With the help of the neural network, Idris Elba spoke in Russian in the Cyberpunk 2077: Phantom Liberty trailer

Cyberpunk 2077 Phantom Freedom Traiiler in English, voiced by the neural network trained on voices from the original video. He took the voices from the same video that came out in English, the new generation neural network is able to study with only 10 second examples of the voice (the president’s voice was trained at 6 seconds of its 2 offers in the video), while the old -generation networks require hundreds of hours for this. The bottom line is that I did not voiced anything with its voice, the network works “Text-Voyus”, that is, voices exclusively by subtitles, but she can understand the context of who and what he says, since she was given a relatively speaking to listen to the original video, and creates emotions in the voice but already in English voices of original actors (the same Idrisa Elba for example). If there is more data for training, then everything DLC ​​will be announced in automatic mode in better quality. Nevertheless, tearing off the voices from the video where a lot of background sounds and music is like that.

Also, the author gave an example how the synthesis of the voices of real characters of the game (Vi, Adam Sweeter, Johni Silly Super)

The actors of the voice acting will soon remain without work.

In, with such requests)))

Soon – in 20 years? To voice something decent in this way we need power. it will cost more than the actors of voice actors in the foreseeable future. This is not even given that a living person in any case conveys emotions better

They will not remain. Do not forget, they are actors, not working voice actors. And good voice acting differs not just in voices – it differs in the game, handed over in voice. The neural network can make words in the right voice, but it cannot make the game a voice.

This has already been said about unmanned cars who will leave drivers without work. And how many of them are not driven in tests – they still work fully only in factories, in the form of trucks that ride along the same route.

Time of time. A start.

The workers of the 18th century also thought that cars in factories would not replace manual labor, and then only workers who serve the machines remained.
It will be the same here: for example, one person translates the trailer with the help of AI. Whereas now it is necessary to hire actors, directors, sound engineers, coordinate time, rent a room, then mixing, etc.

Cool of course, but I read in his posts that it takes 25 days to voice acting. In total, let’s say in the game of dialogs for 30 hours – this is 750 days, not a real volume) so purely trailers voice)

so buy him a computer and will be faster

there the whole game is in addition to 30 hours, I think dialogs from the force of hours by 5-7 hours

And if this is done not in one haru, but a team of 20-50 people, then there will be much less human arms.

You can rent performance, many who are without supercomputers, but who needs computing power for a specific one -time task, only the price of the price remains, whether it will be lower than that of the studio, which is ready to take up voice acting or not.

Well, if 10 hours of AI voiced a minute of the conversation.

then will be thrown away to pay for the mains

And three cars of video cards to the farm.

10 hours do not need. For a long time, a neurona has been on the network where a 10 second scrap of a voice can be taught to talk, and with emotions without any robotic voice.

And training takes only 5-10 minutes

God, how cool it sounds!!

And what I said, with the advent of neural networks, Oo -Ovuchakina go to the free cash desk in the five)))) otherwise you understand 20 million

Sagles, coders are many times cooler than any type of voice!

and the neural network is studying? Isn’t it at work "Vlavchulkinov"?

A neural network will not replace a person, in voice acting so accurately. The neural network does not know how to build intonations, does not understand the context, and accordingly stress suffers, moreover, it sounds artificially. Plus it is a matter of ethics.

So it is, I hope this will begin to regulate this by law, otherwise we will lose their voice actors with at such a pace

I am completely satisfied with the voice acting of the neural network. For the rest, a free cash desk)

This can already really fake speech

The update came out or not?I do not get it.

I’m not waiting for this feces

I sympathize with your unhealthy, not pain

I would also understand what it is. You have a random tick for some?

your campaign is something serious, here medicine is powerless

The coolest thing is that the transfer of actors with an original voice is very atmospheric!

It would be nice with RDR 2. And then wait until 10 Lyamov will be collected for her voice acting, and only then, work will begin on voice acting. Of course, the neuro cannot be compared with acting. But certainly better than nothing

Very cool of course. I would play like that with pleasure.

Well, here you don’t need to worry about the voice acting

To be honest, you hear how they wrote before, "announced by professional programmers"

Lol, Grozny Nick.

After how much update 2.0 comes out? or what time in Moscow time?

It seems like at 18:00 Moscow time, immediately after Stream with development

Goga just went out, already on the leap

Translation using a neural network is good, but what is missing is emotions. Too dry voice.

This people so want English voice acting at Cyberpunk 2077, that they created what the non -huts can be lifted by the loot. Bravo.

I liked the first version of male vi. I hope the author will continue to develop his creation. I was amazed how the voices were made in English. Bravo to the author!

Someone voiced the video and then the vote shook the neural network

The neural networks will not yet voice the games

Even if the neural networks for translation of text in games do not work perfectly

I would not be so sure – there Yandex and already completely translates and voices vidos from YouTube in real time. The translation of more in places is crooked, but this is only the beginning.

So that on the BG3 there is already complete voice acting neurosity.

So there it seems not neural network, but just a robot and voice acting old or new did ?

It seems to be norms, but some kind of cotton wool selected in the voice, I thought it would be better, and the processing of the voice is the same

Big Dizlaik. You can’t just take and steal the voices of the actors in the voice acting. Otherwise there will be no voice acting games in the future.

Um. The fact of the development of voice acting neural networks suggests that everything will soon have a voice acting, even a teenager will be able to release an indie game and say 100 characters in it will be voiced, moreover, in different voices. And it will not be necessary after the review of the game that she is cool to write on the great broken English *** please Add English Language!*** —- on which the developers put their fragrant (after a working day) (XD) sausages (XD) in the skinny (after paying for the next AAA of Offer Price) players rolls.

Purely my opinion, I do not want to argue

You, like most players, are just a consumer for whom only the final product is important, and not how it was created. Imagine, if you were engaged in the voice acting of something, or was an actor or singer, and then saw yourself in the main role in the movie about which you never heard or heard a song where you sing, but never participated in the record of this song? How will you feel? You need to be at least a little empathic to understand that there is nothing good in this and something needs to be done with this. I do not think that this person contacted the actors and received the good to use their votes.

Purely your opinion)))

And I ask you not to go to persons, you don’t know and you can’t judge me in any way, there is too little information for this 🙂

fair and logical, but it will not change anything. You can also argue about the fact that "If you were a developer/publisher of the game and would look at the statistics, as a product you released a week ago, 60 bucks per copy of the downloads from torrents is already a million times. ", But who does not want to pay, he will not pay. And no matter how much you block the torrent of sites, their endless mirrors immediately come in their place or people simply turn on VPN. So, I think this (development of neural networks) can no longer be stopped, only to look and oh *** at what speed they develop. And nothing can be done here.

Oh, how sad! Something like the moaning of the Luddites in due time. Just imagine a thousand weaving hands was replaced by one high -performance machine. There was no limit to wild hatred. There were widespread killings of prominent engineers and scientists, sabotages and arson, a revolution nearly broke out. Only now you know why suddenly after a few years the pogroms stopped? One fine day, the former weavers, whose salaries were hardly enough for the product they made suddenly found that thanks to progress, the market was saturated with machine -produced tolerable quality and at an affordable price.

So why am I? Oh yes – progress does not stand still. Those who were everyone yesterday – today suddenly became anyone – this is life. This is a struggle. Before the invention of the tractor, 90% of the population of the planet were busy treating agricultural lands. So now in the case of actors and singers, when they felt the sunset of their era and the end of their huge fees. What you described in the case of them not quite correctly. In each individual work – the creator is either a director or composer, and the beneficiary is a producer. An actor or singer is only a hired worker – they take a minimum of time in the creative process. And in addition to this, imagine that thousands of actors themselves or involuntarily lost their own fateful roles of Ford instead of Connery in Star Wars, Reeves instead of Will Smith in the matrix, a lot of songs were performed not by the singers who the composer and so on. But we did not lose these works from the refusal of individual mercenaries. And this applies to the sphere of cinema and songs. But the characters of the games are sets of landfills collected in the model. Their appearance or voice tied to a separate famous actor is only a PR move for promoting the product and no more. A lot was lost or acquired that Vann Damm refused to participate in the very first Mortal Kombat, which was replaced by the Joni Cage pixel clone performed by a third -party actor who was set to resemble Claude? Or the fact that instead of the planned Mashkov in the GTA4 turned out a half -clone, which practically looked like Mashkov sounded like Mashkov, but a third -party compromise performer played the role?

And why appeal to the feelings of the actors in the case of the game I do not understand? Well, you got a bunch of famous actors. Well, you gave appearance and voice to ten to the heroes from your game. And what about the rest of the hundred? Why should the similarity of the game model with a famous actor bother me more than the similarity of the random NPC with Vova Frolov from the next entrance? Almost all men’s non -grades in the recent Starfield remind me either Gazmanov or Kirkorov, well, and they don’t pay the poor fellow. Well, is it possible. Why suddenly the lulling ram of Idris Elba suddenly more unique than the honey timbre of the Nejidi Vasya’s neighbor after a white check of Belaya? Who later went and digitized for a couple of thousand for neurosettes his vote or textured external? Than it is bad when the creator of the games later saved a couple of million on the similarity of models and voice acting, investing in technical and gameplay quality? After all, the same thing happened 10 and 20 and 30 years ago! Only now you do not need to waste time on compromise actors who should look and sound as the director intended. He just presses the desired button. Progress occurs.