The Witcher 3 should welcome a new unofficial mod, but that has an element for which it would happen to be. Translated as A night to remember, has new content and voices, the modder has used a trained AI to generate the voice of Doug Cockle and pretend to lend his voice to Geralt.

“If this is true, it is heartbreaking“it said in Twitter Jay Brutton, voice actor in Divinity: Original Sin 2 and Pathfinder. “Yes, AI might be able to replace things, but should it? We literally have to decide. Displacing voice actors with AI is not only walking on mines, it is also heartless“.

This controversy is also accompanied by the fact that Obsidian release a video about his work with Sonanticas they used voices with AI. But only initially, then there were actors, although they explain that the developer saw this as something useful to see what worked and what didn’t.

As usual, it has made voice actors nervous. “There is the instinctive concern about ‘it’s going to take our jobs away’ and I think that in some cases that will happen“said Natalie Winter, a voice actress in games like Assassin’s Creed Valhalla. With ever-new technology, the dubbing could be homemade.

“Sad to think that AI voices could be so good to be widely used“Winter continued. The modder used a software called CyberVoice to create new voices and dialogue for Geralt, it is something created by Simulation Lab and also created CyberMind, which voices NPCs.

CyberMind gives used information – for example, Geralt’s knowledge of the monsters in The Witcher – and CyberVoice gives voice. Mind Simulation CEO Leonid Derikyan explains that technology is necessary. “We created digital versions of voices so that NPCs can answer player questions, outside of missions, with the same voice“.

“And since they form their answers independently and you remember new facts, it is impossible to give them a voice in advance. It would be strange if, in that case, they spoke with a different voice, that’s why we need this AI“. Simulation Lab has worked with voice actors, Geralt was created mixing audios with another voice.

Derikyants explains that Simulation Lab takes all the engineering of manually changing the voice to make it similar to the original. The company trains speech synthesis in audio, something they define as repeating like a parrot, but the quality is very good; although this can bring a lot of problems.

The Top Secret images of the WitcherCon trailer could have been deciphered

What for a solo study or AAA to use someone’s voice to say something, racist, homophobic or of any kind without consent? Even though that him Sonantic CEO Zeena Qureshi does not agree with the AI ​​for the mod, says that at least for his company the offensive should not be of concern.

Qureshi points to the “disclosure system,” which allows Sonantic to pass content through actors before adding anything. “If they are not satisfied with something, then it is not done“, He says. “We take misuse very seriously“.

Fountain: Input Mag