sajalmadan0909/

hinglish-stt-tts-deepgram

A Hindi-English code-mixed (Hinglish) speech dataset for automatic speech recognition (ASR) and text-to-speech (TTS) research. The dataset contains 23,543 timestamped speech segments from conversational recordings. Transcript replacement was performed using Deepgram where a non-empty result was available; otherwise, the original transcript was retained.

	Dataset structure




	

Column Type Description

text… See the full description on the dataset page: https://huggingface.co/datasets/sajalmadan0909/hinglish-stt-tts-deepgram.