NCSoft Unveils AI Model for Automatic Character Voice Generation

On the 8th, NCSoft announced the release of an artificial intelligence (AI) model called 'Zero-shot Multi-verse TTS' that can be utilized throughout the entire game voice production process.

TTS (Text to Speech) is a speech synthesis technology that creates voice content such as character voices by inputting natural language. NCSoft's newly released 'Multi-verse TTS' enables ▲various speech style transformations ▲cross-lingual generation with high speaker voice similarity ▲and multilingual voice production using only a 3-second prompt voice sample.

NCSoft plans to apply the Multi-verse TTS technology across the entire game voice production process. By utilizing this model, it is possible to produce high-quality and rich AI character voices with limited voice resources, reducing time and costs compared to traditional voice work.

Another notable feature is that it operates with a single model. Multi-verse TTS provides TTS capabilities for multiple languages and functions within one model to create multilingual voice content. By using one optimized model, it offers high-quality voice generation services at relatively lower operational costs compared to competing TTS models.

NCSoft also demonstrated its global technological prowess by publishing a paper on the Multi-verse TTS model, which generates various styles of languages and voices with a single model, in the prestigious AI-related technical conference 'EMNLP (Empirical Methods in Natural Language Processing)'.

Currently, NCSoft is conducting R&D on multilingual voice AI for global game launches. Starting with this Multi-verse TTS, the goal is to produce voices for 100 game characters within the year and continuously develop adjustment functions that create voices tailored to the personality and situation of non-player characters (NPCs).

This content was produced with the assistance of AI translation services.