Festim Peposhi, a graduate in the Faculty of Sciences in Tirana, works today in Germany in technology companies and AI. He and his team started a successful project two years ago. ZERI.app provides you to listen instead of reading the news.
Festimi noticed that one of the genuine services was missing in Albania. The services that automatically generate a written text into audio, or in other words text-to-speech conversion.
Google offers such a service through Google Translate, but in this case, we are dealing with a computerized voice, which is far from natural sound. Also, this service for the Albanian language is not very suitable for professional and reliable use.
Genesis of Zeri started working on a project to create a system that collects and presents to the reader the latest informative articles from all Albanian-speaking information platforms. It was an opportunity to assist readers in listening to information. This way the reader can hear the news while walking, driving, running. The information would be accessible even to people unable to see.
"After looking for a service that could perform this function, I realized that such service was missing in the Albanian language. This is where it all started," he said.
After research work in many different articles and publications, he begins to consider the development of a service that enabled audio of materials written in Albanian. He points out that "the more I read, the more challenges appeared in terms of professional plan, experience, time or resources."
"We have been working with ZERI for a long time and in addition to studying, testing and coding, we have managed to record nearly 100,000 words in the Albanian language, which has taken a long time to provide. We have collaborated with professionals in the field. With all these voices recorded and processed, it is possible to have a coverage of over 80% of the most used words of the Albanian language", says Festimi.
But he points out that all of these resources cannot be provided to users without the use of Machine Learning (ML) and Artificial Intelligence (AI) technologies, which enable us to continuously improve our voice and text-to-audio conversion.
The App itself presents many challenges, ranging from text processing, converting it to something known for algorithm, audio / sound generation, application of machine learning for intonation to the final product.
"The work with Zeri continues, but we decided to publish it in order to make it known as a service and to be used by the people. And why not have a feedback when possible. In this way we can improve this service to best suit your needs," he said.
To convert a text to audio is already very simple. Just create an account in Zeri.app and enter the text you want to convert. You can download the generated voice offline, or you can copy the link and share it, depending on your needs.
0 comments