How to translate a video with a voice-over
You have a video that works in one language: a product tour, a lesson, a tutorial, a talk over slides. Subtitles reach some viewers in other languages, but many would rather listen. Re-recording the narration means a translator, a voice actor and an afternoon of lining clips up by hand.
There is a shorter way. The video already holds everything a voice-over needs: the words and the second each one was said. Transcribe it, translate the transcript, and voice every line at the time it belongs to. What comes out is one audio track as long as the video, which goes under the picture with nothing to line up.
Here is the whole workflow in Speecho, start to finish.
What you’ll need
- The video file (MP4, MOV or WebM).
- A Speecho account. The first 15 minutes of transcription are free.
- Ten minutes for a ten-minute video, most of it spent reading.
Step 1: Upload the video
Open Transcribe and drop the video in. Your browser takes the sound out of the file and uploads only that, so the video itself stays on your computer. The transcript comes back in a few minutes, in whichever of 99 languages the video is in.
Step 2: Check the transcript
Read it once before anything else. A product name heard wrong, a term written as two words, a number: whatever you fix here carries into the translation and then into the voice. It is far cheaper to correct a word now than to voice a line twice.
Step 3: Translate, and keep it short
Choose Translate and pick the language. The first two translations of a transcript are free.
Then read the translation with one question in mind: will this fit in the time the original took? Translations often run longer than the source. German and French easily take a fifth more words than English for the same sentence, and a line that is longer than its slot has to be said faster or start late. Where a sentence has a shorter way to say the same thing, take it. Everything you see in the translation is what the voice will read.
Step 4: Voice it on the original timing
In the translation’s Voice menu, choose Voice on the original timing. Text to speech opens in Subtitles mode with every line and its time already filled in.
Pick a voice and play it free first. Three things to look at before you press the button:
- Lines that may not fit. The panel names the lines with more words than their time is likely to hold. Shorten them here, or let them be sped up a little.
- More than one person. If the video has two people talking, each gets a voice of their own. Who says each line comes from speech recognition and can be wrong, so check it: the name sits at the start of each line, and changing it moves the line to that voice.
- The price. It is on the button: all the lines are priced together as one text, by their characters.
Each line is voiced on its own and placed at the second it was said. A line that runs over its time is sped up by up to 25%, with the voice keeping its pitch. If that is still not enough, the next lines start a little late until a pause takes the delay up.
Step 5: Listen, then fix single lines
When the last line is voiced, the history card puts the track together and plays it. The card also says which lines were sped up and which started late, and by how much.
To fix a line, open the lines, correct its text or give it another voice, and voice it again. Only that line is charged, and the track is put together again by itself. Every take comes out a little different, so a line that sounds flat the first time often sounds right the second.
Step 6: Put the track on the video
Download the MP3. It starts where the video starts and is as long as the video, so there are three easy ways to use it:
- In the browser. Drop the video and the MP3 into the free replace audio in video tool. It copies the picture as it is and writes only the sound, without uploading anything. You can remove the original sound or keep it quietly underneath.
- In your editor. Mute the original audio and place the MP3 at the very start of the timeline.
- On YouTube. Where your channel can add audio tracks in YouTube Studio, upload the MP3 as the track for the new language, and add the translated subtitles from the same transcript.
What this does not do
It is worth knowing the limits before you start, because they decide which videos it suits.
- Lips. The voice starts at the same seconds as the original speech, but nothing changes the picture. In a close-up of someone talking, the mouth will not match. Tutorials, screen recordings, courses, slides and narrated footage, where a voice explains what is on screen, are where it works best.
- Music and sound effects. The track holds only the new voice. The free tool can keep the original sound underneath at a low level, but the original speech stays in it too. For a clean result, add the music again in your editor.
- Overlapping speech. Two people talking over each other become two lines one after the other.
What it costs
For a ten-minute video with steady narration, about 9,000 characters:
| Step | Credits |
|---|---|
| Transcribe 10 minutes | 100 |
| Translate (first two languages) | free |
| Voice on Standard / Enhanced / Premium | about 450 / 1,000 / 1,450 |
A pack of $5 is 2,400 credits, so a ten-minute video in one more language costs a few dollars, and credits never expire.
The short version
Upload, check, translate short, voice on the original timing, fix single lines, put the track back. The first video takes a little longer because you are learning where the tight lines are; after that, most of the time goes on reading.
If you already have subtitles in the target language, skip the first three steps: open the SRT or VTT file in Text to speech’s Subtitles mode and voice it the same way, as described on the subtitles to voice-over page.
