Where the translation actually runs
In a browser-based translator, the code runs inside your own browser session. Polyglot Tool parses the SRT and executes translation in a Web Worker on your device; the subtitle text does not need to be sent to Polyglot Tool's server.
- The SRT parser and the translator run client-side
- Files are opened and processed locally
- The server never receives subtitle content
What still gets downloaded
Translating requires a language model. Polyglot Tool loads a small model for the languages you choose and caches it in the browser, so the first translation downloads the model and later ones reuse it.
What, if anything, reaches the server
Before translation starts, Polyglot Tool verifies a Cloudflare Turnstile token to protect the endpoint. The server also delivers public configuration. Neither call includes your subtitle file or its text.
- A Turnstile token is checked before each translation
- Public configuration comes from the config endpoint
- No subtitle content is included in these calls
How to verify it for yourself
Open your browser's network panel and run a translation. The only requests you will see are the model download and the Turnstile verification; you will not see your subtitle content leave the browser.
Your operating system's network monitor tells the same story at a lower level. A local translation produces traffic only for models and verification.
The honest limits of local translation
Local translation trades the convenience of a large server fleet for privacy and cost control. Quality depends on the models available for your pair, longer files need more time, and the first run downloads a model.
- Quality varies by language pair
- Large files take longer on your device
- The first translation downloads and caches a model