Video translation
Translate a video into another language
Video translation covers three different jobs that often get sold as one. This explains which one you are actually buying, what each is good for, and how subtitle translation works here.
Three things people mean by video translation
| Translated subtitles | The original audio stays. Translated text appears on screen. The speaker keeps their own voice, and viewers hear the real delivery. |
|---|---|
| Dubbing | The original audio is replaced with a new voice track in the target language. Expensive to do well, and lip sync is a hard problem. |
| Synthetic voice-over | A generated voice reads the translation over the original. Cheaper than dubbing, but the result sounds generated, which some audiences discount. |
This tool does the first one. Translated subtitles keep the speaker audible and are the option viewers trust most, because nothing about the original performance has been replaced.
How the translation works
Transcribe the speech
Speech recognition writes down what was said, with timings, in whatever language it was spoken.
Translate the transcript
A language model translates the transcript into your target language, working with the surrounding lines rather than one line at a time.
Fit it to the screen
The translated text is split into subtitle lines timed to the speech, since translations rarely come out the same length as the original.
Burn it in
The subtitles are rendered into the picture and encoded to MP4, so they play the same in every player.
Why translated lines rarely fit the original timings
This is the part that separates a readable translated video from an unreadable one. The same sentence takes a different amount of room in every language, and the timing was set by the original speech, not the translation.
- Romance languages such as Spanish, French, Italian, and Portuguese generally run longer than English, so lines get split more often.
- Chinese, Japanese, and Korean pack more into fewer characters, so translated lines are usually shorter than the English they replaced.
- Finnish, Hungarian, and Turkish stack endings onto words, producing long single words that cannot be split across two lines at all.
- German, Dutch, Swedish, Danish, and Norwegian build long compound nouns for the same reason, with the same effect on line breaks.
What you get back
- An MP4 with the translated subtitles burned into the picture.
- An SRT in the target language, ready for anywhere that accepts subtitle files.
- An SRT in the original spoken language, which is worth keeping as the master copy.
- Plain TXT transcripts of both, useful for show notes, articles, and search.
Choosing a target language
There are 30 to choose from, and each behaves differently once it is on screen. Right-to-left languages need the line laid out from the other edge; languages without spaces between words need line breaks chosen by meaning; scripts with stacked marks need more vertical room. See the full list for what changes in each.
Translate your first video with 50 free credits. No card, no install.
Continue with Google50 free credits. No credit card needed.Frequently asked questions
Does it translate the audio or just add subtitles?
It adds translated subtitles. The original audio is untouched, so the speaker keeps their own voice and viewers hear the real delivery.
Do I need to say what language the video is in?
No. The speech is transcribed in whatever language it was spoken. You only choose the language you want the subtitles in.
Can I translate one video into several languages?
Yes. Run the job once per language you need. Each run transcribes the audio again rather than reusing a previous transcript, so budget a review pass for each language you publish.
How accurate is the translation?
It depends heavily on the audio. Clear speech with little background noise transcribes well and therefore translates well. On the Pro and Studio plans you can edit any line before you download.
Can I edit the translation before downloading?
Yes, on Pro and Studio. The subtitle editor lets you change wording and timings, then reburn the video with your corrections.