About three to four and a half seconds. That is how long DubMog's natural AI voice took to start speaking after the person on screen started talking, in our own tests this month. The median was 3.3 seconds, both in the Chrome extension and in our next iPhone update. The free Standard mode voiced each phrase about 4 seconds after it began, and it never fell behind: in our runs, no phrase had to wait in line or get skipped.
We build DubMog, so we time it all the time. This post shares what we measured, where the seconds go, what happens when you pause a video or two people talk at once, and how to get the fastest live translation on a computer or a phone.
Why live dubbing can't be instant
A dub can't start before the meaning does. Real-time dubbing is speech-to-speech translation with an AI voice: it has to hear enough of a sentence to know what it says, and languages don't put words in the same order. In German, Ich habe gestern im Park meinen alten Freund getroffen ends with getroffen, "met". An English dub can't say "Yesterday I met my old friend in the park" until that last word arrives.
Human simultaneous interpreters work a few seconds behind the speaker for the same reason. Live AI dubbing does the same job while the video keeps playing: it listens, translates and speaks. What matters is a delay that stays short and steady, so you can follow along without pausing.
Our numbers
| Where | Mode | Delay we measured | Sample |
|---|---|---|---|
| Chrome extension | Pro Live (natural AI voice) | 3.2 to 4.5 s to the first dubbed words; median 3.3 s | 7 sessions, Oct 13–14 |
| Chrome extension | Standard (free) | About 3.9 s from a phrase's first word to its dubbed voice | 43 phrases, 2 sessions |
| iPhone app, next update | Pro Live | Median 3.3 s from each line on screen to its dub | Our standard test clip |
| iPhone app, today's version | Pro Live | Median 4.5 s | The same clip |
| Android app | Pro Live | Same live voice as the iPhone app. We're adding the same stopwatch to Android and will publish its numbers here. | Coming soon |
How we measured. We played everyday conversation videos, mostly Spanish dubbed into English, and logged two moments: when speech began and when the dubbed voice began. The browser numbers come from a Mac running Chrome 156 on home Wi-Fi, the phone numbers from an iPhone 14 Pro. Microsoft Edge runs the same extension and the same live voice. A slower connection, a busier video or a different language pair will move these numbers a little either way, and we'll add runs as we make them.
Where the seconds go
Most of the delay is listening. In the free Standard mode, about three quarters of it went to waiting for the speaker to finish a phrase. The translation itself took a few hundredths of a second, and the voice took about a second to start.
Pro Live works differently. It listens and speaks in one continuous stream instead of phrase by phrase, which is why it starts sooner and sounds more natural. In our next iPhone update we cut about a second from it by letting the voice start earlier, after checking on a real iPhone that it still doesn't stutter.
What happens when you pause the video
Pausing in the middle of a sentence is the hardest moment for live dubbing: half of that sentence has been heard and the rest never comes.
Before our October update to the Chrome and Edge extension, the end of a long sentence could arrive late. You paused, nothing more was said, and when you pressed play the leftover words came out glued to the next line. Version 0.7.12 fixes that. In our test, the dub finished the sentence about 2 seconds after the pause, then went quiet until the video started again.
Two people at once, and other things that trip up a live dub
We listen to our own dubs blind, without knowing which version is playing, and score them. The same few problems show up in every version, because they come from how live speech translation works today:
- Two voices at once. When people talk over each other, the dub follows one of them and can drop words from the other. Lectures, interviews and tutorials, where one person speaks at a time, dub best.
- Lines that mix two languages. A short line that switches language halfway can be heard as the wrong language for a moment.
- Words already in your language. If someone in a Spanish video says a phrase in English and you dub into English, that phrase can be spoken twice.
- The first words after a long silence. After a stretch with no speech, the dub can take a beat longer to catch the first words.
We'd rather tell you than pretend they don't happen.
How to get the fastest, clearest dub
- Use Pro Live when speed matters. In our tests it started about half a second sooner than Standard and sounded more natural. Standard stays free.
- Give it a steady connection. Pro Live speaks from the cloud, so weak Wi-Fi adds delay. The free Standard mode in Chrome hears the video on your computer.
- On a computer, start DubMog on the tab that's playing. Click the DubMog icon on that tab, or right-click the page and choose "Dub this video with DubMog".
- On a phone, pick the mode that can hear the sound. The microphone can't hear a video that plays into your headphones. For a video on the same phone, use Dub a Video or Dub Any App instead of the talk mode.
- In Standard, set the video's language. Standard needs it to listen in the right language.
Live dubbing, subtitles or upload-and-wait?
| Live dubbing (DubMog) | Subtitles | Upload-and-wait dubbing | |
|---|---|---|---|
| When you get it | While the video plays, a few seconds behind | While it plays, but you read instead of listen | After the whole file is processed |
| Works on | Any tab with sound, live streams, apps on your phone | Videos with captions in your language | Files you can upload |
| Best for | Lectures, interviews, live streams, following along | Quiet rooms, exact wording | Finished videos you plan to share |
DubMog does the third one too: Dub a file on iPhone and Android makes a finished dubbed copy you can save. For everything you watch in the moment, a video translator you listen to beats one you have to read. More about speech-to-speech translation →
Quick answers
How long is the delay in real-time dubbing?
In our October 2026 tests, DubMog's natural AI voice (Pro Live) started speaking 3.2 to 4.5 seconds after the person on screen, with a median of 3.3 seconds in Chrome and in our next iPhone update. The free Standard mode voiced each phrase about 4 seconds after it began. Your connection and the video move these numbers a little.
Is real-time translation instant?
No. Any live speech translation has to hear enough of a sentence to know what it means, and languages put words in different orders. Human simultaneous interpreters work a few seconds behind the speaker for the same reason.
Which is faster, Standard or Pro Live?
Pro Live. In our tests it started about half a second sooner than Standard and sounded more natural. Standard is free.
Can I dub a live stream?
Yes. The DubMog extension for Chrome and Microsoft Edge dubs any tab that plays ordinary sound, including live streams, webinars and lectures, while they happen. Some players protect their audio (DRM) and deliver silence to extensions; DubMog tells you when that happens.
Can I translate a YouTube video into English in real time?
Yes. Open the video in Chrome or Edge, click the DubMog icon or right-click the page and choose “Dub this video with DubMog”, pick English and press Start. On iPhone, Dub Any App dubs the video while it plays in the YouTube app.
Does it work on iPhone and Android?
Yes. DubMog is free on iPhone and Android with the same account and minutes as the extension, and Pro Live uses the same natural AI voice everywhere.
Try it and time it yourself
DubMog is a free AI video dubbing extension for Chrome and Microsoft Edge, and a real-time translator app on iPhone and Android, all on one account. Want the course version of this guide? Read how to dub Udemy and Coursera courses live.
DubMog is also free on iPhone and Android, with the same account.