It has taken a month to collect all the evidence for this prosecution, and it is about to be submitted to the court. The lawyer said that the next step is to mediate first, and then wait in line for trial. There are so many cases like this now, so we can only wait in line. It is estimated that the wait will take two to three months, or even longer. Although the time and cost of defending rights were huge, Zhang Jiaming did not want to give up. He wanted to "set an example" through this lawsuit, until the lawyer told him that the actual user of this account was only a 13-year-old minor.
"I have no way to sue a child, but all the upfront costs have been invested." Zhang Jiaming said that this situation is not an exception. There are minors and people with weak legal awareness who just want to make a living by making these AI videos to earn traffic. “In the large number of cases where AI has been used to steal my voice, I have no idea how to do it in many cases.”
So far, he has spent a lot of time and energy on defending his rights, but due to various practical obstacles, not one case has been successfully prosecuted. Even if the rights protection is successful, the costs of infringement and rights protection are not equal at all. According to lawyers, even if the court verdict is won, the amount of compensation related to most infringement cases is not large, and the other party will use the excuse of lack of money to make it difficult for the court to enforce.
Zhang Jiaming concluded that one thing that many people who use AI infringement have in common is a weak legal awareness. He realized that many people have no concept of "authorizing the use of other people's voices." When communicating with the infringing bloggers, they asked them to take them off the shelves or use formal authorization. Instead, the other party said, "Why do you think this voice is yours?" They often hold the view that using the voice of a public figure does not constitute infringement as long as it does not break the law or cause malicious vilification.
[2] One character combines the voices of multiple actors, and infringers rely on "guerrilla warfare" to avoid reporting.
Full-time voice actor Xiaomu is also troubled by having his voice stolen by AI. She did not have a professional background, but due to her hobby, she plunged into the dubbing industry. She has dubbed for thousands of films, TV shows and vertical screen shorts. At most, she can record three or four dramas a day, and her income can be tens of thousands of yuan at the peak.
In February this year, she discovered for the first time that her voice had been "stolen." Someone posted a costume-themed audio in a voice actor group: the same character, her voice was used in the first two sentences, followed by another colleague's voice. It was obvious that many people's voices were merged into one character. It may be that the material obtained was not clean enough, and the sound source sounded a little distorted, but people who are familiar with her can tell it apart at a glance.
What shocked her even more was that not long after, her colleagues told her that her voice was in an AI live-action drama on the short video platform. She clicked on the second episode, and sure enough she heard her own voice appear: in a fusion splicing mode, the first few lines were hers, and the last few lines were someone else's. There may be hundreds or even thousands of voice actors' voices in one drama, and each character's voice is unstable.
When the platform generates AI dubbing, it also charges users, which is equivalent to using the voices of Xiaomu and other dubbing actors to make money without anyone's consent or even a greeting.
Then Xiaomu and other infringed voice actors wanted to defend their rights together, but after consulting a lawyer, they learned that the road to defending their rights was extremely difficult – not only did they have to record a large number of dry sounds for voiceprint identification, but different voice lines also had to be individually identified. The cost of pursuing liability was extremely high. Currently, there is no clear legislation on sound copyright protection at home and abroad.
The cunning infringers rely on "guerrilla warfare" to avoid infringement reports – they remove the infringing dramas to avoid the limelight, and within a few days they change their vest and launch a new AI drama, still stealing their voices.
[3] Under the impact of AI dubbing, some voice actors are delivering food, working as real estate agents, and selling used cars.
Voice actor Zhang Maimai believes that a good voice actor is often honed over time. For those who are new to the industry, a few years ago they might have relied on some audiobooks to exercise and support themselves, but now they have become dubbing comics and short plays. Under the impact of AI, the number of orders received by veteran voice actors like him has dropped sharply, not to mention how difficult it will be for newcomers to grow. What’s more serious is that AI has not only squeezed out most of the sinking market, but also stolen the unique and vital voices of actors.
Voice actor Zhang Mai's response. Picture/Screenshot of Tencent News
He remembered that AI began to appear after 2024. At that time, the technology was not as mature as it is now, but the number of orders received had begun to decrease significantly. From more than a dozen audition messages a day, there are only three to five, and the price has also been reduced. When he first encountered this situation, he was so anxious that he couldn't even sleep because of it, wondering if his abilities were too poor. But after a while, he could only look away and accept this "wave of the times". The technology itself was not at fault. "We are not resisting AI, we are resisting abuse and infringement."
Zhang Xiaomai believes that the voices produced by AI and the interpretations brought into the characters by voice actors are just like the difference between clothes cut out on the assembly line of online shopping and private customization. AI does not have a "soul".
Zhang Jiaming also believes that with the arrival of the AI era, voice actors have been squeezed out of their living space. However, he is also actively accepting and embracing AI, and cooperates with the AI production center on AI comics and other content. "The only thing we haven't been replaced now is that our voices still have a sense of living, breathing, and some fireworks. Maybe this is why everyone likes the dubbing of "Nezha"."
Xiaomu feels that AI can imitate the dubbing skills of voice actors with a high degree of similarity, but it is certainly not as flexible as real people in terms of detail processing and emotional changes.
But for many producers pursuing low cost, this is enough for now. She observed that, especially for newcomers, current AI incorporates the works of outstanding teachers in the industry. It is too difficult for newcomers to surpass these "copies". The cost of learning is high and the employment prospects are unclear, resulting in fewer and fewer fresh blood willing to enter the industry.
Many of Zhang Maimai’s students often tell him that they are under great pressure to survive after the number of orders they receive decreases. They are often full of enthusiasm for this industry. They only have an income of two to three thousand yuan in the early stage of entering the industry. In order to accumulate and improve themselves and look for greater opportunities, they basically accept orders from anyone who comes. Speaking from his personal experience, he only "made it out" after 10 years in the industry. During his time in Beijing, he basically recorded some minions, some small advertisements, and some street hawking in the recording studio.
He said that after many of his colleagues who used to dub together got married, their incomes dropped due to the impact of AI and they chose to change careers. Some people are delivering food, some are real estate agents, and some are selling cars.
[4] Experts say there are practical difficulties in determining AI imitation infringement and how to supervise it.
If the voice is "stolen" by AI, it not only infringes on the voice personality rights of the dubbing actor, but may also seriously damage the image of the IP. Zhang Jiaming said that there are still people who use AI to transform the images and dubbing of Taiyi Zhenren and Shen Gongbao to create vulgar and inappropriate content, and even make the characters say bad things such as swear words. Many minors like to imitate and learn. He is worried that such content will have a very bad guidance and negative impact on them.
In order to prevent his voice from being "stolen" by AI, Zhang Maimai adds an anti-counterfeiting mark to his voice every time he auditions. In addition to adding background music, he also adds a "beep, beep, beep" noise. "But this is also a way to guard against gentlemen and villains." Zhang Maimai said that the other party actually has a way to remove these watermarks, but it takes a little effort and time. Some people may think that cloning is too troublesome.
Regarding the legal boundaries of AI dubbing, according to CCTV News, many industry insiders and legal experts believe that in practice, there are currently practical problems in how to identify AI dubbing infringement and how to supervise it.
The first is that the evidence collection chain is weak. Most of the AI voices spread on the Internet are fragmented and secondary processed, making it difficult to trace the source and fix the complete infringement link.
Another difficulty is the difficulty of “identification”. Sounds can be easily edited, tampered with, and changed in speed. AI can fine-tune timbre, intonation, etc. to achieve a high degree of similarity, and can even pinch the voiceprints of several people together. It is very difficult to defend rights from a technical level.
Zhao Zhanhan, a special researcher at the Intellectual Property Center of China University of Political Science and Law, said that industry self-discipline and platform rules are also an important part of building a compliance ecosystem. In the absence of regulatory policies, network platforms and AI software platforms can prohibit users from using other people's voices to generate AI without authorization.
However, how to embed watermarks in AI sounds still requires unified and standardized technical standards. In addition, high-quality and commercially valuable voice actors can register their voices and voiceprints separately for better accountability in the future. With the gradual improvement of the supervision system, labeling system and evidence collection system, an effective protection mechanism for the voice of dubbing actors can also be quickly established.
