NCsoft unveils AI model 'Multiverse TTS' that can be used to produce game voice
Oct 09, 2024
|
Text to Speech (TTS) is a voice synthesis technology that produces voice content such as character voice by inputting natural language. NCsoft's 'Multiverse TTS' can transform various speech styles, generate cross-lingual languages with high speaker tone consistency, and produce multilingual voices with just 3 seconds of prompt voice.
NCsoft plans to use 'Multiverse TTS' technology throughout the game voice production process. NCsoft emphasized that this model enables the production of high-quality and rich AI character voices by utilizing limited voice resources, which can significantly reduce the time and cost spent on existing voice work.
Another feature is that it can be driven by a single model. 'Multiverse TTS' provides TTS with multiple languages and functions as a model to produce multilingual voice content. As it utilizes one optimized model, it provides high-quality voice generation services at a relatively low operating cost compared to competing TTS models, according to the company.
NCsoft published a 'multiverse TTS' model paper, which produces various styles of language and speech with a single model, in the world-renowned AI-related technology association 'EMNLP (Empirical Methods in Natural Language Processing)', and said it has also succeeded in demonstrating global technology.
NCsoft also said it is focusing on researching and developing multilingual voice AI for the global game launch. Starting with this 'Multiverse TTS', the goal is to continue to develop a control function that produces 100 kinds of game character voices within this year and produces voices according to the nature and situation of NPCs.
bluesky@sportschosun.com
