Level-5 Rushes to Reupload Trailers Amid Growing AI Backlash

Following intense online criticism, Professor Layton developer Level-5 has issued a new apology for the distrust caused by its use of generative AI in a recent showcase for upcoming titles, while also claiming that some examples flagged by fans were simply human error.

The studio further insisted that certain mistakes identified as AI-generated by sharp-eyed social media users were actually human mistakes, and has now pledged to correct these problematic areas.

For context, Level-5 streamed its 'VISION 2026 II: Dreams' presentation on YouTube on September 10. Narrated by a cartoon version of CEO Akihiro Hino, the showcase highlighted the studio's upcoming releases, including Yokai Watch 2: Haunted Domain, Decapolice, and a first glimpse of Professor Layton and the New World of Steam.

However, soon after the video went live, social media erupted with comments and reaction videos. Viewers scrutinized backgrounds and artwork, pointing out inconsistencies and elements that suggested generative AI use, with many accusing Level-5 of producing 'AI slop.' Comparisons between the September 9 Nintendo Direct trailer for Yokai Watch 2 and the VISION 2026 trailer for the same game released just a day later also led fans to identify AI-like errors. The Nintendo Direct trailer for Professor Layton and the Curious Village (a Switch remake of the Nintendo DS original) also had internet sleuths spotting elements that appeared to be AI upscales of original backgrounds.

Oh no Level-5 is DEFINATELY using Gen AI to "update" Professor Layton. Especially when you compare the original backgrounds to the new. So many weird details and inconsistencies were added. https://t.co/AjVNfgetkk pic.twitter.com/0tRQ9vKn03

This backlash prompted CEO Hino to make a statement on X, admitting that the developer had experimented with AI processing techniques for the VISION 2026 II showcase out of a 'desire to make the presentation flashy.' After apologizing for any offense caused, Hino insisted that key elements of Level-5's games (such as story, character designs, and settings) were created by people, and that AI tools were used to speed up the process of digitizing artists' creations. 'Please be reassured that there is no simplistic, AI-generated data in our released games,' Hino commented, before noting that the company's use of AI tools to make processes more efficient was giving its developers more time to be creative, and that they are currently in a trial-and-error process of shortening the 5-year development cycle for major titles down to two years.

Hino's statement went down like a lead balloon, however. The CEO's comments about 'using AI in various ways' to enhance the visual appearance of the showcase drew particular ire, with many Japanese users seeing this as a lazy 'style over substance' approach. 'Isn't the point of a trailer to express creativity, not flashiness?' one person said.

Many commenters in both Japanese and English also pointed out the ethical issues with how generative AI has been trained on copyrighted works of artists, and how Level-5's attitude towards gen AI creates distrust among fans. 'I think human spirit is more important than the speed of game releases, so wouldn't it be better to not use AI, even if it takes five years? Even if just one asset is AI-generated, people will never stop worrying about whether a song, image, or line of dialogue is AI or human-made. Isn't maintaining trust the most important thing?' another commenter responded in Japanese. As a side note, Level-5's concept image for its TGS booth also got flagged as a suspected AI image, prompting further angry comments.

In its latest statement issued on September 17, Level-5 said: 'We deeply apologize for the distrust in our production process that our inexperienced use of AI and simple human errors have caused. Please note that AI has not been used directly in the contents of the games themselves. Moving forward, we will pay even greater attention to the accuracy and quality of the work we produce.'

Level-5 also said that it will clean up 'errors caused by AI use, as well as human mistakes that look like AI errors,' and upload corrected versions of the trailers to coincide with Tokyo Game Show, which opens to the public this weekend. At the time of writing, Level-5 has uploaded TGS versions of trailers for Decapolice and Holy Horror Mansion to its official YouTube channel.

In its September 17 statement, Level-5 published a side-by-side of Yokai Watch 2's promotional illustration before and after mistakes had been edited out. Addressing comments made by users regarding the telephone lines looking 'unnatural,' the company stated that its artist had used a real photo of telephone lines they had taken themselves as a reference, and intentionally simplified the lines to fit the background, which resulted in this 'being misconstrued as AI.' 'These telephone lines were actually hand-drawn by our in-house designer, however, to prevent any further misunderstandings, we have corrected them so that they look more natural,' Level-5 said.

But what about other details flagged by commenters as possible signs of gen AI usage? Level-5 said that, 'Similarly, other issues that people pointed out, such as with the vending machine and guttering, were also the result of human error in omitting these details. The brake levers on the bicycle were not drawn in past Yokai Watch 2 games. However, we feel they should have been drawn for this latest game, considering the higher level of detail.'

⁉️ Dans le nouveau trailer de Yo-kai Watch 2 : Haunted Domain, Nathan a… changé de visage ?! 😭#YOKAIWATCHHier (gauche) / Aujourd’hui (droite) pic.twitter.com/Vr3rzwNPA7

Many commenters also picked up on how Kenta/Nathan's face looks different in the Level-5's VISION 2026 II showcase and the Nintendo Direct that was released on the previous day. In response, Level-5 said, 'We did not make any alterations or corrections using AI.' It said it was unable to correct Nathan's low-poly appearance in time for Nintendo Direct's submission deadline but had updated the model by the time of its own showcase, which explains why Nathan looks different in trailers that were made public just one day apart. Level-5 also stated that 'no AI was used in Professor Layton and the New World of Steam,' however it did not comment on whether or not AI was used for the Curious Village remake.

Level-5 wrapped up its statement by apologizing again for the distrust caused by its 'inexperienced use of AI and human error,' promising to take more care in the future to avoid such mistakes. 'From now on, we will limit our AI usage to enhancing efficiency, and will avoid directly using it in videos released directly to our fans.'

This statement also did not go down well. 'I don't think they understand WHY using AI is bad. We weren't mad at the little errors that AI produced, we just used those errors to determine if it was AI or not, because using AI at all is bad. Now they are going to continue using AI but they are going to hide it better,' one commenter responded.

It's not about reducing the use of AI, but about eliminating it entirely. pic.twitter.com/ziJuMrxBsm

Many Japanese comments also criticized the Level-5 statement for blaming human error. 'You promised not to use gen AI going forward, but aren't you just shifting the blame onto the staff?' replied one commenter. 'I think the main reason for all this criticism is President Hino's stance of repeatedly promoting generative AI usage without confronting its fundamental issues (such as unauthorized use of data and environmental pollution).'

Level-5 has been experimenting with gen AI tools to speed up game development and improve efficiency for at least the last three years. Back in December 2023, the studio released an extensive presentation on its use of AI tools to a Japanese government study group (source: Automaton/ITMedia). In short, it has been using Stable Diffusion to generate layout ideas for promotions, and to create 3D versions of the original 2D models from older games, like its Yokai Watch series. However, the company pointed out then that these generative AI outputs were just used as guidelines for its human artists, who produced the final output. In the same presentation, it described combining parts of AI-generated images with creations by its artists to make the in-game backgrounds (including stadium crowds and buildings) for soccer RPG Inazuma Eleven: Victory Road.

However, it remains to be seen how much the studio's attitude towards AI usage costs it in terms of players' trust in the company.