Zevo Tech, Orange Fab startup, develops products for speech transcription, synthesis and processing

Zevo Tech, Orange Fab startup, develops products for speech transcription, synthesis and processing

Using artificial intelligence and machine learning algorithms, Zevo Tech has developed products to help in tasks such as transcribing conversations and identifying speakers, synthesizing speech from text written based on human voice templates, or identifying emotions and feelings.

Started as a spin-off from the “Speech and Dialog Research Laboratory” of the Polytechnic University of Bucharest, the startup Zevo Technologies has a team of professors and doctoral researchers in artificial intelligence who have published over 100 scientific articles and won multiple awards for speech processing applications.

We talked to Horia Cucu, co-founder and CEO of Zevo Tech about the startups current products and future plans.

The beginnings of Zevo Tech's collaboration with Orange Fab

We had our first contact with the Orange team in the summer of 2021 when we I submitted a proposal for a research and development project in the consortium Politehnica University of Bucharest (UPB) - Zevo Tech - Orange. Although the project proposal was not funded, the Zevo Tech application for Orange Fab resulted from the collaboration and so here we are.

The team behind Zevo Tech

Horia, the leader of Zevo Tech and its team of young researchers, finished his PhD in speech recognition 10 years ago, the result being the first automatic speech transcription system dedicated to the Romanian language. Alexandru Caranica got his doctorate in the same field, now being in charge of Cloud security and dev ops. The youngest of the three, Lucian Georgescu, is the main developer of speech and speaker recognition components. He will finalize his doctorate in speech recognition in 2022.

Zevo Tech products and the target audience of the startup

Zevo Tech’s main product is Zevo Speech Transcripts, a real-time speech transcription solution that works with an accuracy between 90% and 98%, depending on the speech domain, addressing primarily companies with Call Center departments or Call Center system integrators. In this industry, Zevo Speech Transcripts can be used to transcribe all telephone calls and generate reports on recurring customer complaints, customer satisfaction, or operator efficiency.

Zevo Speech Transcripts can be used to deploy intelligent telephone robots that interact with customers by voice. For this second field of use, the product Zevo Text-To-Speech (Zevo TTS) is also required, a product that can be used to generate natural speech from the text.

Zevo Speech Transcripts is aimed at both companies that want to transcribe on-site or online team meetings, as well as public institutions such as courts and local councils. Zevo Speech Transcripts can also be used in other areas, such as media monitoring, indexing and searching in multimedia databases, transcription of interviews, reports, and press conferences. For such scenarios, the product has also been expanded to have a user-friendly graphical interface where the user can view, correct, and download the transcript of the multimedia materials he wants to process.

Zevo Speech Dictation is another product of Zevo Tech that can be integrated into various applications that require dictation functionality. This functionality is useful primarily to doctors who could complete medical records with voice notes during the consultation.

The future of Zevo

In Zevo’s future plans we include the development of the B2B call center and media monitoring direction to adapt the application to concrete market needs and to obtain recurring revenues that can fuel the continuous development of products.

Find out more about Zevo Tech from the ZF Tech Day show (a discussion in Romanian):