29 provider adapters
- 21 source services + embedded extractor + 7 Subliminal wrappers
- ASS-first bonus (+50) and anti-MT penalty (−40) in the score
- SHA-256 deduplication catches identical files
- AniDB resolution for anime with alternate episode order
Sublarr is now 1.10.1 — stable. Subtitle search, scoring and *arr integration are production-ready; LLM translation stays experimental.
v1.10.1 released
Follow on GitHub →
v1.10.1 · 29 provider adapters · 12 translation backends · OpenAPI live
Subtitle automation for
TV, film and anime.
Sublarr runs in Docker on your own server and works with Sonarr,
Radarr, Jellyfin and Emby: search, score, sync, clean up.
Built for the hard cases — absolute episode numbering, ASS styling,
fansub groups. Whatever handles those handles the rest of your
library without breaking a sweat.
Docker
docker run -d -p 5765:5765 ghcr.io/abrechen2/sublarr:latest
localhost:5765 — Library
29
Provider Adapters
63
Subtitle Languages
0
Cloud Calls Unless You Enable Them
GPL-3.0
Open Source
Features
Everything your subtitles need.
Built for power users with self-hosted media stacks.
Multi-engine sync
- Fallback chain: ffsubsync (audio-feature) + alass (alignment)
- Auto-sync after every download
- Large drift (>2s) → alass first, fallback ffsubsync
- Per-engine audit log with score + correction
Waveform editor
- Audio waveform with snap-to-wave (frame-accurate)
- ASS/SRT syntax highlighting in the cue editor
- Active-cue bar + region labels above the waveform
- Auto-repair on save: BOM, encoding, cue overlap
*arr integration
- Sonarr & Radarr webhooks: Download events
- Multi-instance: multiple Sonarr/Radarr/Jellyfin/Emby in parallel
- Tag-based language profiles (tag "anime" → DE+EN)
- Library refresh via API for Jellyfin & Emby
Wanted scanner
- Event-driven with optional periodic scans
- Batch search with backlog reserve against quota bursts
- Slow-mode backoff for unfindable items (4h → 24h)
- Subtitle upgrade job: better ASS to replace existing SRT
Reliability
- Per-provider circuit breaker (CLOSED → OPEN → HALF_OPEN)
- State persists across container restarts
- Budget manager with 429 backoff + refunds
- Multi-key pool: rotates API keys before hitting limits
Library & sidecar tools
- Sidecars as first-class citizens on detail pages
- Inline HI removal, timing shift, format convert
- Embedded track extraction via ffmpeg
- Trash with 30-day retention instead of hard delete
Localization
- Full UI in German + English with one click
- 63 subtitle languages in language profiles (up from 20)
- Per-profile cutoff stops searches when target hits
- audioExclude skips items with matching audio language
Cleanup operations
- 6 ops: language, format, orphan files & DB, backups, foreign tracks
- Dry-run preview shows every file before apply
- Trash with 30-day retention — restore supported
- Path-mapping aware, scheduler-integrated
Optional translation Experimental
- 12 backends: Ollama, DeepL, Claude, Gemini, OpenAI, DeepSeek, …
- Live queue with per-backend cost tracking
- Off by default — opt in via Settings → Translation
- Local models experimental, cloud backends more reliable
Screenshots
See it in action.
A look at the real Sublarr interface.
localhost:5765 / dashboard
Workflow
From download to subtitled — automatically.
01
Connect your stack
Link Sonarr, Radarr, Jellyfin or Emby. Sublarr receives webhooks whenever new media arrives.
02
Search & Score
29 provider adapters searched in parallel. Subtitles scored, deduplicated and ranked — ASS preferred.
03
Translate & Sync
Optionally translate with a local LLM, sync timing with ffsubsync, then deliver to your media server.
Installation
Up and running in minutes.
# Pull and run
docker run -d \
--name sublarr \
-p 5765:5765 \
-v /path/to/config:/config \
-v /path/to/media:/media \
ghcr.io/abrechen2/sublarr:latestThen open http://localhost:5765 in your browser — or read the Installation Guide for the full setup.
Trust
What Sublarr does with your files — and what it doesn't.
A subtitle manager reaches into your media library. So here is what happens, without the marketing voice.
Your videos never leave the server
Hash-based search computes a checksum locally and sends only that — never the file itself.
No telemetry without consent
Anonymous usage statistics are opt-in. Nothing is sent until you actively agree, and you can withdraw at any time.
Deleted is not gone
Removed subtitles go to a trash folder and stay restorable for 30 days. No immediate hard delete.
Preview before it acts
The invasive tools — cleanup, repair, the automation dry run — show you what they would change first.
Hardlinks stay intact
Files hardlinked into your torrent client are skipped on remux by default — rewriting them would break the seed.
Cloud only if you want it
Translation is optional and off. If you use it, you choose between local Ollama and cloud services.
Support
Like Sublarr? Buy me a coffee.
Sublarr is free and open source — built and maintained in my free time. If it saves you hours of subtitle hunting, a small donation goes a long way to keep the project alive.
- Keeps the project maintained and updated
- Funds server costs for the translation model
- Motivates new features and bug fixes
No account required. Any amount appreciated.
Documentation
Everything you need, in the docs.
From your first docker compose up
to the plugin API — the docs are part of the project.
Open source, versioned, no tracking.
-
Getting Started
Installation, onboarding and first run.
-
User Guide
Dashboard, library, wanted list and activity.
-
Editor & Sync
Edit subtitles, waveform sync and batch processing.
-
LLM Translation
AI-powered translation with cost and glossary management.
-
Settings
Providers, languages, scoring, automation and more.
-
Tools & Integrations
*arr integrations, plugin marketplace, AniDB and more.
-
Troubleshooting
When something breaks — common issues and performance tuning.
-
Development & API
Architecture, REST API, database schema and plugin development.
Full overview in the docs — searchable, with sidebar and full-text search.