---
license: apache-2.0
language:
- hi
pipeline_tag: automatic-speech-recognition
---

int8 quantized [ctranslate2](https://github.com/OpenNMT/CTranslate2)-compatible version of [vasista22/whisper-hindi-large-v2](https://huggingface.co/vasista22/whisper-hindi-large-v2).
This means the 5.7GB model is compressed into 1.6GB :).

Model created using

```
ct2-transformers-converter --model /path/to/vasista22/whisper-hindi-large-v2 --output_dir whisper-hindi-large-v2-ct2-int8 --copy_files tokenizer_config.json preprocessor_config.json added_tokens.json special_tokens_map.json --quantization int8
```

For monospeaker audio, use either of

1. [ctranslate2](https://github.com/OpenNMT/CTranslate2)
2. [faster-whisper](https://github.com/SYSTRAN/faster-whisper)

For multispeaker audio with english diarization, use [whisperX](https://github.com/m-bain/whisperX/).

For multispeaker audio with non-english diarization, use [whisper-diarization](https://github.com/MahmoudAshraf97/whisper-diarization/).