Speech2Text Based on Whisper

This repository provides a fast inference API to complete Speech2Text task.

Model

The base.en from OpenAI is selected as the backend model to do the inference.

For demo purpose, we select LibriSpeech dataset (test-clean) to show workflow.

pip3 install git+https://github.com/openai/whisper.git git clone https://github.com/openai/whisper.git

pip3 install -r requirements

python3 ./whisper_split_gt.py

The script is able to split each file's ground truth and save to path.

python3 ./whisper_infer.py

The script inferences the inputs and generate transcribes.

python3 ./whisper_wer_cal.py

The script is able to calculate output's WER and save results into csv file.

or all in one script

./run.sh

There are several testing data put in ./LibriSpeech folder.

whisper official repo: whisper

Please contact me, if you are interested in this project or have any questions.