Subtitle Translator

Why translating a subtitle file is more than running text through a translator

When you localize a video for a new audience, the captions have to land on screen at exactly the moment each line is spoken. Paste an .srt into a generic document translator and you get back translated prose with the timing destroyed — the cue numbers, the --> timing lines and the blank-line structure that a video player reads are gone, and you are left rebuilding the file by hand. This subtitle translator keeps the file intact: it translates only the spoken text of each cue and returns a real, playable SRT or VTT with every index and every timestamp exactly where it was — so a client deliverable comes back ready to use, not ready to re-time.

Timestamps and cue structure survive — and we re-check them

The model is instructed to translate the words and leave every index and timing line byte-for-byte unchanged. Then, before you ever see the result, the translated file is re-parsed and re-validated with the same subtitle engine our SRT toolkit uses, so what you download is a clean, correctly formatted file rather than a best guess. Here is what that looks like for a single cue:

Before (English SRT):

12
00:01:04,200 --> 00:01:06,800
We ship to over forty countries.

After (translated to Spanish):

12
00:01:04,200 --> 00:01:06,800
Enviamos a más de cuarenta países.

The index (12) and the timing line (00:01:04,200 --> 00:01:06,800) are identical; only the spoken line changed. That is the whole point — the captions still sync to the picture.

Reading speed and line length change between languages

A translation is rarely the same length as the original. German and Russian often run longer than English; Japanese, Korean and Chinese pack more meaning into fewer characters. Because each cue keeps its original on-screen duration, a longer translation raises the characters-per-second (CPS) a viewer has to read in that fixed window. The tool is told to respect subtitle conventions — keep lines short, around 42 characters per line where the language allows — so the result stays readable. After translating, it is worth a quick pass over any cue that grew a lot; if a few are too dense, shortening the wording (not the timing) fixes it.

SRT and VTT, in and out

Both common caption formats are supported. SRT (SubRip) uses a comma before the milliseconds (00:00:02,000) and numbers every cue; WebVTT — the format an HTML5 <track> element needs — starts with a WEBVTT header and uses a dot (00:00:02.000). By default the translator returns the same format you gave it, but you can ask for the other and translate and convert in one step, with the timestamps carried across unchanged.

Supported languages

You can translate into Spanish, French, German, Italian, Portuguese, Dutch, Polish, Russian, Arabic, Hindi, Japanese, Korean, Chinese (Simplified) and English. The source language is detected automatically, so you only pick the language you are translating into. One tool with a language selector covers the whole set — there is no separate page per language to hunt for.

How much you can translate at once

Translation runs on AI, so each job uses a small amount of AI credits and is sized to one clean pass. Without an account you can translate files up to 100 cues or 8,000 characters; a free account raises that to 500 cues or 40,000 characters. For a feature-length file beyond those limits, split it into parts and translate each — every part comes back with its timings intact.

Private by default

Your subtitles are processed securely to perform the translation and are not stored afterwards, and you do not need an account to try it. See our privacy policy and terms for the details.