting Best Practices

OpenAI provides users with a comprehensive guide for creating prompts for this model. The guide emphasizes the importance of providing explicit structure, positioning, lighting, and style descriptions when creating prompts. Additionally, the model supports complex multi-step instruction sets and iterative refinement processes.

What Are the Best Alternatives to GPT Image 1.5?

  • •
    DALL-E 3: Prior to GPT Image 1.5, the strongest image model produced by OpenAI that was good at generating creativity with images but weak in rendering and editability of text. Ideal for artists working with a free-flowing process versus structured editable output. Generation is slower than GPT Image 1.5. (openai.com)
  • •
    Midjourney: A Discord-based AI-generated art tool that has great artistry, community support and user features. Great for stylizing illustrations, does have limitations due to lack of an API, limited editing capabilities and less accurate text generation compared to GPT Image 1.5. Has a subscription model. (midjourney.com)
  • •
    Stable Diffusion 3: An open source model that can be run on local machines or through a platform called Replicate. Allows for customization and is more cost effective when it comes to high volume, however the model requires some level of expertise in creating and setting up prompts versus the ease and accuracy of GPT Image 1.5. (stability.ai)
  • •
    Google Imagen 3: Enterprise image model provided by Google through Vertex AI with robust safety features and integration options. Best suited for regulated industries, however the API for this model will likely be more complex and costly compared to GPT Image 1.5. (cloud.google.com)
  • •
    Adobe Firefly: The ability to generate commercial safe images through a plugin to Photoshop that includes strong editing tools. Best suited for professionals who are already using Adobe products, provides a complementary model to GPT Image 1.5. (adobe.com)

What Is GPT Image 1.5's Model Overview?

Developer
OpenAI
Version
GPT-Image-1.5
Release Date
2026
Architecture
Autoregressive Multimodal (GPT-5 integrated)
Open Source
No
Status
Generally Available

How Does GPT Image 1.5's Model Versions Compare?

VersionRelease DateKey Improvements
DALL-E 22022Initial text-to-image generation
DALL-E 32023Improved prompt understanding, ChatGPT integration
GPT Image 1.520264x faster generation, precision editing, superior text rendering

What Is GPT Image 1.5's Image Generation Specs?

Max Resolution
1536x1536
Supported Ratios
1:1 (1024×1024), 2:3 (1024×1536), 3:2 (1536×1024)
Output Formats
PNG, JPEG (via API)
Generation Speed
15-45 seconds
Batch Size
Multiple variations supported

What Generation Modes Does GPT Image 1.5 Offer?

Text-to-Image

Images generated from natural language prompts.

Image-to-Image

Transformation of existing images with text prompts.

Inpainting

Editing of precise regions while maintaining surrounding context.

Region-Aware Editing

Editing of specific elements without modifying the surrounding environment.

What Style Capabilities Does GPT Image 1.5 Offer?

Photorealism

Natural looking images that include accurate light and materials.

Illustration

Artistic digital illustration and style.

Graphic Design

Posters, infographics, and UI layouts.

Text-Heavy Visuals

Rendering of dense, structured text.

Product Photography

E-commerce style renders.

How Does GPT Image 1.5's Benchmark Scores Compare?

BenchmarkScoreRankNotes
ELO (Text-to-Image)1260#1Artificial Analysis
ELO (Image Editing)1270#1Artificial Analysis
Prompt AdherenceSuperiorvs prior OpenAI models
Text RenderingLeading#1Dense/small text accuracy

What Is GPT Image 1.5's Access Licensing?

Open Source
No
License
Proprietary
Self-Hosting
No
Platforms
ChatGPT, OpenAI API (gpt-image-1.5)
Availability
API and ChatGPT Plus/Pro

How Does GPT Image 1.5's Generation Pricing Compare?

TierCost Per ImageResolutionSpeedNotes
Standard (API)20% cheaper than GPT Image 1Up to 1536×153615-45s4x faster than predecessor
ChatGPT PlusIncluded1024×1024 to 1536×1024VariableSubscription access
High QualityPremium tier1536×153645sMaximum fidelity

What Creative Controls Does GPT Image 1.5 Offer?

Precise Instruction Following

Adherence to complex, multi-step prompts.

Quality-Latency Tradeoff

Ability to adjust the trade-off of speed and fidelity.

Iterative Editing

Maintaining context during iterative refinements.

Aspect Ratio Selection

Automatically optimize generation based on input dimensions.

Seed Control

Reproduction of generations through use of API.

What Is GPT Image 1.5's Content Safety Status?

NSFW FilterBuilt-in moderation
Content ModerationAPI-level filtering
Identity PreservationRobust facial consistency
C2PA WatermarkingProvenance metadata
Usage Policy EnforcementOpenAI safety systems

Expert Reviews

📝

No reviews yet

Be the first to review GPT Image 1.5!

Write a Review

Similar Products

Reve Image 1.0
reve.art
FLUX.2 Dev
blackforestlabs.ai
MAI Image 1
microsoft.com