He switched automatic dubbing on across all his videos in one evening. In the first week views rose, and he was pleased.
Then he read a comment in English mocking a line he had said about a neighbourhood in his city: the translation had rendered the neighbourhood's name literally, turning it into an odd phrase that meant nothing. Afterwards he opened the analytics and found that watch time in the new language was far below the Arabic, which is to say people were opening and leaving.
Views rose and the channel gained nothing.
Three different levels, not one thing
Before any decision, separate them, because each has its own cost and its own risk.
1. Subtitles. A written subtitle file in another language, with the original audio untouched. The least risky of the three and the easiest to fix, because the error is visible and can be corrected in a minute.
2. A read-aloud voiceover. A synthetic voice reading the translation over the video. It solves the problem for people who do not read, but it loses your delivery and makes the video feel commercially automated.
3. Dubbing in a clone of your voice. It says your words in another language in something like your own tone. The most effective and the most dangerous, because a mistake in it appears to come from you.
Most of the people who regret it today started at the third. The right order starts at the first.
Before anything: who would you be serving?
Do not translate because the option exists. Ask one question: is there a real audience for this content in that language?
And look at your own numbers first: do you already get views from countries that do not speak your language? Have comments arrived in another language? Does your subject matter to people outside the region? A film about a place in the Arab world interests non-Arabs; a video about local tools or local prices interests nobody abroad.
And the practical ways of learning what your audience wants, current and potential, are in how do you find out what your audience wants.
Start with subtitles
This is the most useful step and the cheapest, and it has a benefit everyone overlooks: subtitles help your existing audience too. Many people watch with the sound off, or in public, or at low volume, and subtitles are what let them finish.
And the rules for on-screen text, how lines break in Arabic, and what works and what does not are all in on-screen text and subtitles.
The practical order: subtitles in your own language first, then English, then a third language if you have a real reason.
What machine translation always ruins
This list is constant across every tool, so check it on every video:
- Names of people and places. Translated literally they become odd phrases. Keep a list of names that must stay as they are.
- Proverbs and idioms. An idiom rendered literally means nothing, and sometimes means something embarrassing.
- Jokes and irony. They lose their timing and become either a flat sentence or an incomprehensible one.
- Technical terms. The word you use in your field has an agreed equivalent, and the tool may pick a different one.
- Forms of address. The line between formal and informal address differs between languages, and tools get it wrong.
- Figures, currencies and units. Watch currencies especially: a figure left in a local currency with no conversion confuses a foreign viewer.
Three places to always check yourself
You do not need to review the whole video, only three places that set the impression.
1. The title and description. These decide whether anyone opens the video at all. A literally translated title looks strange instantly, and nobody opens what looks strange. Write a title in that language fresh rather than translating one.
2. The first minute. If the error is at second twenty, the viewer has left and never learned that the rest was good.
3. The list of names and terms. You build the list once, then use it on every video.
And if you do not know the language, get somebody who does to review only those three. Twenty minutes of a human being rescues the whole video.
A clone of your voice: questions before the technology
Cloning your voice is not only a technical decision. Ask three questions before you do it.
Who owns the voice afterwards? Read the tool's terms: what it does with the voice model, whether it keeps it, whether it uses it in training, and how it is deleted.
Is it clear to the viewer that this voice is generated? Disclosure protects your audience's trust. And anything that looks like you said it yourself when you did not harms you if it comes out against your will.
And other people's voices? Do not clone the voice of somebody who appeared in your film without written permission for that specifically; their consent to being filmed is not consent for their voice.
How to tell whether it worked
Do not judge by view count in the new language, which is a misleading figure. Look at three things:
- Watch time in the new language against the original. If it is far lower, the translation is bringing people who do not stay, which does not help you.
- Comments in that language. One comment engaging with something in the video is worth more than a thousand silent views.
- Did subscribers from those countries increase? Passing views do not build a channel.
And if the numbers are weak after a reasonable number of videos, stop without embarrassment. The decision is not permanent.
When not to translate
- Purely local content: prices, tools, or services in your country.
- A video that rests on wordplay, poetry, or dialect.
- A video full of local names that mean nothing outside the place.
- A channel that has not settled yet. Steadying the base matters more than doubling the fronts.
In short
Subtitles are a step with almost no risk, so start there today. Dubbing is a powerful tool that needs human review in three places, and needs a real reason rather than the mere existence of the button. And where AI helps across the rest of your work and where it ruins it is set out in where AI helps in video making and where it ruins it.