Whisper Hindi2Hinglish Prime — GGML

GGML (whisper.cpp compatible) conversion of Oriserve/Whisper-Hindi2Hinglish-Prime.

Transcribes Hindi audio directly into romanized Hindi (Hinglish) in Latin script — not Devanagari, not English translation. Hindi-English code-switching is preserved naturally.

Files

File	Format	Size
`ggml-hindi2hinglish-prime.bin`	Q5_1 quantized	1.1 GB

Usage with whisper.cpp

whisper-cli -m ggml-hindi2hinglish-prime.bin -f audio.wav -l hi

Conversion details

Source: Oriserve/Whisper-Hindi2Hinglish-Prime (HuggingFace Transformers safetensors)
Architecture: Whisper large-v3 (32 encoder + 32 decoder layers, d_model=1280)
Conversion: whisper.cpp's convert-h5-to-ggml.py
Quantization: Q5_1 via whisper.cpp's quantize tool

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for curiophile/whisper-hindi2hinglish-ggml

Base model

openai/whisper-large-v3

Finetuned

Oriserve/Whisper-Hindi2Hinglish-Prime

Finetuned

(1)

this model