Translate Subtitles to Spanish

Spanish usually runs 15 to 25 percent longer than the English it comes from, and subtitles do not get extra time for it. A line that sat comfortably on screen for two seconds in English may be too dense in Spanish.

So each translated cue is measured against about 17 characters a second, the common guideline for Spanish subtitles, and lines longer than 42 characters are flagged. Only lines the translation made harder to read are marked, so you are not sent to fix what was already long in your original. Shorten those in place, often by dropping a pronoun or choosing a shorter verb, and the download includes your wording.

The translation uses neutral Spanish that reads naturally in Spain and Latin America, with opening ¿ and ¡ marks and speaker dashes kept at the start of each speaker's line. If your audience is in one country, read the lines with local vocabulary (car, computer, bus) and adjust the few that matter.

Upload a subtitle file and it comes back in another language, cue by cue. Every cue keeps its number, its start and end time and any position settings, so the translated file drops onto the same video with nothing to re-time.

You check the result on this page before you download anything. The original sits beside each translation, and any line can be edited in place. Your edits go into the file you download.

It works on the SRT or VTT file you already have: from YouTube Studio, from a transcription tool, from an editor such as Premiere, Resolve or Subtitle Edit, or from a video you were sent. There is no video upload and nothing to render.

Only the words are translated. Everything a player or an editor depends on is kept out of the translation and written back verbatim:

• Cue numbers and VTT cue identifiers.
• Start and end times, to the millisecond, including overlapping cues and long gaps.
• VTT cue settings (position, line, alignment, region) and SRT position coordinates.
• Italics, bold, colours and positioning codes such as {\an8}, which stay around the words they belong to.
• VTT header lines, STYLE, REGION and NOTE blocks. The header's Language line is updated to the new language.
• The file's line endings and byte-order mark.

Before a result is shown, the translated file is parsed again and compared with the original: the same number of cues, in the same order, with the same identifiers and timings. If anything disagrees, the job stops and nothing is charged.

Subtitles split sentences across cues, and word order differs between languages, which is where most subtitle translation goes wrong. Each group of cues is translated with its neighbours visible as context, so a sentence split over three cues is translated as a sentence and then kept in three pieces, each in its original cue and timing.

Every reply is checked: each cue must come back exactly once, and no cue may be invented or merged away. A reply that fails the check is asked for again, in smaller groups if needed.

Keeping the timings does not make a longer translation comfortable to read. Spanish, French and German usually run longer than English, and a line that fitted in two seconds may not fit in the same two seconds translated.

So the result marks the lines worth a second look: cues that read faster than about 17 characters a second (20 for English), lines over 42 characters, and cues that grew past two lines. Lines that were already long in your original are not flagged. Use "Show only lines to check" to go straight to them, then shorten the wording in place.

The flags are guidance, not rules. A two-word reply that runs slightly fast is fine; a dense line of dialogue over a quick cut is not.

Translate into English, Spanish, French or German. The language of your file is detected from its text, so there is nothing to set on that side.

The list is short on purpose. A language is added once its translations have been read by a fluent reviewer, rather than because a model can technically produce it. If you need another language, email hello@sonomir.com: requests decide what comes next.

Ten minutes of subtitles a day, across up to three files, are free, and so are the downloads. A ten-minute video's subtitles translate completely at no cost.

A longer file is translated up to the free limit, stopping at the end of a whole cue, never mid-sentence. You see that part, can edit it, and can download it as a shorter file that is still valid. The page shows the full price before anything is charged.

Beyond that, it is 1 credit per minute of subtitles, measured from the first cue to the last. A 45-minute episode costs 35 credits after the free 10 minutes. Credits are bought once, never expire, and work on every tool here. A job that fails is not charged.

It does not change timings, re-split cues, dub audio or burn subtitles into a video. It does not convert SRT to VTT or back, because each format carries things the other cannot, and dropping them silently would be worse than not offering it.

Files must be UTF-8 text, up to 2 MB, 10,000 cues and 100,000 characters. An old SRT saved in another encoding needs re-saving as UTF-8 first; the page says so when that is the problem, and names the line when a timestamp is broken.

Machine translation needs a human pass before anything is published. Check names, jokes, and anything where tone matters.