原始内容
pi-chinese-motivation
Plays motivational music while Chinese-origin models (DeepSeek, MiMo, MiniMax, Kimi, Qwen, etc.) are thinking — so the longer DeepSeek takes, the more intense the soundtrack gets.
- Starts as a quiet murmur at 15% volume
- Ramps up +4% every 1.5 seconds to 75%
- After ~20 seconds of thinking: 💥 bit-crusher distortion kicks in
- Distortion worsens from 10-bit → 3-bit (completely blown out)
Stops instantly when the model responds. Only mpv is needed for full modulation; ffplay/vlc work as fallback players with cached audio.
Showcase
https://github.com/user-attachments/assets/75bfd71e-ab70-4339-881e-fbb9d25b576a
Install
# From npm (once published)
pi install npm:pi-chinese-motivation
# From git
pi install git:github.com/kaueDM/pi-chinese-motivation
# Local development
pi install ./path/to/pi-deepseek-motivation
Usage
Enabled by default for Chinese-origin models. Disable with CLI flag or slash command:
pi --no-motivation-music
| Command | Effect |
|---|---|
/motivation-music |
Full status |
/motivation-music on/off |
Toggle |
/motivation-music cache |
Pre-download audio to /tmp/pi-motivation-music/ |
/motivation-music url <url> |
Change the YouTube soundtrack |
Prerequisites
- mpv — for YouTube streaming + real-time modulation
- yt-dlp — required by mpv for YouTube; also used for caching
- Optional: ffplay or cvlc — fallback players for cached files
# Ubuntu/Debian
sudo apt install mpv yt-dlp
# macOS
brew install mpv yt-dlp
# Arch
sudo pacman -S mpv yt-dlp
How it works
agent_start fires
├─ mpv starts via IPC socket (--input-ipc-server)
├─ Volume ramp: 15% → 75% over ~22s
├─ After 20s: lavfi-acrusher activates
└─ agent_end → killed instantly
The extension hooks into pi's agent_start/agent_end events and only triggers when the active model matches known Chinese AI providers (DeepSeek, Moonshot/Kimi, Alibaba/Qwen, ByteDance/Doubao, Zhipu/GLM, Baidu/ERNIE, Tencent/Hunyuan, Xiaomi/MiMo, MiniMax, StepFun, 01.AI/Yi, Baichuan, SenseTime).
License
MIT