---
license: mit
pipeline_tag: automatic-speech-recognition
library_name: whisper
---

# OPENVINO Model Decoders for @ggerganov's whisper.cpp ASR

This repository extends the [existing one](https://huggingface.co/Intel/whisper.cpp-openvino-models) providing a limited variety of models by Intel.

The repository contains: 
- Pretrained model files in GGML format as the base format to load in [whisper.cpp](https://github.com/ggerganov/whisper.cpp)
- OPENVINO decoder binary files converted with original OpenAI's [Whisper](https://github.com/openai/whisper) pipeline
- OPENVINO handler files in XML format

The conversion was made with original OPENVINO 2024.0.2.20231213 and modified [Whisper](https://github.com/openai/whisper) installation compatible with the latest PyTorch.
ASR models were tested on the same OPENVINO version and Intel Core i7-6700 CPU @ 3.40GHz and original [whisper.cpp](https://github.com/ggerganov/whisper.cpp) compiled with OPENVINO support turned on.