// the find
Ajatt-Tools/mpvacious
🍜 Adds mpv keybindings to create Anki cards from movies and TV shows.
An mpv user script for language learners doing sentence mining — you watch a show, hit a keybind, and it creates an Anki flashcard with a screenshot, audio clip, and subtitle text. Built specifically for the AJATT (All Japanese All The Time) community but works for any language with subtitle support.
The workflow is genuinely fast once configured: subtitle selection, audio trimming, and Anki card creation happen without leaving mpv. Profiles let you switch between language configs (deck name, note type, fields) mid-session with one keypress. The new note timer is a nice touch — it watches for cards added via Rikaitan or GoldenDict and automatically attaches the audio/image without you pressing anything. Platform support is real: separate platform modules for Linux/Windows, PowerShell install script, explicit handling for Flatpak, portable mpv builds, and iOS Opus container quirks.
Test coverage is effectively nonexistent — the 'test suite' is setting an env var and watching mpv start up, which tests nothing meaningful. The dependency chain is fragile for a video tool: AnkiConnect must be running, curl must be on PATH, clipboard tools differ by display server (xclip vs wl-copy), and if ffmpeg isn't installed the encoder silently falls back to mpv's built-in which may not support your codec. The config file name is still `subs2srs.conf`, a leftover from the predecessor project that causes confusion when searching for docs. Custom subtitle filtering requires creating a dummy empty `main.lua` to prevent mpv from throwing errors, which is a workaround for a missing proper plugin API.