The “Generating speech in the voice of a player of a video game” patent describes a “computer-implemented method of generating speech audio in a video game”.
This method includes inputting data into a “synthesizer module” representing speech content — both audio and text.
Tom’s Hardware said the patent appeared to suggest the system could work on-device or by sourcing external data — like from a server running while the player is in-game.
It reported that although similar technology as described in the patent already existed, it questioned why someone would want to insert themselves into games in this manner as it would be a quick way to break immersion.
One game that already features a similar system requiring players’ voice input is Inworld Origins.
The game sees the player take on the role of a detective who must ask non-playable characters (NPCs) questions to solve the mystery behind an explosion.
The NPC’s answers are generated in near real-time using AI software.
EA’s implementation is different in that it will likely use some AI to process the player’s voice to generate output that they did not record, but not for “improvised” responses from NPCs themselves.