- Speech dialogue
- Media generation
High-quality voice synthesis technology based on deep generative models
Achieves compact AI voice synthesis comparable to recorded voice.
- Developed technology based on DNN technologies and as a new RECAIUS voice synthesis method, achieving high quality that is comparable to human voice.
- Demonstrates ToSpeak’s lightness and ease-of-use as well as high audio quality, e.g., applied/sequential generation functions and adjustment functions of compact technologies.
- We offer proposals in line with customer requirements as a technology with a broad range of applications, from built-in fields to advanced content applications.
Applications
- Content Applications: Reading news, announcements, and train information in facilities, as well as voice output for translation devices and generative AI dialogue apps.
- Built-in Applications: Reading menus as an accessibility function, audio responses in devices, and operations guide voices.
- A Wide Range of Applications: From calm-tone narrations to recreating the voices of actors and characters.
Benchmarks, strengths, and track record
- Toshiba has strengths in technologies that provide high-quality voice synthesis using the latest DNN as compact middleware (SDK).
Inquiries
RECAIUS * inquiry site (In Japanese)
Please include the title “Toshiba AI Technology Catalog: High-quality voice synthesis technology based on deep generative models” or the URL in the inquiry text.
Please note that because this technology is currently the subject of R&D activities, immediate responses to inquiries may not be possible.
* Toshiba Digital Solutions licenses ToSpeak voice synthesis middleware to customers.
References:
- Hiruta, Tamura; “Studies of pitch synchronization analysis applications in encoder-decoder voice synthesis”; Acoustical Society of Japan 2022 Autumn Meeting.
- Toshiba Review, Vol. 78, No. 5 (September 2023) (PDF)
- Toshiba Review, Vol. 74, No. 4 (July 2019) (PDF)
- DiGiTAL T-SOUL Vol.30 (In Japanese)
- Toshiba Review, Vol. 73, No. 5 (September 2018) (PDF)