logo

Author: Jeremy Huang

On March 25, 2025, OpenAI officially announced that its flagship language model, GPT-4o, is now fully integrated with native image generation capabilities, bringing more advanced AI-powered visual creation tools to ChatGPT users. Starting today, image generation is a built-in feature of ChatGPT and is available to Free users, Plus subscribers, Pro users, and Team accounts. The functionality will soon be extended to Enterprise and Education tiers. Additionally, GPT‑4o's image generation capabilities are now integrated into OpenAI's video generation platform, Sora, with API access expected to roll out in the coming weeks.

Users can simply describe the image they want through natural language prompts and receive high-quality images, with support for specific parameters such as aspect ratio, transparent backgrounds, and HEX color codes.

Key Feature Highlights

Compared to OpenAI's earlier DALL·E model, GPT‑4o not only offers more diverse styles but also introduces several comprehensive upgrades:

• Native Integration and Conversational Creation Experience
  Image generation is embedded directly within the GPT‑4o model architecture, eliminating the need to switch tools or interfaces. Users can generate, refine, and edit images through natural multi-turn conversations.
• Enhanced Contextual Understanding and Image-Text Fusion
  GPT‑4o can analyze user dialogue context and uploaded images in real time, incorporating them into image outputs. It supports use cases like image-to-image and text-to-image transformations, far surpassing the single-input model limitations of DALL·E.
• Accurate Text Rendering
  GPT‑4o can precisely render elements such as signage, menus, and invitations—ideal for marketing assets, infographics, and educational materials—solving a common issue where DALL·E often produced distorted text.
• Improved Object Control and Detail Accuracy
  Capable of rendering 10–20 objects with associated properties, GPT‑4o enhances object-attribute binding accuracy, reducing misinterpretations that were frequent in DALL·E’s complex compositions.
• Support for Diverse Styles and Photorealism
  From photorealistic photography to digital illustrations, postcards, vintage film effects, and hand-drawn science posters, GPT‑4o can consistently generate stylistically coherent and detailed images with effects like gradients, shadows, and lighting enhancements.
• Expanded Use Cases: Gaming, Character Design, and Branding
  GPT‑4o allows for consistent visual identity across game character development and branding materials such as logos, typography, and layouts—making it a powerful tool for creators and designers.

Ghibli-Inspired Boom Drives ChatGPT User Surge

Following the rollout, the new image generation tool has gone viral globally, especially for its ability to transform user-uploaded photos into distinctive artistic styles. One particularly popular trend is the creation of “Ghibli-style” images, drawing inspiration from the renowned Japanese animation studio.

The process is simple:

1. Upload a clear photo of a person or pet on the ChatGPT website or app
2. Enter the prompt: “Please convert this image into Studio Ghibli style.”
3. Wait a moment for the result to generate

According to a Reuters report on April 1, this trend helped push ChatGPT usage to a new record, temporarily overwhelming servers and leading to usage restrictions. Market research firm Similarweb reported that average weekly active users exceeded 150 million for the first time by the end of March. OpenAI CEO Sam Altman also posted on X (formerly Twitter), saying: “We added one million users in the last hour,” highlighting the contrast with ChatGPT's early launch, when it took five days to reach one million users.

Further data from SensorTower indicated that by the end of March, ChatGPT hit record highs in app downloads, active users, and in-app purchase revenue. Compared to the week before, app downloads increased by 11%, weekly active users rose 5%, and in-app revenue grew 6%.

However, the surge in demand also led to minor outages and errors. OpenAI co-founder noted, “We're getting things under control, but users should expect some delays, occasional bugs, and slower service speeds as we handle capacity challenges.”

Copyright Disputes Over Ghibli-Style Image Generation

The widespread creation of Ghibli-style images using AI tools has also raised copyright and intellectual property concerns. In the Reuters report, Evan Brown, a partner at Neal & McDevitt law firm, said:
“The legal risks of generating AI images that mimic Studio Ghibli's style remain in a gray area. Copyright law generally protects specific expressions, not artistic styles themselves.”

So far, OpenAI has not publicly commented on whether its image generation models were trained using licensed data. However, an Associated Press (AP) article dated March 28 cited OpenAI's technical documentation, which stated that the company takes a “conservative approach” when mimicking the styles of individual living artists.

“A refusal is triggered when a user attempts to generate an image in the style of a living artist,” the document reads. Still, the company clarified that “broader studio styles are permitted, which have inspired many delightful and original fan creations.”

Studio Ghibli declined to comment on the matter. Nevertheless, a CNN report resurfaced a widely circulated video clip from 2016 in which Studio Ghibli co-founder Hayao Miyazaki expressed deep discomfort after viewing an AI-generated animation demo. Known for his frame-by-frame hand-drawn animation, Miyazaki said:

“I am utterly disgusted. If you want to create creepy things, go ahead—but I would never incorporate this kind of technology into my work.”
“I strongly feel this is an insult to life itself.”

Taiwan's Government Response

In a March 31 report by CNA, Taiwan's Intellectual Property Office (TIPO) under the Ministry of Economic Affairs responded to the rising concerns. The agency emphasized that whether large-scale use of copyrighted works to train generative AI constitutes “fair use” remains to be clarified and recommended that developers seek prior consent or authorization from rights holders.

While emulating an artist's style does not necessarily violate copyright law, TIPO noted that if a dispute arises over whether AI has plagiarized expressive elements or mimicked a style too closely, it must be determined by the courts on a case-by-case basis.

The agency added that it has hosted forums on generative AI and copyright, and continues to monitor global discussions and legal developments. Given that intellectual property law is highly harmonized across countries, Taiwan will adopt a balanced approach to protect creators' rights while supporting the healthy development of the AI industry.

Conclusion

The release of GPT-4o's image generation capabilities marks a new era in intuitive and powerful creative tools for AI applications. However, the Ghibli-style image boom also underscores unresolved legal and ethical questions in the world of generative AI. As current copyright laws struggle to define whether "style" is protected, governments and creators alike are seeking clarity.

Taiwan has begun taking steps to address these challenges, recognizing that striking the right balance between safeguarding artistic integrity and fostering innovation will be crucial for sustainable AI development. As technology continues to evolve, so too must the frameworks that govern its use.

📌 Disclaimer:

The content of this article is for reference and informational purposes only. The AI image generation features mentioned herein are described purely in technical terms. Actual use must comply with platform guidelines and applicable copyright laws in each jurisdiction. If users generate images using generative AI that involve specific artistic styles or the likeness of public figures, they are advised to carefully assess any potential issues related to intellectual property or image rights. Users are solely responsible for any associated legal liabilities.

Reference

經濟部：AI抄襲或模仿風格若有爭議由法院個案認定

Ghibli effect: ChatGPT usage hits record after rollout of viral feature | Reuters

ChatGPT's viral Studio Ghibli-style images highlight AI copyright concerns | AP News

Viral Studio Ghibli-style AI images showcase power – and copyright concerns – of ChatGPT update | CNN