The project,being part of Kagglex BIPOC Mentorship Program final project, aims to train two separate Hindi ASR models using the Facebook Wav2Vec2 (300M parameters) and OpenAI Whisper-Small models, respectively. The goal is to compare their performance, with a target WER of less than 13%, across various Hindi accents and dialects.

Sound Splitter is a free Windows application (with Python code) that converts a large audio/video file to a low-quality 16bit 16kHz wav or mp3 file and also splits it into several pieces of around 60 MB in size, detecting silent places in the original record.

Transcription of an mp3 file into text (Speech Recognition)

Text-to-Speech (TTS) using python.

Tigerra File Convertion API - NODEJS SDK

One local platform to satisfy remote deeds