Our YouTube channel, BizMate AI, teaches AI tools in Vietnamese. It has 8,800+ subscribers, and 95.6% of its viewers are in Vietnam. In October 2026 we decided to take it global without opening a second channel: same videos, same voice, plus AI-dubbed audio tracks in other languages. This case study shows where we started, what YouTube’s free auto-dubbing actually did for a Vietnamese channel, what broke, and what our first paid dubbing test cost. We update it with real numbers as the experiment runs.
Some links in this article are affiliate links. We may earn a commission if you sign up through them, at no extra cost to you. Every number below comes from our own YouTube Studio and ElevenLabs account. Nothing here is a projection presented as a result.
Key takeaways so far
- Starting point: in 28 days (Sep 6 to Oct 3, 2026), the English auto-dub brought 217 views and 2.2 watch hours, just 0.2% of total watch time.
- Auto-dubbing from Vietnamese only goes to English. Spanish, Portuguese or Indonesian need a dub you make yourself and upload as an extra audio track.
- 6 of our 22 most-viewed videos got no dub at all (“Processing failed”, no reason shown).
- YouTube’s machine-translated titles were the weakest part: several were cut off mid-sentence. We rewrote 18 titles by hand.
- First paid test: we dubbed one 9:22 video into English, Spanish, Portuguese and Indonesian with ElevenLabs. It cost about $0.50 per minute per language and took minutes, not days.
- Our 90-day goal: dubbed tracks bring 3 to 5% of total watch time by early January 2027.
Why one channel instead of an English channel
The classic advice is to open a separate channel per language. For a small team that means double the uploads, double the community management, and a new channel that starts from zero subscribers. YouTube now lets one video carry several audio tracks, and viewers hear the track that matches their language settings. So we chose one channel and one library, and let the audio do the translating.
The trade-off is honest: a dubbed video will never feel as native as a video recorded in that language. The question this case study answers is whether it is good enough to reach new viewers at almost no extra production time.
Our baseline: what the English auto-dub did in 28 days
YouTube turned on automatic dubbing for our channel by default. Before changing anything, we pulled the numbers from YouTube Studio’s advanced analytics, broken down by audio track (Sep 6 to Oct 3, 2026).
| Audio track | Views | Share of views | Watch hours | Share of watch time | Average view |
|---|---|---|---|---|---|
| Vietnamese (original) | 42,475 | 99.3% | 1,305.5 | 99.7% | 4:38 |
| English (auto-dub) | 217 | 0.5% | 2.2 | 0.2% | 1:12 |
| Total | 42,774 | 1,309.8 | 4:37 |
By country, Vietnam brought 95.6% of views. The United States came next with 297 views (0.7%), followed by Japan, Australia and Canada with a few dozen each.
Two things stood out:
- The English track exists, but almost nobody finds it. English titles were machine translated, often badly (see below), so the videos rarely surfaced for English searches.
- English viewers leave much earlier: 1:12 on average against 4:38 on the original. Our guesses, which we will test: Vietnamese text on screen, a synthetic voice that does not match the speaker, and examples that only make sense in Vietnam.
What YouTube auto-dubbing does for a non-English channel
According to YouTube’s help center, automatic dubbing works in two directions only: from English into about 20 languages, and from 29 languages, Vietnamese included, into English. There is no Vietnamese-to-Spanish or Vietnamese-to-Indonesian auto-dub. For a non-English creator, auto-dubbing is an English door, nothing more.
Useful details for your own channel:
- It is on by default for eligible channels. In YouTube Studio you can choose to review dubs before they go live, and you can unpublish or delete any dub later.
- Videos over 120 minutes, videos with little speech, unusual speech pacing, and videos with copyright claims are not dubbed.
- Viewers hear the track that matches their language preference, and can switch tracks from the settings (gear) menu.
Step-by-step instructions for all of this are in our YouTube auto-dubbing guide.
Audit: our 22 most-viewed videos
We checked the 22 most-viewed videos one by one (213,000+ views combined).
16 had an English dub, 6 had none
Six videos showed “Processing failed” for the English track, with no reason and no retry button. Two of them are Vietnamese adaptations of English-language talks, which fits YouTube’s rule about copyright claims. For the other four we have no explanation yet. They are ordinary screen-recorded tutorials, so speech pacing is our main suspect.
Machine-translated titles were the real bottleneck
Titles and descriptions are translated separately from the audio, and that is where we found the worst problems. A Vietnamese title with quotation marks came out in English as “Everyone uses NotebookLM, but 99% overlook this \”, cut off right where the quote began. Others were literal translations of Vietnamese clickbait (“Google Shocks the World!”) or ran past 90 characters.
We rewrote 18 English titles by hand in YouTube Studio (Languages → English → Title), following four rules:
- Write the way English speakers search (“What Is OpenAI Codex? A Beginner-Friendly Explainer”).
- Stay under about 70 characters so nothing is cut off.
- Keep the original promise, without exaggeration or income claims.
- Fix what the machine got wrong instead of polishing what it got right.
Four videos we chose not to localize
Four of the 22 were Vietnamese versions of English-language content. English speakers already have the original, so a dub back into English helps nobody and risks being treated as reused content. We left those four in Vietnamese only.
What we changed for new videos
- Hand-written English titles and descriptions on every upload, written at the same time as the Vietnamese ones.
- Less Vietnamese text on screen, or bilingual captions when text is needed.
- Language-neutral links: we say “funova dot ai slash v slash…” instead of a Vietnamese phrase. Our short links send viewers to the English page when their browser is not Vietnamese.
- Clear, unhurried speech, because a dub can only be as good as the transcript it is built from.
Test 1: one video, four languages, with ElevenLabs
Since YouTube will not auto-dub Vietnamese into other languages, we made the dubs ourselves and uploaded them through YouTube’s multi-language audio feature. We picked NotebookLM vs Gemini: Which One Should You Use, and When?: a 9:22 comparison that will not date quickly, and one of the six videos whose English auto-dub had failed. We dubbed it into English, Spanish, Portuguese and Indonesian with ElevenLabs, using its dubbing API on the Creator plan.
| What we measured | Result |
|---|---|
| Credits per language (9:22 video) | 25,761, the same for all four languages |
| Credits per minute of video | About 2,750 (the pricing page lists 3,000) |
| Cost on the Creator plan ($22 for 121,000 credits) | About $4.68 per language, about $0.50 per dubbed minute |
| All four languages | 103,044 credits, about $18.70 |
| Processing time | About 2 minutes for one language; about 10 minutes for three languages run in parallel |
| Output | MP3, mono, 128 kbps, exactly as long as the original (562.06 s), so no re-syncing |
Each language also got a hand-written title and description that says the audio is an AI dub. Viewers whose YouTube language is Spanish, Portuguese, Indonesian or English now see a title in their language, and YouTube plays the matching dubbed track based on their language preferences.
What tripped us up
- The API does not accept YouTube links. We downloaded the original from YouTube Studio, extracted the audio and uploaded that instead. Audio-only uploads are smaller and faster.
- A failed auto-dub blocks your own English track. We had to delete YouTube’s failed English auto-dub before the Add button appeared for our own file.
- “Add descriptive audio” is not the dub. In the language dialog, upload under the Audio row. The descriptive audio option is an accessibility track for blind and low-vision viewers, and our first English upload landed there by mistake.
- Uploading still happens by hand in YouTube Studio (Languages → language → Audio → Add), one file per language.
Next we will compare views, watch time and average view duration per audio track for this video after 30 days, then decide which videos to dub next. We will also test a local dubbing workflow later, to keep the process tool-agnostic.
Results log
We add a row at every checkpoint.
| Date | Change | Dubbed share of watch time | Notes |
|---|---|---|---|
| Oct 5, 2026 | Baseline, before any change | 0.2% (English) | 18 English titles rewritten the same day |
| Oct 6, 2026 | ElevenLabs dubs (EN, ES, PT, ID) live on one video | measuring | About $0.50 per dubbed minute per language |
| Mid-October 2026 | First check after the new titles and dubs | coming | |
| Early January 2027 | Day-90 review | coming | Goal: 3 to 5% |
FAQ
Can YouTube auto-dub a Vietnamese video into Spanish?
No. For Vietnamese, YouTube’s automatic dubbing only produces English. For other languages you create the dub with a tool or a voice actor and upload it as an additional audio track.
Does an auto-dub hurt the original video?
The dub is a separate track. It only plays for viewers who choose it, or whose language settings match it, so Vietnamese viewers keep hearing the original. In our baseline the original track still carried 99.7% of watch time. We will report any change at each checkpoint.
How do viewers find the dubbed version?
YouTube plays the track that matches the viewer’s language preference and shows the translated title and description. That is why titles matter so much: a broken English title means the dub is rarely discovered.
Should I open a separate channel for each language instead?
If you can record natively in each language and manage several communities, separate channels feel more native. If you are a small team, one channel with dubbed tracks costs a fraction of the effort. Our numbers in this case study will show how far that gets you.
