Modder Universal-Modder has introduced a new modification to Grand Theft Auto V named NPCTalk. This mod enables gamers to engage in conversations with non-player characters, receiving immediate spoken feedback. By incorporating this feature, the gameplay experiences a shift from the traditional scripted exchanges to dynamic and intelligent dialogues powered by AI.
Open-source tool combines NVIDIA and Qwen models for dynamic NPC interactions
This enhancement opens up new possibilities for those aiming to immerse themselves further in the open-world setting. The ability to interact and respond in real-time adds another layer to player-driven experiences, closing the gap between player actions and NPC responses.
The functionality of the NPCTalk mod is powered by the conversion of the player's voice input into an AI-driven stack of models. Specifically, NVIDIA Nemotron ASR plays a crucial role in converting the voice data into text that can be understood by the game engine. This text data is further processed using the Qwen 3.8 27B model to produce the appropriate dialogue response based on context, all done in real-time.
After obtaining the dialogue text output, the project integrates Fish Audio Drama 3 to synthesize the voice audio. This part of the process enables voice cloning, ensuring the NPC speaks in a voice that either aligns with the character or the player's chosen voice. Overall, the process from voice input to output audio takes between 1.2 seconds and 2.2 seconds for a response to occur.
The NPCTalk project is made accessible to the public as open-source software through GitHub. The user christos-antono hosts the project, making it available for other modders to inspect and potentially expand upon this new integration. This release represents a practical implementation of large language models and voice synthesis technologies in a classic game title.
Source: IXBT
Discussion
0 comments