zsxkib / create-rvc-dataset

Create your own Realistic Voice Cloning (RVC v2) dataset using a YouTube link

  • Public
  • 4.6K runs
  • GitHub
  • Paper

Input

Output

Run time and cost

This model runs on Nvidia A40 (Large) GPU hardware. Predictions typically complete within 79 seconds. The predict time for this model varies significantly based on the inputs.

Readme

RVC v2 Dataset Creation Tool

Introduction

Create vocal datasets for Realistic Voice Cloning (RVC) v2 models with ease. Simply provide a YouTube video URL and let the tool handle the extraction and preparation of vocal data, ideal for training sophisticated voice cloning models. ๐Ÿง ๐ŸŽค

Features

  • Easy Input: Paste the URL and optionally name your dataset. ๐Ÿ“Œ
  • Automated Processing: Isolates vocals and segments them into clips. โš™๏ธ
  • Immediate Download: Access your dataset with a click. ๐Ÿ’พ

How to Use

  1. YouTube URL: Input the link to the video. ๐Ÿ–‡๏ธ
  2. Dataset Name: Choose a name or stick with the default. โœ๏ธ
  3. Run: Hit ‘Run’ to begin the magic. ๐Ÿš€
  4. Download: Grab your .zip file full of vocal clips. ๐Ÿ“ฆ

What You Get

  • The output is a .zip file titled dataset/<your_dataset_name>/.
  • It houses split_<i>.mp3 files, each containing a clear vocal extract. ๐ŸŽถ

Training Your Model

Once you have your dataset, head over to https://replicate.com/zsxkib/train-rvc-v2 to train your RVC v2 model on your newly created dataset. ๐Ÿš€

Acknowledgments

This tool is adapted from work by the talented ardha27, who authored the initial codebase available here. ๐Ÿ™