Hub84/faster-whisper-hinglish-prime

Whisper-Hindi2Hinglish-Prime (Faster-Whisper / CT2 Optimized)

This repository contains a highly optimized, CTranslate2 (CT2) converted version of the original Whisper-Hindi2Hinglish-Prime model, originally developed and open-sourced by Oriserve.

🌟 Acknowledgements & Credit

All core credit for the base architecture, Hindi-to-Hinglish fine-tuning, and original training data belongs entirely to the Oriserve team.

This specific repository simply applies int8_float16 quantization and CT2 conversion to their brilliant work. This conversion compresses the model from ~6GB down to ~1.5GB, allowing it to run natively in Faster-Whisper environments and perform blazing-fast inference on hardware with limited VRAM (such as 4GB GPUs) without crashing.

License: Apache 2.0 (Inherited from the original Oriserve repository)