Meta AI Voiceover and Subtitle Translation: Video Ad Localization for Cross-Border Sellers
What problem these features solve
More than 85% of Facebook and Instagram video ads are watched with the sound off. Meta’s own data has said this for years, and it hasn’t changed. Most sellers know this, which is why they lean hard on subtitles rather than voiceover.
But sound-on moments do happen. When a user taps to unmute, if there’s nothing to hear beyond background music, they leave. AI Voiceover is aimed at that gap: it adds a narration track that delivers product information when users choose to listen. Silent auto-play, voiced on demand.
The bigger practical problem AI Subtitle Translation solves is cost and turnaround time. If you’re running ads in five markets, the traditional workflow is: produce one video, hire a translator for each language, hire a voice actor for dubbed versions, re-edit and re-export, upload separately. Five languages done properly runs $1,200-2,000 in outside costs and two to three weeks of calendar time. Meta’s tool does this inside Ads Manager for free, in about 15 minutes.
That time and cost difference matters most when you’re running a time-sensitive promotion or testing whether a new market is worth pursuing at all. You can spin up five localized versions of a test ad for near zero marginal cost, see what the data says, and then decide whether to invest in a polished localized version.
How to enable AI Voiceover and Subtitle Translation in Ads Manager
Both features live under the “Creative Tools” section at the Ad level in Meta Ads Manager. You’ll find it below the asset upload area after you attach a video to your ad.
Enabling AI Voiceover
After uploading your video, click into “AI Voiceover” within Creative Tools. The setup prompts you through three choices:
- Target language for the voiceover
- Voice preset (more on these below)
- Script source — you can let Meta’s AI extract text from the video automatically, or type a script manually
Write the script manually. Auto-extraction from video is unreliable for product names, specifications, and pricing — the exact content where accuracy matters most. Keep it short: two to three sentences covering the core value proposition, under 100 words. That’s enough for a 15-30 second product video.
Enabling AI Subtitle Translation
In the same Creative Tools section, find “Subtitle Translation.” If your video already has an SRT file attached, Meta will translate from that. If it doesn’t, the system runs speech-to-text first, then translates.
Select the languages you want. The system generates all translations in parallel, so selecting five languages doesn’t take five times as long. After generation, a preview pane shows each subtitle line by language. You can edit individual lines directly — worth doing for product names, numerical specs, and any legally sensitive copy before publishing.
Language coverage and voice presets
Language support as of May 2026:
| Language | Subtitle translation | AI Voiceover | Quality notes |
|---|---|---|---|
| English | Yes | Yes | Best quality |
| Spanish | Yes | Yes | Best quality |
| French | Yes | Yes | Best quality |
| Portuguese | Yes | Yes | Specify pt-BR vs pt-PT |
| German | Yes | Yes | Good |
| Italian | Yes | Yes | Good |
| Japanese | Yes | Limited | Honorifics need review |
| Korean | Yes | Limited | Honorifics need review |
| Arabic | Yes | Beta | Manual review recommended |
English, Spanish, and French have the best output quality by a visible margin. Portuguese works well but you need to specify Brazilian Portuguese (pt-BR) versus European Portuguese (pt-PT) — they differ enough that using the wrong variant sounds wrong to native speakers.
Japanese and Korean translations come out grammatically accurate but sometimes miss register. Japanese in particular has formal and informal speech patterns that matter a lot in advertising context. Using the wrong level of formality can make a product ad sound either too stiff or too casual for the intended audience. If Japan or Korea is a serious market for you, treat the AI output as a first draft.
AI Voiceover currently offers four voice presets: a measured explanatory tone, a conversational tone, a professional broadcast-style delivery, and an upbeat promotional tone. There’s no custom voice option and no way to upload a voice sample to clone a specific speaker. For brands that have established a distinctive audio identity, this is a real limitation.
Cross-border workflow: one video, multiple markets
Here’s how the workflow looks in practice for a seller manufacturing in China and running ads in the US, UK, Mexico, Brazil, and France.
You have one source video in Mandarin, shot for domestic platforms. The content is solid: good lighting, clear product shots, a presenter walking through key features. You need five localized versions.
In Ads Manager, upload the source video and open Creative Tools. Enable Subtitle Translation and select all five target languages simultaneously. While that runs, write a short English voiceover script covering the three main selling points — this takes about five minutes. Enable AI Voiceover, select the English version, choose a voice preset that matches your video’s pacing.
When the translations come back, work through each language’s subtitle preview. Pay attention to: product name spelling and pronunciation, any numeric specs (dimensions, weight, battery life), and pricing if mentioned in the video. These are where errors cluster. For the French and Spanish versions, also read through any superlative claims (“best,” “most powerful”) — translation sometimes drifts these into territory that sounds overclaimed.
Total time from upload to five localized versions ready for review: around 30-45 minutes, including your own review pass. Compare that to the three-week external localization timeline.
Quality limitations and risk checklist
Voice preset mismatch
The four presets cover common cases, but they’re generic. A soft-spoken wellness brand and a loud discount retailer both have to pick from the same four options. If the preset clashes with the visual tone of your video, the ad feels disjointed. Preview carefully before running.
Product name and model number pronunciation
AI voiceover stumbles on branded terms and model numbers. “ProMax X9 Ultra” might come out as a strange syllable string. Fix this by writing the script around the problem — describe the product rather than naming it, or spell out how the name should be pronounced in the script text.
Health claims and legal copy
If your product category involves health, efficacy, or safety claims (supplements, skincare, electronics with safety certifications), review translated text line by line against the source. AI translation can drop limiting qualifiers or shift “may help” into “treats.” Those changes matter legally. Don’t rely on spot-checking.
Japanese and Korean register
Covered above, but worth repeating as a checklist item: if these are target markets, budget for a native speaker review of the AI-generated translations before publishing.
Pre-launch checklist
- Product names and model numbers pronounced correctly in voiceover
- Numerical specs (dimensions, weight, price) translated accurately in all languages
- Health or efficacy claims retain their original limiting language
- Subtitle timing aligns with on-screen visuals
- Correct Portuguese variant selected (pt-BR for Brazil, pt-PT for Portugal/Europe)
- Voice preset matches video pacing and brand tone
- Japanese/Korean translations reviewed for appropriate register if applicable
阅读本文中文版: Meta AI配音+字幕翻译：跨境卖家视频广告本地化实战
Related Articles
Google AI Brief in Practice: Three Layers of Natural Language Control for AI Max
AI Brief lets advertisers steer Gemini inside AI Max using plain English across three layers: Messaging Guidelines, Matching Guidelines, and Audience Guidelines. Here is how to use each layer and what cross-border sellers should write in each field.
Google Veo 3 in Google Ads: Generate Video Ads From Product Images at No Cost
Since May 6, 2026, Veo 3 image-to-video is built into every active Google Ads account for free inside Asset Studio. Upload product photos, get a 10-second HD video with ambient audio. Here is the full workflow and what it actually produces.