Mizo audio segmented ASR and TTS

About Dataset

Developed by Mizomade Private Limited, this dataset represents a significant effort toward advancing digital speech resources for the Mizo language. The audio recordings are sourced from publicly available open-source resources, and each recording has been carefully segmented into smaller utterances before being paired with accurate transcripts. We have taken great care in segmenting the recordings, preparing and validating the transcripts, normalizing text, and organizing the dataset to ensure high quality, accuracy, and consistency. Multiple rounds of review and quality assurance have been performed to improve transcription accuracy, segment alignment, and overall reliability. While every effort has been made to maintain high standards, speech data is inherently diverse and context-dependent, with variations in pronunciation, accent, dialect, speaking style, speaking rate, and recording conditions. Users are encouraged to perform task-specific evaluation and validation when deploying models trained on this dataset in real-world applications. Our goal is to contribute to the growth of Mizo language technology, promote language preservation and digital inclusion, and foster innovation in speech AI for Mizo-speaking communities.