Speech Synthesis Middleware: ToSpeak™

Provides far more realistic speaking voice than the current closed-loop training (CLT) technology with about the same amount of memory.

ToSpeak™ uses a new Toshiba-proprietary technology to reduce sound-quality variations. This technology selects plural speech units and fuses them together to make the synthesized voice sound more realistic and natural.
The memory-resident version of ToSpeak™ consumes only as much memory as the current CLT technology (approx. 2 MB) to implement a speech corpus.

This figure shows the comparisons of Sound Quality and Memory Requirement between the Conventional and New Speech Synthesis Technologies.

Comparisons of Sound Quality and Memory Requirement between the Conventional and New Speech Synthesis Technologies

General Specifications

	Japanese	Chinese	English
Required Memory (Incl. Code and Dictionary)	Approx. 5.5 Mbytes^*1	Approx. 8 Mbytes^*1	Approx. 8.5 Mbytes^*1
System Config. Example	Japanese TTS: Approx. 4.0-Mbyte ROM and approx. 2.0-Mbyte RAM^*1
API Specification	Toshiba TTS API Specification (Android^*2 TTS API can also be supported.)
Input Text	Shift JIS	GB18030	UTF-16, Latin9
Output Speech Format	Signed 16-bit linear PCM
Output Speech Format	22.05 kHz

*1: Values depend on conditions.
*2: Android is a registered trademark of Google, Inc.

Product Introduction

Two versions of the speech synthesis middleware are available according to application requirements.

This figure provides an overview of the Speech Synthesis Middleware (SYN & TTS).

Features of the SYN middleware

Takes a string of phonetic symbols as input and adds natural prosody and intonation
Eliminates the need for linguistic analysis and thus saves memory.

This figure provides an overview of the Features of the SYN middleware.

Features of the TTS Middleware

Accepts plain text as input and converts it into speech.
Supports phonetic transcriptions as input.

This figure provides an overview of the Features of the TTS Middleware.

Benefits of Using Speech Synthesis

Click this image to hear a sample voice.

Route guidance of a car navigation system

ToSpeak™ enables route guidance including a huge number of proper nouns with a natural-sounding voice. ToSpeak™ can be upgraded through simple maintenance of a phonetics database; it is unnecessary to record utterances spoken by professional narrators.

Drive assist

Draws a driver's attention with voice warnings.

Hands-free phone calls

You know who is calling without taking your eyes off the road.

Reading out the delivered information

You can make ToSpeak™ read out content delivered from information service providers.

Reading out emails

Why not add voice messages to make your emails friendlier?