Scope & limitations
This page exists so you can decide against Sublarr quickly, if that is the right call. Everything below is a current constraint, not a roadmap teaser.
Is Sublarr for you?
Section titled “Is Sublarr for you?”You probably don’t need Sublarr if
Section titled “You probably don’t need Sublarr if”- Your library is English-only live-action and Bazarr already covers it. Bazarr is mature, has more hands on it, and does that job well. Replacing a working setup with a younger tool buys you very little. A longer side-by-side is at Sublarr or Bazarr.
- You never touch anime. Absolute episode numbering, AniDB resolution, ASS/signs handling and dubtitle detection are where most of the engineering went. Without anime you are carrying machinery you will not use.
- You want subtitles fetched once and never thought about again. Sublarr rewards being tuned — language profiles, scoring, automation rules. Set-and-forget works, but a simpler tool would too.
- You need real multi-user access control today. Sublarr has one shared password, not accounts. See Authentication below.
Sublarr’s value compounds if
Section titled “Sublarr’s value compounds if”- You run anime, especially releases with absolute ordering, signs/songs tracks, or several English subtitle tracks per file — where picking the wrong one desyncs what you read from what you hear.
- You care about the format, not just the presence, of a subtitle: ASS with its styling intact rather than a flattened SRT.
- You maintain more than one language and want per-language rules instead of one global setting.
- No subtitle exists at all for what you want, and you would rather have a machine translation you can inspect than nothing.
- You want to know why a particular file was chosen, and to be able to change the answer.
Known limitations
Section titled “Known limitations”Translation is experimental
Section titled “Translation is experimental”LLM translation is off by default and stays labelled experimental. Quality varies sharply by backend, model and language pair — the cloud backends (DeepL, Claude, Gemini) are the reliable end, local Ollama models are usable but need tuning. Never point it at a library you have no backup of.
Authentication
Section titled “Authentication”Sublarr’s UI auth is a single shared password — no usernames, no user accounts, no roles, no per-user audit trail. That is adequate for a single-tenant box on your own LAN and nothing more.
For genuine multi-user setups, put Sublarr behind a reverse proxy that
authenticates for it. Sublarr reads proxy auth headers natively (Authelia,
authentik, and anything else that sets a trusted Remote-User-style header) —
see Reverse proxy and the
proxy_auth_enabled setting.
One instance per library
Section titled “One instance per library”The scheduler assumes a single primary. If you run additional replicas
against the same database, every replica must set
SUBLARR_SCHEDULER_ROLE=disabled or each recurring job fires once per replica.
There is no worker pool and no multi-host distribution — Sublarr ships as one
container per library.
No auto-update
Section titled “No auto-update”By design. The UI tells you when a newer release exists; you decide when to run the upgrade. Read the CHANGELOG before you do — this project moves quickly.
Docker is the supported path
Section titled “Docker is the supported path”Running from source works and is documented for development, but bare-metal installation is not a path that is tested or supported. If you are not running Docker, you are on your own.
Provider coverage depends on your accounts
Section titled “Provider coverage depends on your accounts”Of the 28 providers, several need credentials or an API key. With none configured you are effectively running a much smaller subset than the number suggests. See Providers.
Anime-first scoring defaults
Section titled “Anime-first scoring defaults”ASS receives a large scoring bonus over SRT. That is deliberate and correct for anime; on an English live-action library it mostly means the scoring has less to say about which candidate wins.
Solo-maintained
Section titled “Solo-maintained”One person builds, tests and answers issues here. Bug reports get read; response time varies with real life. Discord is usually the fastest route.
These docs are written by hand
Section titled “These docs are written by hand”They can drift. If the documentation and the running app disagree, trust the app — and please open an issue or say so on Discord so it gets fixed.