CD Projekt uses AI to adjust Geralt’s Polish voice in a new Witcher expansion

CD Projekt has used AI tools to adjust Geralt’s Polish voice lines for The Witcher 3: Wild Hunt: Songs of the Past, an upcoming expansion. The lines were recorded by Jacek Rozenek, the actor who has voiced Geralt in Polish since the series began, and then processed by voice technology company Respeecher to make the new performance sound closer to his earlier work.

The studio says Rozenek gave his consent. The decision followed a change in his voice after a stroke in 2019, which affected his ability to speak. Rather than replace the actor or generate a performance without him, CD Projekt had him record the new material and used AI to modify those recordings.

The actor recorded the new lines

Rozenek’s performance is a defining part of the Polish version of The Witcher. After his stroke, he worked to regain his speech and returned to voice acting, but recording new lines could be physically demanding. CD Projekt’s localization producer said the studio wanted to understand how Rozenek’s voice had changed before deciding how to handle Geralt’s dialogue in the expansion.

The actor recorded his new lines for Songs of the Past. Respeecher then used the recordings and prior voice work to adjust the sound toward the voice players heard in the original game. The reported process is therefore different from generating dialogue with no new performance from the actor: Rozenek supplied the words and delivery, while the AI-assisted work altered the vocal character.

That distinction is central to the story. The technology was used to bridge a change in an actor’s voice, rather than to remove him from the role. CD Projekt says Rozenek approved the approach, and he has publicly expressed trust in the studio’s handling of his performance.

Why the studio chose this approach

The studio’s concern was continuity. New dialogue appears alongside existing lines recorded years earlier, and a noticeable change in Geralt’s voice could make the character sound different from scene to scene. CD Projekt chose to use Respeecher after evaluating Rozenek’s current voice and discussing the change with him.

The actor’s health situation is also why the studio has described this as a special case. CD Projekt has acknowledged that AI use in voice acting raises concerns, particularly around consent and the possibility of replacing performers. In this instance, the actor remained involved in recording the new lines, and the studio says it had his approval before the processing took place.

Consent does not settle every debate around voice technology. Questions remain about how voice models are trained, what control performers retain over their likeness, how they are compensated and whether an approach used for one health-related case could become routine elsewhere. Those concerns are broader than this expansion, but the details of this case matter when considering them.

A sensitive use of a contested tool

The story arrives while game developers, actors and audiences are debating how generative AI should be used in production. Voice work is especially sensitive because a performer’s voice is both a creative contribution and a recognizable part of a character. A studio using AI to alter that voice needs to be clear about the actor’s role and the consent behind the decision.

Here, CD Projekt says Rozenek recorded the new dialogue and approved its subsequent adjustment. That offers a more specific account than a vague claim that AI was used to “improve” the game. It also gives players a way to understand why the studio made the choice: continuity with Geralt’s established Polish performance, while keeping the original actor involved.

The expansion will put the result in front of players when it releases. Whether the adjusted performance feels seamless is one question; whether this model of consent and collaboration can be maintained in other projects is another. CD Projekt’s explanation makes the circumstances clear, but it does not remove the need for scrutiny as studios adopt voice technologies.