Speech Slicing

A repository for extracting setence-by-sentence speech files from audio files or video files. This repository utilizes Whisper V3 and GPT-4o-mini to extract speech files.

---

Usage

Environment

Install torch 2.0+ compatible to your environment, e.g.:

conda install pytorch==2.3.1 torchvision==0.18.1 torchaudio==2.3.1 pytorch-cuda=12.1 -c pytorch -c nvidia

or

pip install torch==2.3.1 torchvision==0.18.1 torchaudio==2.3.1 --index-url https://download.pytorch.org/whl/cu121

Install ffmpeg. If you are using conda, run

conda install conda-forge::ffmpeg

Install the other dependencies:

pip install -r requirements.txt

OpenAI API Key

Fill api_key.txt with your own OpenAI API Key.

Inference

python slice_speech.py --input_path ./YOUR_INPUT_PATH --extension YOUR_FILES_EXTENSION

For further details, check the help messages by python slice_speech.py --help.

Implementations

• Audio Slicing with Sliding
• Recursive Whisper
• LLM Merging
• Postprocessing with length and VAD