v1.0.0

First public release. Subtitle Scout has been running on my own NAS for months; this is the point where the docs, the setup wizard, and the deploy path are solid enough to hand to someone else.

What it does

An LLM agent fetches subtitles for a media library. For each file missing a subtitle in your target language, it chooses which sources to search, reads the actual results and reasons about them, verifies each candidate against the video (duration, framing, embedded tracks), and installs the one that fits — or nothing, if nothing does. A second agent translates from another language when no direct source exists. It scans folders directly, with no dependency on Plex, Jellyfin, Emby, Sonarr, or Radarr.

Highlights

• Seven sources, added over time: OpenSubtitles, SubDL, ASSRT, r3sub, Jimaku, SubHD, Zimuku.
• Per-season coverage tracking with a daily re-check, so you can see exactly which episodes are still uncovered.
• Multi-language: deepest coverage for Chinese; English, Japanese, and Korean fully supported; six more configurable.
• Agent-driven install: the repo ships an Agent Skill. Paste one prompt into Claude Code, Codex, or Cursor and it handles deployment end to end. Validated on a wiped Synology with a third-party agent.
• Credentials stay local in the app's own SQLite. Nothing phones home.

Install

git clone https://github.com/fancydirty/subtitle-scout && cd subtitle-scout
mkdir -p cache && cp .env.example .env
docker compose up -d

Prebuilt image at ghcr.io/fancydirty/subtitle-scout:latest. Then open http://<host>:8099 for the setup wizard. Full guide in the README; read-only demo at https://demo.subtitlescout.com.

Built with an LLM in the loop across many iterated sessions. Test coverage backs it: 3,600+ backend tests and 1,200+ frontend tests in CI, with behavioural fixes written red-first.

AGPL-3.0.