whisper-large-v3-nepali-ggml

whisper-large-v3-nepali GGML model

This repository contains the GGML format (whisper-large-v3-nepali-ggml.bin) of the fine-tuned Whisper Large V3 Nepali model originally created by kiranpantha.

Model Details

• Original Model Creator: kiranpantha
• Base Model: whisper-large-v3-nepali
• Fine-tuned For: Nepali language speech recognition
• Conversion: This GGML model is converted from the original fine-tuned PyTorch checkpoint (safetensors) to GGML binary format usable with whisper.cpp and compatible apps like VoiceInk.

Usage

This GGML model can be used with whisper.cpp CLI or any compatible Whisper GGML runtime for efficient CPU inference.

Example command:

./whisper.cpp/build/bin/whisper-cli -m whisper-large-v3-nepali-ggml.bin -f input.wav -l ne

Make sure your input audio file is 16kHz, 16-bit mono WAV for best results.

About GGML

GGML is a quantized and efficient binary format designed for CPU-friendly lightweight Whisper model inference.

License

Please refer to the original model license from the kiranpantha Hugging Face repository for license details and usage rights.

---

This GGML file is a conversion artifact and no fine-tuning has been done in this repository. Please support the original model creators!