The AI image generation battlefield has never been more intense. In December 2024, OpenAI finally unveiled its highly anticipated new image generation model—GPT Image 1.5—integrated directly into ChatGPT. This move is seen as a direct challenge to its main competitor, Google’s Gemini 3 and its Nano Banana Pro image model.
According to TechCrunch, this is OpenAI’s latest counterattack following Sam Altman’s declaration of “Code Red.” However, the outcome of this showdown isn’t simply about “who’s better.” This article distills five surprising and impactful findings from hands-on testing comparisons.
Highlight 1: A Secret Feature Hidden in Plain Text—Likeness Retention
One of the most intriguing new features in this update is something OpenAI chose to “completely hide.” It’s called “Likeness Retention,” which allows users to perform a one-time portrait upload so ChatGPT can learn your appearance and reuse it in future image generations without re-uploading each time.
This feature didn’t appear in the official press release but was discovered by sharp-eyed users in the “prompt text” of an infographic. The concept is very similar to a feature called “Cameo” in OpenAI’s video model Sora 2, except this is the image version. For creators who frequently need to generate images of themselves—like making YouTube thumbnails—this feature will undoubtedly save tremendous time.
Highlight 2: Faster & Cheaper—4x Speed Boost & 20% Cost Reduction
According to OpenAI’s official API documentation, GPT Image 1.5 brings two key practical improvements:
- Speed: The new model generates images up to 4x faster than its predecessor
- Cost: API costs reduced by approximately 20%
These aren’t just numbers on paper. Faster generation speeds fundamentally change how frequently people use these tools and their mindset toward them, transforming image generation from an occasional novelty into a practical everyday tool. For developers and enterprise users, this means lower operational costs and higher efficiency.
Highlight 3: Finally Able to Truly “Read” and “Write” Text
In the past, getting AI to accurately generate readable text in images has been a major challenge. GPT Image 1.5 has made breakthrough progress in this area.
In OpenAI’s demonstration, the model successfully generated an extremely realistic newspaper photo containing a complete Markdown-formatted article with perfectly accurate headlines, formatting, and numbers. According to VentureBeat, this is a game-changer for many practical applications—whether creating product mockups, marketing materials, UI concepts, or product photos, accurate text rendering capabilities significantly enhance practical value.
Highlight 4: ChatGPT Image 1.5 vs Nano Banana Pro—An Intense Tug-of-War
So how does GPT Image 1.5 compare to Google’s Nano Banana Pro (Gemini 3 Pro Image)? Based on extensive testing, the answer isn’t black and white.
Five Key Comparisons
| Test Category | ChatGPT Image 1.5 | Nano Banana Pro |
| Multi-step Editing | Better at remembering entire prompt sequences | Better at maintaining image consistency |
| Text Rendering | ✅ Clear winner, sharper text | Good performance |
| Crowd Generation | Good performance | ✅ More realistic and natural crowds |
| Brand Consistency | ✅ Tie | ✅ Tie |
| Face Preservation | ✅ More stable overall | Good performance |
Real-World Test Cases
Based on hands-on comparisons, different tasks have different optimal choices:
- Initial generation of “kitesurfer”: Nano Banana Pro performed “much better” with more accurate body proportions
- Editing the same image (adding seagulls and altitude): ChatGPT Image 1.5 “did better” at editing
- Creating YouTube thumbnails: From a “graphic design perspective,” Nano Banana Pro was superior; but for making “faces look like the actual person,” ChatGPT Image 1.5 excelled
This perfectly illustrates why debating “who’s the absolute winner” is meaningless. When generating initial concepts, you might choose Gemini; but when you need precise face preservation or subsequent editing, ChatGPT may be your more reliable partner.
Highlight 5: A Brand New User Experience—Designed for “Regular People”
Beyond the model itself, OpenAI has introduced an entirely new image generation experience within ChatGPT. There’s now a dedicated “Images” tab in the sidebar, marking image generation’s transformation from an add-on feature to one of OpenAI’s core services.
This new interface offers:
- Preset visual style options (pop art, sketch, plush toy style, etc.)
- An explore feature recommending trending prompts and usage ideas
- An image library for managing generated images
This is clearly OpenAI optimizing for regular people—those who just want a style without becoming prompt engineers.
Strategic Goal: Not “Surpassing” but “Keeping Pace”
Behind this update lies a grander strategic story: OpenAI’s primary goal isn’t to completely surpass Gemini, but to reach parity, thereby eliminating reasons for users to switch platforms.
From the results, OpenAI has successfully accomplished this mission. Combined with the recent GPT-5.2 model update, users already in the OpenAI ecosystem now have “few reasons to switch to Gemini.”
GPT Image 1.5 Technical Specifications
| Item | Specification |
| Model Name | GPT Image 1.5 (gpt-image-1.5) |
| Generation Speed | 4x faster than predecessor |
| API Cost | 20% lower than predecessor |
| Key Improvements | Instruction following, image editing, text rendering, face preservation |
| Integrated Platforms | ChatGPT, OpenAI API |
| Commercial Use | Permitted (users responsible for content) |
Conclusion: This Is Its Worst Version
GPT Image 1.5 is undoubtedly an impressive upgrade, but more importantly, it’s a milestone in a rapidly evolving process. AI technology advances daily, and today’s surprises quickly become tomorrow’s norm.
As industry insiders say: From now on, this is the worst it will ever be.
ChatGPT and Gemini have essentially reached parity in image generation and editing capabilities. Now, user choices depend more on personal preferences, existing workflows, and specific but critical needs like “precise face preservation in image editing.” This competition has evolved from a pure technical race into a battle over user experience and ecosystem.
Further Reading
Want to learn more about the latest developments in AI image generation? Check out: