When discussing such orders, the price per volume is usually immediately assigned. The task was not really explained, but this often happened, especially because of the NDA. I sent a demo, then I went through a big casting. And only after agreement did I find out what kind of company the order was planned for, and they sent me a contract for approval. At first glance, everything was as usual, the standard wording about the performance and the obligation to write down the volume of texts. However, only on the recording itself it turned out that this is the volume of texts for creating a neural network.
Representatives of the bank were present at the first recording, and when I started asking them, I was told that my voice was recorded only and exclusively for the internal tasks of the bank, it would help train Oleg and replace the call center, that’s all. I was then upset that I did not know all the information about the task and could not adequately evaluate the project, but since an agreement was reached and the studio was rented, I decided to take my word for it and do my job honestly.
A few years have passed since then, and a year ago a friend of mine asked me if I had voiced porn ads. I answered – no, you probably got it mixed up. But he kept saying that it was my voice. Soon, other acquaintances and colleagues began to talk about it and send me recordings of screens, on which, when the video is turned on, explicit porn advertising is launched, without any censorship, which is voiced by the synthesis of my voice.
18+ content, very 18+ content https://disk.yandex.ru/i/DiiinHWOcHISZg (screen recording of that unspeakable)
I began to google and found that my voice is openly available for any synthesis on the bank’s website and is sold on many third-party sites. I started looking for a lawyer, and also trying to contact the lawyers of Bank T. However, in my contract there were no contacts for communication, the sent letters – physical, electronic – the lawyers did not answer, and the bank employees whom I contacted for help and contacts the legal department, refused to give them to me. And all this time they continued to synthesize my voice, they sounded erotic chats, subscriptions and donations on broadcasts, commercial advertising. Even the studios that I previously worked with as an actress began to sell my synthesis to TV channels and streaming services (for example, the documentary “Space: New Horizons”).
Why couldn’t I protect myself? Remember, this was 2019. There is no pandemic, the first voice assistants, Alice and Oleg, are just being introduced, no one can synthesize their voices anywhere, and there is no talk of mass synthesis at all. In addition, the employees promised me that my vote would remain only for the internal tasks of the bank, and I hoped for honesty on the part of such a large corporation. in vain
But what about the contract? It does not contain the word “synthesis”, nor the mention of neural networks, nor any details about the use of records. It’s a standard contract that I create a performance, deliver it, and allow it to be processed. At that time it was an ordinary standard contract, we sign such contracts several times a month. Coordination has always been on the conscience of the studios and intermediaries. Of course, I carefully read the contract, but in this case, the agreement was actually “blind”, since there is not a word about the main thing – that we are writing a synthesis.
In the future, they turned to me for additional recording of the material, and when I asked why and where it would go, they did not answer me and stopped contacting me. I myself made a request with a request to the bank to confirm the said agreement for me even before the publication of the syntheses and remained unanswered.
And I am not alone in this situation. A year later, my colleague, Pavel Dorofeev, also recorded a male voice for the bank, he already knew a little more about the work from me, asked the same questions as I did, clarified whether this material would go somewhere else.
Pavel was told that the synthesis was absolutely necessary for the internal tasks of the bank, and the clause on the transfer of voice to third parties was needed only in order to transfer material for technical synthesis to their subsidiary.
As a result, Pavel’s voice is now voiced in various videos on the Internet, including videos aimed at condemning the actions of Russian troops and a special military operation. We have absolutely no control or responsibility for what our voice is involved in. We cannot be responsible for what people will hear us in. And many are no longer able to distinguish where the synthesis speaks, and where the living person speaks. And the poor performance of synths also undermines our credibility as good announcers and actors.
I spent almost a year trying to contact the bank through everyone I could, as a result I received an answer that I have no right to any claims, however, the bank admits its guilt and is ready to pay me absolutely ridiculous compensation, about the cost of a month or two of my work, provided that I completely waive all my claims. This option does not suit me. Therefore, I filed a lawsuit and will sue Tinkoff Bank. However, during this year I have learned a lot about syntheses and their development, as well as what is happening now in connection with this in the legislation in Russia and in the world.
I found many voices of my colleagues in the public domain, some of them actually recorded syntheses, but in almost all cases the conditions were initially discussed differently or there was no talk of syntheses. Almost all projects were recorded long ago, before the mass distribution of neural networks. I talked to various lawyers in search of someone who would take on my case, and found out that almost all the voices you know who recorded synthesis or voice assistants for someone were somehow deceived and resorted to the services of lawyers. I saw other people’s contracts with other companies in which, like mine, there is no mention of synthesis and neural networks. I also found voices that did not record synths, but still appeared in the public domain, which means that someone, perhaps someone from the studios we work with, used our work with them to create such neural networks. Basically, these votes are uploaded to sites registered to already inactive individual entrepreneurs, like such and similar counterparts.
The apogee of all this is various telegram bots, like Silero or SteosVoice, which opened files from famous computer games, including companies such as Blizzard, CD Project Red, Riot Games and others, and synthesized the voices of actors without their knowledge and without them. permission from the copyright holders, and also sell these votes. But at the same time, they say that they allegedly received permission to use the votes.
Many have heard about artists, that neural networks were trained on their work without their knowledge. But if we are talking about voices, the voice is not synthesized from some common mass, the voice of one specific person is used: his own timbre, his manner, his intonations, diction. It’s very personal, it’s one of the things that basically make up a part of a person’s personality.
Right now, the list of people who were synthesized without authorization is huge, it includes well-known masters and young actors – there are Vsevolod Kuznetsov, Olga Pletneva, Mikhail Tikhonov, Tatyana Shitova, Olga Zubkova, Yulia Gorokhova and many others. Quite recently, a cyberpunk add-on trailer was released, voiced entirely by synthesis of voices from the game.
It is surprising that the authors of these bots not only did not hesitate to steal voices, but also began to write to the actors with a proposal to make a synthesis for their project under the auspices of “protecting voices from theft”.
Don’t you see any contradiction in this? I directly asked the creators of these bots how it happened, and this is what they answered me.
I heard the same rhetoric from Tinkoff bank managers when I spoke to them, and from other people who were not interested in any way. Is it true? Here is what lawyer Tamara Bogdanova, who deals with legal issues and is now working a lot with the topic of speech synthesis, says:
The voice is an intangible good, it belongs to every person from birth. It is part of his personal biometrics, like a fingerprint or an eye pattern. Banks also use voice recognition in their system to verify identity. Many voices we easily identify. And when an actor or announcer signs an agreement to create a Performance, it goes without saying that he is ready to create the result, but the customers believe that they have the right to create a synthesis from a personal cast of the voice and use it as they please, completely regardless of the damage they cause to a person. Although, in accordance with subparagraph 4 of paragraph 1 of Article 1315 of the Civil Code of the Russian Federation, the performer has the right to inviolability of the performance, including the production, the right to protect the performance from any distortion, that is, from making changes that lead to a distortion of the meaning or to a violation of the integrity of perception performance, in its recording, in broadcasting or by cable, when bringing the performance to the public, as well as in the public performance of the performance.
(The quote is not from the video, but from our materials, Tamara tells much more in the video)
As a result, various customers, studios, or third-party people in general are trying to refer to the fact that once the actor has allegedly given the right to process his voice, then you can do anything with him. However, processing has always meant sound engineering processing: cleaning the sound, adding music and other actions, but not creating a public clone of your voice. And now it turns out that, by abusing their right, any studio or customer can dust off a 10-15-20-year-old contract for any audiobook or other project and create a neural network from an actor. And it’s already happening! There is a story of one speaker who recorded a large amount of materials for the department of phonetics already in the early 2000s! And right now, he was synthesized, of course, without his knowledge.
And, as I said earlier, absolutely anyone can be defenseless. Synthesis creators say that technology is now so advanced that they only need a small cast of a person’s voice to create a copy of it. Literally just a couple of minutes.
In the public domain, there are already sites that promise to make a synthesis from a small piece of audio. Imagine what scammers can do with the ability to imitate your speech in real time! Yes, and even use the already available voices – my voice is the voice of the bank’s call center, anyone can now synthesize it and call you on behalf of the bank. The voice of Vsevolod Kuznetsov is the voice Alfa Bank – and his voice for synthesis was stolen from video game files. Now we are not talking about a robotic voice that can be identified – modern technologies record a voice over speech, repeating intonations and even a melody! So, for example, songs are synthesized from the voices of different performers, for example, there are already many records on the network with the voice of Gorshka from KiSha. Only a year or two, and the synthesized voice cannot be distinguished from the real one. And speaking of the Pot, what about ethical issues? Syntheses are already being created from the voices of dead artists and singers – among our colleagues, for example, Vladimir Vikhrov or Andrey Yaroslavtsev. How ethical is it to use their voices in relation to, for example, their relatives? I think that they should at least give their consent, and ideally, relatives should receive royalties, as is the case with copyright.
And everything I’m talking about is happening not only in Russia! Voices from foreign games are stolen in the same way – for example, recently released Erotic mod for Skyrim, voiced by synthesized voices from the game, it thrilled the actors. Speakers who have worked with Apple and their subsidiaries were not aware that their votes are used for AI training. Right now in Hollywood passes a strike of actors, including one related to neural networks – film companies revive dead actors with the help of deepfakes, and many tried to prescribe in contracts the scanning of an actor with a one-time payment and its subsequent use forever. Is such an attitude towards workers from giants, who turn back millions, fair? However, we do not have such guilds, and what I am talking about also concerns the only thing that is on our side now – justice and honesty. […]
iXBT.com, 08/31/2023, “Tinkoff stole his voice with a neural network? The dubbing actress demands 6 million rubles from the company and, together with other actors and announcers, seeks protection from synthesis”: Dubbing actress Alena Andronova filed a lawsuit against Tinkoff Bank and demands 5,960,000 rubles, including 5 million in moral damages. […] Representatives of Tinkoff Bank commented to iXBT.com: The information provided is not true. Alena Andronova and Tinkoff Bank JSC concluded an agreement, according to which Alena performed the work on voicing the texts and transferred the exclusive rights to them to the Bank in full, including the rights to rework, make any changes, reductions, additions. At the same time, the contract with Alena directly provided for her consent to the use and processing of records by third parties. Thus, the Bank has the right to use Alena’s voice recordings by any means, including for training neural networks.
Synthesized materials based on the voices of the speakers are used in the Tinkoff VoiceKit technology, on the basis of which service users can create voice robots and recognize speech. Companies use synthesized voices to automate call centers, create voice assistants, and other business processes. However, in 2022, some users of the service tried to use it in bad faith in violation of the rules. Having identified such cases, Tinkoff immediately blocked such users. Due to the fact that the situation that happened caused a negative reaction from the announcer, we offered Alena Andronova monetary compensation, because. look forward to further cooperation. However, Alena did not agree with the proposed amount of compensation. — Inset K.ru