You've been logged out of GDC Vault since the maximum users allowed for this account has been reached. To access Members Only content on GDC Vault, please log out of GDC Vault from the computer which last accessed this account.

Click here to find out about GDC Vault Membership options for more users.


The Number One Educational Resource for the Game Industry

Session Name: Machine Learning Summit: GPT-3 Powered Text to Lifelike Speech and Animation for NPCs
Speaker(s): Dao Si
Company Name(s): NUVERSE
Track / Format: Machine Learning Summit

Did you know free users get access to 30% of content from the last 2 years?

Get your team full access to the most up to date GDC content

Overview: Performance-driven narrative video games needed NPCs' performance to be realistic and depict a wide range of believable emotions. Accurate sentiment analysis and semantic understanding of the text can better help games' audio and animation content generation. This session describes a novel system in 'Earth Revival', using GPT-3 to measure sentiment and extract semantic features, to automatically synthesize emotional voices and high-quality emotional, expressive full-body animations for talking NPCs. In this system, the speech synthesis system introduces paralinguistic elements to achieve realistic emotional expression, which can produce natural-sounding voices for final game releases or content updates. What's more, the automatic full-body animation generation model uses the multi-modal context of speech text, audio, and speaker identity to produce the arbitrary beat and semantic full-body animation together.This system of GPT-3 powered text to lifelike speech and animation can significantly improve the narrative process and minimize time and cost.

Game Developers Conference 2023

Dao Si


free content

Machine Learning Summit