Text To Speech

with OpenAI

Abstract

One of the many services provided by OpenAI is a text to speech service. It claims to be multi-lingual. In this text, we will compare the quality of its output for English, Czech and Russian. Spoiler alert: it is very good.

English

Input:

My name is Jan Palášek and I’m a Machine Learning Engineer. I graduated from the Faculty of Mathematics and Physics at Charles University with a Master’s in Artificial Intelligence. Nowadays, I mainly program in Python and focus on solving various tasks using machine learning techniques.

Output:

Czech

Input:

Jmenuji se Jan Palášek a jsem Machine Learning Engineer. Vystudoval jsem Matematicko-fyzikální fakultu Univerzity Karlovy v magisterském oboru Umělá inteligence. V současné době programuji převážně v Pythonu a zaměřuji se na řešení různých úloh pomocí technik strojového učení.

Output:

Russian

Input:

Меня зовут Ян Палашек, я инженер по машинному обучению. Я окончил факультет математики и физики Карлова университета со степенью магистра искусственного интеллекта. Сейчас я в основном программирую на Python и фокусируюсь на решении различных задач с использованием методов машинного обучения.

Output:

Conclusion

From my point of view, the English variation is perfect.

Czech variant is one of the best I’ve ever heard. Unfortunately, sometimes the pronuanciation isn’t correct. It says “r” letter as if the speaker was American. The letter “ř”, which is very special for Czech language, is also sometimes noticable.

I also cannot notice any errors in the russian. I’m not a native speaker though.

In summary, the voice is fluent and very well done. The pronuanciation isn’t perfect for non-english languages, but I still find it very usable.