Skip to content

Latest commit

 

History

History
85 lines (65 loc) · 3.84 KB

README.md

File metadata and controls

85 lines (65 loc) · 3.84 KB
Raw

gutural_nlp

Scream: Fine-Tuned Whisper model for automatic gutural speech recognition 🤟🤟🤟

The objective of this project is to leverage the Whisper state-of-the-art automatic speech recognition (ASR) system developed by OpenAI for automatic speech recognition of gutural and scream sounds mainly present in heavy metal music.

I take advantage of the HuggingFace ecosystem to:

1) Create a dataset using .srt files and audio .wav files collected from Youtube in order to annotate vocal sound extracts.

2) Fine-tune a large version of the Whisper model on the dataset created.

3) Deploy a demo on Space repository using Gradio.

Folder structure

|-- README.md
|-- LICENSE
`-- Notebooks and Scripts
    |-- Dataset Preprocess Notebook.ipynb
    |-- EDA_gutural_nlp.ipynb
    |-- gutural_nlp.ipynb
    |-- SRT_downloader.ipynb
    |-- preprocess.py
    |-- youtubetowav.py
`-- images
    |-- spectogram 1.png
    |-- spectogram 2.png
    |-- spectogram 3.png
`-- Metadata
    |-- Dataset Lookup.csv
`-- Dashboard
    |-- Gutural Dashboard.pbix
    |-- Dashboard SS.JPG
`-- SRT Files

HuggingFace Profile

Dataset

Speech-to-text dataset with +2.5h of annotated audio, comprised of clean vocals and different gutural sounds, including: low, mid, & high fry screams. The data was built with an eclectic sample of artists from varied sub-genres of hard metal music. Some of the artists included are: Suicide Silence, Lamb of God, Cradle of Filth, Cannibal Corpse and much more (full YouTube playlist available below).

Model

Gradio App

Live demo of the ASR engine. With the transcription, I analyze the emotions present in the lyrics to identify: anger 🤬, disgust 🤢, fear 😨, joy 😀, neutral 😐, sadness 😭, surprise 😲 in a user-friendly and interactive spider chart.

YouTube Playlist - Gutural Speech Recognition

The audio files for the dataset can be downloaded using the following command with the python script in this repository:

!python youtubetowav.py https://www.youtube.com/playlist?list=PLkCTyMdVt0AHgp-80jqskjUtfHo-Ht4xy

or using this OneDrive link

Dashboard

Alt text

Spectogram visualization

Alt text Alt text Alt text

License

MIT License

Copyright (c) 2023 Juan Pablo Díaz Pardo

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.