Embark re-records some Arc Raiders AI lines but keeps text-to-speech

Embark Studios re-recorded some AI-generated Arc Raiders voice lines with human actors after players criticised the game’s use of synthetic speech. The move added more human performances, but did not amount to a pledge to remove AI voices from the game.

The studio has not disclosed how many lines were remade, which lines changed or when players would hear the revised audio. It has said recording every possible line would be impractical, particularly for the game’s ping system, where additions of items and features can create further dialogue requirements.

Eurogamer reported that the re-recordings had been made after the game’s launch. Chief executive Patrick Söderlund said there was “a quality difference” between AI and human performances, adding: “A real professional actor is better than AI; that’s just how it is.”

Embark’s approach relies on actors being paid for studio recording sessions and, for selected uses, for approving the licensing of their voices through text-to-speech. The studio has described those uses as less central to immersion, chiefly ping-system audio. It also brings many actors back for game updates.

The studio said it uses AI chiefly to test dialogue internally and to assist some content creation, rather than invariably replacing performers. Stefan Strandberg said text-to-speech could expand the game in areas requiring repetitive work, while maintaining that Arc Raiders’ gameplay experience itself did not use generative AI.

PC Gamer reported that players had described some AI lines as lifeless, lacklustre and odd. Embark acknowledged that some criticism was fair and said it used the technology to release updates faster. The company said actors’ voices had supplied the training data used to prototype dialogue before final recordings, but that bringing performers in for every fast-changing ping-system variation was not feasible.

The response leaves the central boundary intact: human actors remain in key roles and some lines have been replaced, while licensed text-to-speech continues to be used for other dialogue. AI and Games criticised that model as offering little player benefit and argued that it raised wider ethical questions about the use of trained voice models in big-budget games.