A woman surnamed Wang is a verified voice actor on a certain APP, where she regularly performs dubbing and publishes works to earn compensation.

In January 2025, Wang learned from a friend that Company A had allegedly used her voice in its internet promotional campaign for user acquisition. After verification, Wang believed the voice used in the campaign was likely AI-synthesized, so she notarized the relevant audio to preserve evidence and filed a lawsuit with the People's Court.

Wang argued that she had never cooperated with or authorized Company A in any way, and that Company A's unauthorized use of her voice constituted infringement. She demanded that Company A cease the infringement and pay 300,000 yuan in compensation for economic losses.

Company A argued that the audio in question was indeed AI-generated by the company, but the specific source and the materials fed into the AI could not be confirmed due to the departure of a former employee. The company claimed it had previously been unaware of Wang's identity as a voice actor and had never collected or used her voice. It also argued that existing evidence could not directly prove that Wang's voice was used in its promotional campaign.

During the trial, to prove the facts of infringement, Wang conducted a forensic voice analysis comparison between the notarized audio in question and her own voice recorded on the spot. The forensic opinion showed that among 28 formant acoustic indicators, 24 indicators had a deviation of less than 10% between the audio in question and Wang's own audio, with 16 of those being less than 5.36%. The portions that were relatively similar and highly similar reached 90%.

The court ruled that Company A's conduct constituted AI-generated voice infringement and caused economic losses to Wang, and ordered Company A to pay Wang 50,000 yuan in compensation.

Both parties were dissatisfied with the first-instance judgment and filed appeals. The Shanghai First Intermediate People's Court, upon hearing the case, held that using a natural person's voice as training corpus without their consent to imitate that person's timbre, tone, and pronunciation style to generate a synthesized voice capable of identifying that person should be recognized as infringing upon the natural person's voice rights.

In this case, Wang, as a voice actor, had provided evidence proving that her dubbing works were published on public online platforms, and Company A had information channels and possibilities to obtain the source of her voice. Meanwhile, according to the forensic report analysis, the synthesized voice had a small deviation from the human voice audio, and the two were highly similar, sufficient to enable ordinary members of the public to associate a specific identity, thus possessing identifiability.

Although Company A denied this, as the party in control of the original training data, it failed to provide sufficient evidence regarding the legality of the source of the fed materials, the non-use of the natural person's voice characteristics, and the absence of public confusion or misidentification. It should bear the adverse consequences of failure to provide evidence.

Accordingly, it can be determined that Company A, without Wang's consent, commercially used a synthesized voice capable of identifying an individual's identity to the public, which constituted voice infringement.

Regarding the determination of the amount of damages for voice infringement, the standard should prioritize reference to the rights holder's licensing fees for similar uses; if the rights holder's evidence is insufficient and licensing fees are difficult to ascertain, the loss amount may be comprehensively determined by considering factors such as the duration of infringement, scope of dissemination, manner of use, degree of fault, damage results, and the contribution ratio of personality elements in commercial exploitation.

In this case, the voice in question was available for a relatively short duration and was a procedural broadcast prompt sound effect, not a core selling point or traffic-driving vehicle of the campaign. Additionally, Wang's online following and public activity were not high, and the standalone commercial market value of her voice was limited. Therefore, the court held that the first-instance court's discretionary award of 50,000 yuan in compensation was fair and reasonable.

In summary, the Shanghai First Intermediate People's Court ruled in accordance with the law: the appeal is dismissed, and the original judgment is upheld.