It’s only been a day since ChatGpt’s new AI image generator went live. Social media feeds are already flooded with AI-generated memes in the style of Studio Ghibli, the cult’s favorite Japanese animation studio behind blockbuster films such as “My Neighbor Totoro” and “Spirited Away.”
In the last 24 hours, I saw Elon Musk’s Studio Ghibli version of Lord of the Rings and AI-generated images representing President Donald Trump. Openai CEO Sam Altman appears to have made a new profile photo, probably made with the GPT-4o’s native image generator, Studio Ghibli-style image. It appears that users are uploading existing images and photos to ChatGpt and asking the chatbot to recreate them in the new style.
The latest update to Openai faces Google’s release of similar AI image features in the Gemini Flash model. This caused a viral moment when people used it to remove the watermark from the image at the beginning of March.
With the latest tools from Openai and Google, it’s easier than ever to re-style your copyrighted work, simply by entering a text prompt. Together, these new AI image capabilities appear to rekindle concerns at the core of several litigation against the developers of the AI model. If these companies are training copyrighted works, are they violating copyright laws?
Products like the GPT-4o’s native image generator work in legal gray areas today, according to Evan Brown, an intellectual property lawyer for law firm Neal & McDevitt. According to Brown, the style is not explicitly protected by copyright. In other words, Openai doesn’t seem to break the law just by generating images that look like Studio Ghibli movies.
However, Brown says it is plausible that Open achieved this portrait by training models on millions of frames in Ghibli films. Even if that is true, some courts still have decided whether training of AI models in copyrighted works falls under fair use protection.
“I think this raises the same questions we’ve been asking ourselves for a few years,” Brown said in an interview. “What does piracy mean when you go out, raw the web, copying it to these databases?”
The New York Times and several publishers have filed aggressive lawsuits against Openai, claiming that the company has trained AI models for copyrighted works without proper attribution or payment. Similar claims have been filed in lawsuits against other major AI companies, including META and AI image generation startup Midjourney.
In a statement to TechCrunch, an Openai spokesperson said ChatGpt refuses to replicate “the style of individual living artists,” but Openai allows it to replicate “the wider studio style.” Of course, it is worth noting that there are living artists, such as Miyazaki Hayo, co-founder of Studio Ghibli, who are believed to have pioneered the studio’s unique style.
Obviously, users were also able to use the GPT-4O’s native image generation capabilities to recreate the styles of other studios and artists. Someone else created a portrait of Mark Andreessen in the style of Dr. Seuss, and the married couple recreated a wedding photo in the style of Pixar.
We tested several popular AI image generators, including those available at Google’s Gemini, Xai’s Grok, and Playground.ai to see their abilities tailored to Studio Ghibli’s style. It turns out that Openai’s new image generator has created the most accurate replica of the style of the animation studio.
For now, Openai and Google’s new image capabilities show that AI models can make a leap to what they can generate. Openai delayed the rollout of its new imaging tool to free tier users on Wednesday, citing high demand. That may be the most important thing for these companies today, but they have to wait for the court to consider their legality.
Source link
