Translarr
Translarr
Docker app from jesposito's Repository
Overview
Translarr translates subtitle tracks from the wrong language into your preferred language using AI. It plugs into Sonarr, Radarr, Emby, Jellyfin, and Plex via webhooks and drops translated .srt files next to your media.
When a release lands with Russian fansubs on a Japanese anime, or hardcoded Spanish on a Korean drama, Translarr extracts the track, translates it with a context-aware LLM, adjusts timing for the target language's reading rate, and writes a clean .srt next to the video.
Features:
- LLM-powered translation with context awareness (Anthropic Claude, OpenAI, DeepSeek, Gemini, or local Ollama)
- Automatic reading-rate adaptation so subtitles stay readable
- Tag-based opt-in via Sonarr/Radarr: only translates what you tag
- On-demand translation when you press Play in Emby
- 6 translation presets: Quick & Cheap, Balanced, Best Quality, Local & Free, DeepSeek Budget, Gemini Flash
- Cost guards: daily and per-job spend caps
- Push notifications via ntfy.sh
- Web UI: dashboard, job history, settings, library browser, glossary editor
- All settings are live-mutable, no container restart needed
Runtime arguments
- Web UI
http://[IP]:[PORT:9000]/- Network
bridge- Shell
sh- Privileged
- false
Template configuration
Host port to map to container port 9000 (Web UI and API).
- Target
- 9000
- Default
- 9100
- Value
- 9100
Path to your media library (movies, TV shows). Translarr writes translated .srt files next to source videos.
- Target
- /media
- Default
- /mnt/user/media
- Value
- /mnt/user/media
Persistent storage for database and settings. Include this in backups.
- Target
- /data
- Default
- /mnt/user/appdata/translarr
- Value
- /mnt/user/appdata/translarr
anthropic, openai, ollama, deepseek, or gemini
- Target
- LLM_PROVIDER
- Default
- anthropic
- Value
- anthropic
Model identifier (e.g. claude-sonnet-4-6, deepseek-chat, gemini-2.5-flash).
- Target
- LLM_MODEL
- Default
- claude-sonnet-4-6
- Value
- claude-sonnet-4-6
Required when LLM Provider is anthropic. Get one at console.anthropic.com.
- Target
- ANTHROPIC_API_KEY
ISO 639-1 code for target language (en, de, es, fr, ja, etc.).
- Target
- TARGET_LANG
- Default
- en
- Value
- en
Hard kill-switch in cents (1000 = $10). Translations stop until midnight UTC.
- Target
- MAX_COST_CENTS_PER_DAY
- Default
- 1000
- Value
- 1000
Required when LLM Provider is openai.
- Target
- OPENAI_API_KEY
Required when LLM Provider is deepseek. Approximately $0.10 per episode.
- Target
- DEEPSEEK_API_KEY
Required when LLM Provider is gemini. Flash tier is approximately $0.08 per episode.
- Target
- GEMINI_API_KEY
Max spend per single translation in cents (500 = $5).
- Target
- MAX_COST_CENTS_PER_JOB
- Default
- 500
- Value
- 500
How many translations may run in parallel.
- Target
- MAX_CONCURRENT
- Default
- 2
- Value
- 2
Shared secret for arr webhook auth. Empty = no auth (fine on a private LAN).
- Target
- WEBHOOK_SECRET
Emby server URL for library refresh hooks (e.g. http://emby:8096).
- Target
- EMBY_URL
Emby API key for library refresh and the in-player subtitle picker.
- Target
- EMBY_API_KEY
Jellyfin server URL for library refresh hooks.
- Target
- JELLYFIN_URL
Jellyfin API key for library refresh.
- Target
- JELLYFIN_API_KEY
ntfy.sh push endpoint for job notifications. Empty = off.
- Target
- NTFY_URL
Ollama server URL (for the free local provider).
- Target
- OLLAMA_HOST
- Default
- http://ollama:11434
- Value
- http://ollama:11434
Categories
Details
ghcr.io/jesposito/translarr:mainRun Translarr on Unraid.
Translarr is listed in Community Apps for Unraid OS. Explore Unraid to build a flexible home server, NAS, or homelab.