Audio tools and resources

Guguță (An Artificial Intelligence-Based Educational Platform)

Young Researchers · 25.80012.0807.21TC · 2025–2026 · Project leader: Tudor Bumbu

Tools, models and datasets for Romanian speech recognition and synthesis.

Tools	TTS/STT Dataset Builder (create and manage datasets) Voice Recorder (record speech in your browser)
Models	Whisper large-v3-turbo (STT, Hugging Face) XTTS-v2 (TTS, Hugging Face)
Datasets	Corpus (Hugging Face) Project datasets (Google Drive) (over 1,000 hours, including 200 hours recorded in a studio as part of the project)

Tools

TTS/STT Dataset Builder (create and manage datasets)

Voice Recorder (record speech in your browser)

Models

Whisper large-v3-turbo (STT, Hugging Face)

XTTS-v2 (TTS, Hugging Face)

Datasets

Corpus (Hugging Face)

Project datasets (Google Drive) (over 1,000 hours, including 200 hours recorded in a studio as part of the project)

Paper

Marius Cerescu, Alexandr Parahonco, Olesea Caftanatov, Tudor Bumbu, Nichita Degteariov, Ion Bostan.
Fine-Tuning Neural Speech Models for Romanian Speech Recognition and Synthesis with Moldovan Dialectal Data.
Proceedings of International Conference on System Analysis & Intelligent Information Technologies (SAIIT2026), 20–22 October 2026, Chișinău.