Superkaraoker
Makes karaoke from any youtube video link. The method is based on machine learning methods. After dowloading video from Youtube, the audio is separated by ffmpeg, then spleeter extract voice from recording and whisper convert audio to subtitles. At the end subtitles are hardcoded to the video with extracted audio. Works with any language.
Install / Use
/learn @michalvavrecka/SuperkaraokerREADME
Superkaraoker 0.0.1
Automatic karaoke generator based on recent neural nets. Converts any youtube link to music video with extracted vocals while hardcode big subtitles. Multilingual vocal autodetection.
Installation
Linux
Clone the repository and make output directories:
git clone https://github.com/michalvavrecka/superkaraoker.git
cd superkaraoker
mkdir karaoke & mkdir output & mkdir original
Install dependencies and make shell scripts executable:
pip install -r requirements.txt
sudo apt install ffmpeg
sudo chmod +ax ./superkaraoker
sudo chmod +ax ./multikaraoker
Mac
Clone the repository and make output directories:
git clone https://github.com/michalvavrecka/superkaraoker.git
cd superkaraoker
mkdir karaoke & mkdir output & mkdir original
Install dependencies and make shell scripts executable:
pip install -r requirements.txt
brew install ffmpeg
chmod 755 ./superkaraoker
chmod 755 ./multikaraoker
Windows
Install Python from store, download and install FFmpeg then:
git clone https://github.com/michalvavrecka/superkaraoker.git
cd superkaraoker
pip install -r requirements.txt
Then you can try superkaraoker.bat
Requirements
FFmpeg runs on CPU but GPU is recommended.
Spleeter runs on CPU but GPU version is faster.
Whisper and Whisperx requires at least 2GB VRAM for smallest model, 6GB VRAM for medium model. For the best results 12GB GPU is recommended.
Usage
Convert one song
./superkaraoker https://youtu.be/fUkBPG6viPI Mindslip_Broken_picture_frames
Convert multiple songs
./multikaraoker mysongs.txt
Example
Karaoke (copyright free music video)
https://user-images.githubusercontent.com/16919032/229085294-edb4509e-deeb-4b0b-b43b-e22c81c7268a.mp4
Original video with hardcoded subtitles (copyright free music video)
https://user-images.githubusercontent.com/16919032/229086837-7cf86220-c6e8-4f99-9d2f-82653a87fa00.mp4
Modification
If you have small GPU change the model size in superkaraoker script to --model medium (6GB) or --model small (2GB)
If you want to force the language add --language parameter to whisper command inside superkaraoker
TODO
Automatic subtitles grammar correction
Better subtitles timing with whisperx (work only for specific languages)
License
Copyright © 2023 Michal Vavrecka. Released under the MIT License.
Related Skills
qqbot-channel
352.9kQQ 频道管理技能。查询频道列表、子频道、成员、发帖、公告、日程等操作。使用 qqbot_channel_api 工具代理 QQ 开放平台 HTTP 接口，自动处理 Token 鉴权。当用户需要查看频道、管理子频道、查询成员、发布帖子/公告/日程时使用。
docs-writer
100.7k`docs-writer` skill instructions As an expert technical writer and editor for the Gemini CLI project, you produce accurate, clear, and consistent documentation. When asked to write, edit, or revie
model-usage
352.9kUse CodexBar CLI local cost usage to summarize per-model usage for Codex or Claude, including the current (most recent) model or a full model breakdown. Trigger when asked for model-level usage/cost data from codexbar, or when you need a scriptable per-model summary from codexbar cost JSON.
arscontexta
3.1kClaude Code plugin that generates individualized knowledge systems from conversation. You describe how you think and work, have a conversation and get a complete second brain as markdown files you own.
