ReadSpeaker Korea, a company specializing in voice, recently announced that it has developed a differentiated voice synthesizer with artificial intelligence (AI) technology.
With the recent application of AI technology to speech synthesizer, the sound quality has been further improved and the development period of speech synthesizers has been shortened compared to the existing Unit Selection Synthesis (USS) and Statistical Parametric Speech Synthesis (SPSS).
The AI-applied voice synthesizer has the advantage of creating a natural synthetic sound because it proceeds voice signal prediction with a single learning model without extracting pre-processing processes or parameters (metric information, sound information, sound length information, etc.) from the input text.
However, due to AI application, GPU is required due to high amount of computation. Generally, GPUs are expensive, so expensive equipment is required to use voice synthesizers with AI technology.
Based on its experience of developing voice technology for more than 20 years, ReadSpeaker Korea has developed AI-based voice synthesizers and can operate it only with CPU without GPU.
The sound quality of the voice synthesizer was raised further and the recording time of voice synthesizer was shortened so that the development of voice synthesizer could be accelerated.
In addition, the company has developed voice synthesizers that are manufactured to fine-tune the synthesized sound or to enable the pronunciation through a user dictionary for the pronunciation customers want. In general, voice synthesizers with AI technology cannot be processed in user dictionaries because they do not have a pre-processing process.
ReadSpeaker Korea's AI-applied voice synthesizer can be operated on a platform without GPU, so it can be used in various ways from servers to PCs and terminals, and since GPU does not need to be mounted, it can configure the system at low cost. It also has the advantage of being able to pronounce what users want through user dictionaries.