AI-Generated Cutscenes Pulled From Vapor World After Player Revolt

A studio admits its AI gamble failed

The makers of Vapor World: Over The Mind have explained why they used generative AI for their cutscenes, and they have now reversed that choice. The game, a PC release on Steam from ALIVE and Boltray Games, launched to a "mostly negative" rating of 36%. Coverage of the launch tied the anger partly to bugs and partly to what players called "AI slop" in the cinematic sequences.14 Another account of the launch puts positive reviews at about 25% shortly after release and says many negative reviews named the AI-generated material directly.1 The two figures probably reflect different points in time, but both show the same thing: AI drew more criticism than anything else in the game.

The director's explanation is the most revealing part of the episode. He said the team had tried "for a very long time" to tell the game's story with cutscenes it made itself, and that "each time the result was terrible." He said the team concluded it could not convey the game's complex narrative that way, so it turned to AI.14 One outlet adds that director Young Kim pointed to the studio's limited resources. The hope was that AI would make the cutscenes more expressive while freeing the team to spend more effort on the game itself.1

Kim called the switch a mistake. He said he expected some players to be disappointed, but not that the disappointment would be so widespread.14 A patch has already removed every AI-generated cutscene and restored the original versions the team made before the switch.14 The studio also said it had reviewed all of the AI content and shipped it only as pre-rendered video. Neither point protected it from the backlash.1

Why reviewing the AI work didn't help

The Vapor World case shows something studios often miss. Players aren't only judging how good AI-generated content looks. They are judging the fact that AI was used at all. A September guest analysis on AI in indie development argues that dialogue, voice acting, art and cutscenes are judged on both counts, so even competent output can cause a backlash.1 The same piece notes that the developers disclosed AI use for both voice acting and cutscenes.1

I think this is the main lesson of the episode. Studios that frame AI as a quality question (will the output be good enough?) are answering the wrong question. Kim's account suggests his team thought AI would improve the cutscenes. Players treated the cutscenes as a signal of how the studio values human craft, and quality didn't come into it.

Cutscenes may also be an especially risky place to try this. They are the part of a game where players are most clearly asked to sit back and feel something. Replacing that with generated footage invites more scrutiny than, say, a background texture would.

A pattern across 2026

Vapor World fits a wider pattern this year, in which players spot AI-generated content and studios respond with explanations, apologies and replacement assets.

The highest-profile example is Level-5. Fans watching its "Level5 Vision 2026 II: Dreams" showcase spent more time discussing suspected AI use than the games themselves, which included the Professor Layton remake, Yo-Kai Watch 2: Haunted Domain and DecaPolice.9 CEO Akihiro Hino confirmed the company had used generative AI. He called it "experimentation with the latest technology" and apologized to anyone who found it off-putting.8 He also said the team "cannot deny that we made extensive use of AI this time," because it wanted a presentation that was visually striking and easy to grasp.9 He argued that scenarios, character designs and foundational settings are made entirely by humans, and that AI is mainly used to digitize human-made work, such as converting artwork into 3D polygon models.8

Outlets read the follow-up differently. Polygon said Level-5 claimed some flagged details, including a vending machine, electrical wires and gutters, were drawn by a human artist and were not caused by AI.13 A more detailed write-up says the studio blamed the changes to Nathan's character model on an older asset submitted early to meet Nintendo's Direct deadline. It also notes the studio did not say whether AI was used in the Curious Village remake.7 Level-5's statement promised to limit AI to efficiency work and keep it out of videos released to fans.7 Some Japanese fans said the explanation shifted blame onto lower-level staff.7

Other studios have faced the same problem:

• Pearl Abyss apologized after players found AI-generated 2D props in Crimson Desert. It said the props came from early iteration with experimental tools and were never meant to ship. It promised a full audit of in-game assets and added an AI disclosure to its Steam page.18
• Panache Digital Games said early versions of some portraits and marketing assets had ended up in the prologue for 1666: Amsterdam. It promised human-made replacements and said the full game would contain no AI-generated assets.17
• Jagex pulled a RuneScape Dragon Slayer 2 trailer after players spotted a six-fingered knight. Jagex said an external partner had broken its explicit anti-AI rule. Its marketing director admitted this was the second such incident, which he said made it "frankly even more crap."
• Cygames apologized in January for announcing an AI subsidiary without addressing the controversy around the technology. It said its products contain no generative AI art.19

Where the reporting differs

The coverage agrees on the basic order of events: players notice something, the studio admits it, then apologizes and replaces the assets. It disagrees on how much credit those admissions deserve.

Some commentary praises candor. One analysis of Level-5 argued that Hino's admission confirmed what fans suspected and could build trust in the long run if handled well.5 GamesRadar noted that Larian kept goodwill by admitting it used AI for concept art and then reversing course.17 PC Gamer took a much harder line. It mocked the suggestion that studios are "shocked" to find AI assets in their own builds, and pointed out that Panache apologized "for any upset caused" rather than for the cause of the upset.11 The same report cites a data analyst who found that "AI stigma" on Steam can cut a game's review count by about 53%, with the reviews it does get skewing more negative.11

I side with the skeptics on the placeholder defense, and I think Vapor World supports that view. Kim did not claim AI slipped in by accident. He said it was a deliberate creative choice made under resource pressure.14 That honesty is unusual, and it also exposes how weak the more common stories of forgotten placeholders and mistakes are. At least some AI content ships because someone chose it.

Suspicion is now the default

The second-order effect may matter more for the industry than any single game. Players now look for AI everywhere, and they don't always get it right. Graveyard Keeper 2's flashback cutscenes are a good example. Players on Steam and Reddit cited mismatched fingers and odd poses as evidence of generation, while others pointed to the developers' denials and the nine artists in the credits.102 Level-5 says some of its "AI artifacts" were human error.13 A widely shared Japanese reply to Hino captured the mood: if even one asset is AI-made, players will keep wondering whether any image, song or line of dialogue was made by a person.9

The same distrust is reaching modding. AI-assisted "mashup" videos made by decompiling games have drawn criticism from modders, some of whom use AI tools themselves.3

What it means for game development technology

The industry's case for AI rests on cost and speed. Some coverage notes that big games can take around seven years and millions of dollars to build, which is the pressure pushing studios toward these tools.12 Vapor World shows the limits of that argument. A small team with too few resources reached for AI to fix a narrative problem, and ended up with a review score that hurt the whole game.

The working rule seems to be settling into place. AI kept behind the scenes for testing, physics and digitization is tolerated. AI in anything players see or hear is not, unless it is disclosed and justified, and often not even then. Coverage of Aniimo reflects the same split: the studio distinguished "utility AI" for testing and physics from creative assets.6 Studios that ignore this distinction will keep putting out apologies like these. Vapor World's director is unusual mainly because he said plainly why he took the risk.