Challenge
Big Tech is currently in a state of vigorous development and is expected to continue its evolution all the way through the following decades. Although the industry giants are already offering numerous voice assistive technologies on the market, TTS software is still not fostered to their fullest potential. In addition to this, while English TTS has a clear room for improvement, Dutch speech synthesis is frankly lagging behind in terms of grammar and speech articulation due to the low adoption. Seeking to tackle this issue, Bit teamed up with a Dutch media group to prototype the voice synthesis software that would unleash the innovation's capacities in the context of local digital landscape.
Solution
During the course of the project, Bit has designed an addition to the voice synthesis of Microsoft Azure - a cloud computing service with a speech feature that converts text to lifelike speech. After building up upon the base layer, our research team developed an evolving lexicon of Dutch words that usually happen to be mispronounced by the TTS software and pre-processed the news articles in such a way that it enhances the TTS system by Azure, for instance, pronunciation of large numbers or punctuation. In order to handle this issue, we had to resort to the use of phonetic transcriptions, therefore signifying to the system how certain words and complex sentences should be articulated. Subsequently, the speechmaker dashboard was established for experimenting with various voice settings, such as pitch, speed, and even computerized speakers. After making additional adjustments regarding the structure and going through the general validation, Bit developed a substantial demo version of Dutch TTS, which was furtherly refined and integrated into the digital newsletter of our partner media entity.
Results
As a result, users are now able to play the headlines and enjoy the high quality of the speech synthesis, whilst our client, on the other hand, directly benefits from the developments of Big Tech and remains a forerunner in the Dutch media landscape.