Official Genshin Impact promotional booth, illustrating the HoYoverse title at the center of the AI voice-cloning lawsuit
Genshin Impact, Wikimedia Commons (public domain photo) - resized to 1600px and re-encoded to WebP; distinct from the photo already used as the Genshin Impact guide cover

A Shanghai court just put a price on stealing a video game character's voice: roughly $112,000. The Pudong New Area People's Court ruled that an unnamed generative AI company illegally reproduced the voices of 63 Genshin Impact characters and sold access to them as real-time "voice packs," ordering the company to pay HoYoverse 750,000 yuan (about $112,000) in damages, as first reported by Game Developer.

The first-instance judgment came down on June 30, but it only became final and started circulating widely in specialized outlets like Automaton West and GamesRadar+ after the defendant withdrew its appeal this month. Some outlets cite a slightly lower figure around $105,000 based on a different exchange-rate snapshot, but the underlying award, 750,000 yuan, is consistent across every source, with $112,000 being the most widely cited conversion.

How the AI service got caught

The defendant built a real-time voice conversion product: users could talk normally while the app swapped their voice for that of a Genshin Impact character mid-conversation, using cloned "voice packs" sold to customers. To advertise the product, the company reportedly used official character art and story animations lifted directly from the game.

The technical details make the case notable. The company admitted in court that it trained its voice model on roughly one minute of stolen audio per character, mixing that small sample with other voice data to generate a convincing clone. A voiceprint analysis submitted as evidence found the AI-generated voices carried "nearly identical characteristics" to the original voice actors' recordings across all 63 characters, characters that include some of the game's most recognizable names, like Raiden Shogun, Hu Tao, and Kaedehara Kazuha.

That evidentiary bar matters for the wider industry: it shows Chinese courts are willing to accept voiceprint comparison as proof of infringement, not just circumstantial similarity.

Why the court called it unfair competition, not just copyright

The legal reasoning behind the ruling matters more than the payout. The court found that Genshin Impact's character designs are protected by copyright. It went further: each character's distinctive voice, built up over years of continuous game updates and voiced content, functions similarly to a trademark tied to HoYoverse's brand.

That framing pushed the case into unfair competition territory rather than a narrower copyright dispute, reportedly the first time a Shanghai court has ruled this way on AI-driven voice imitation. A game studio doesn't need to prove a scene-by-scene copyright violation to win this kind of case. It only needs to show a distinctive character voice, cultivated over time as a recognizable brand asset, was cloned and commercially exploited without consent.

This ruling lands in the middle of a much larger conversation the industry has been having all year about where generative AI helps development and where it creates new legal and ethical liabilities, a debate Vostrike has tracked as adoption numbers climb across the board.

What this means going forward

For live-service and gacha publishers, character voices are core IP, as valuable as the character art itself, since studios like HoYoverse invest heavily in high-profile voice casts across multiple languages to build loyalty around specific characters. A legal precedent that treats those voices as trademark-adjacent assets gives publishers a concrete tool to go after unauthorized AI voice services, rather than relying purely on copyright claims that can be harder to win when the AI output isn't a direct audio copy but a newly generated clone.

Key facts from the ruling:

  • Damages awarded: 750,000 yuan, approximately $112,000
  • Characters affected: 63 Genshin Impact characters
  • Training method admitted by the defendant: roughly one minute of stolen audio per character
  • Court: Pudong New Area People's Court, Shanghai
  • Legal basis: unfair competition, with character voices treated as trademark-like assets alongside copyrighted character designs
  • Status: judgment finalized after the defendant withdrew its appeal

The case doesn't ban voice-cloning technology outright, and the AI company behind the service remains unnamed in court filings. But it sets a usable template other studios, in China and potentially beyond, can point to the next time an unlicensed AI tool starts selling access to their characters' voices.

Nova Ashford
Writer

Gamer and writer specialized in FPS/RPG, always on the lookout for the latest releases and the best PC builds.