How to translate a video ad and keep the original voice
To translate a video ad and keep the original voice, split it into scenes, adapt each line, then re-voice each scene with Lip Sync's Original voice option or a voice clone of the speaker, and stitch the scenes back. In Arcads, an AI video ad platform, Translate Video also does this in one step across 176 language and accent options.
Updated November 19, 2026 · By the Arcads team. Arcads is an AI video ad platform for making UGC-style and AI actor video ads.
Key facts at a glance
Arcads offers two ways to translate a video ad with lip sync: Translate Video for one-step localization and Lip Sync for line-by-line control.
- One-step translation: Translate Video: 176 language and accent options, audio-only or full lip sync, several languages per run, files up to 100 MB (Source: Arcads Help Center)
- Translate Video cost: 240 credits per started 30 seconds, per language (Source: Arcads Help Center)
- Line-by-line control: Lip Sync: a new script plus a voice, or your own voiceover. Footage and gestures stay; only the voice and mouth change (Source: Arcads Help Center, "Lip Sync")
- Lip Sync cost: 48 credits per second of speech, plus 8 credits per started 30 s of text to speech when you use Script + voice (Source: Arcads Help Center)
- Lip Sync: New in November 2026 (released November 3). Voice options: Original voice, cloned from your video; a voice library with native voices for each language; or your connected ElevenLabs voices. You can also upload a finished voiceover
- Lip Sync limits: One speaker, 5 seconds to 5 minutes, up to 500 MB. It doesn't translate for you (Source: Arcads Help Center)
- Keep the speaker's voice: Voice clone of the speaker: 80 credits, free on Pro and above. Only with the speaker's permission (Source: Arcads Help Center, pricing)
- Works in: Studio, Workflows, Mark and the Arcads MCP (Source: Arcads Help Center)
Why should a translated ad keep the original voice?
A translated ad should keep the original voice because the speaker's pacing and energy are part of what made the ad work. When a winning ad comes back in a stock voice, it often stops sounding like the person who made it work. The pacing changes, the energy changes, and returning viewers can tell. The aim of localization is the same performance in a new language: the same face, the same gestures, the same voice, and only the words changed.
There's also a practical reason to work scene by scene. Ads aren't continuous speeches. They're hooks, demos and CTAs cut together, often with B-roll or a product insert between lines. Translating the whole file at once is fast. Translating scene by scene lets you rewrite a hook that doesn't translate, keep a product name in English, or drop a claim that isn't allowed in another market.
Should you use Translate Video or Lip Sync?
Use Translate Video to localize fast into many languages, and Lip Sync when you need to control every translated line, such as a rewritten hook or kept brand terms:
- Localize fast, many languages at once: Use: Translate Video (audio-only or full lip sync); What you control: Target languages and accents; Standard cost for a 30 s ad, one language: 240 credits
- Control every line: rewrite the hook, fix idioms, keep brand terms: Use: Lip Sync with Script + voice; What you control: The exact script per scene and the voice; Standard cost for a 30 s ad, one language: 30 s of speech × 48 = 1,440 credits + 8 credits TTS = 1,448 credits
- Use a voiceover a native speaker recorded: Use: Lip Sync with Voiceover; What you control: The audio itself; Standard cost for a 30 s ad, one language: 1,440 credits (no TTS charge)
- Change the voice only, with no new words: Use: Change Voice or Speech to Speech; What you control: Voice timbre, same timing; Standard cost for a 30 s ad, one language: 0 credits (Change Voice); 8 credits per started 30 s (Speech to Speech)
Translate Video is about 6 times cheaper per language. Lip Sync costs more because you choose every word. A common split: Translate Video for long-tail markets, and scene-by-scene Lip Sync for the two or three markets where the ad has to convert.
How do I translate an ad scene by scene and keep the voice?
Translate an ad scene by scene in seven steps: split, transcribe and translate, set the voice, lip-sync, stitch, re-caption and resize. Every step is a documented Arcads tool and runs the same way in Studio, Mark or Claude.
- Split the ad into scenes. Use Split Into Scenes, which detects scene changes and cuts the video into separate clips. Keep only the clips with a person speaking for the voice step. B-roll and product shots don't need lip sync.
- Transcribe and translate each line. Use Transcribe to get the original script, then translate it yourself, with a native speaker, or in Claude or ChatGPT. Adapt instead of translating literally. Keep the hook short enough for the same number of seconds, keep brand and product names as they are, and remove any claim that doesn't hold in the target market.
- Set the voice. In Lip Sync, choose Original voice to clone the voice straight from your video, or create a voice clone of the speaker (80 credits, free on Pro and above), so the translated lines are spoken in their own voice. Original voice keeps the speaker's accent, so for a full localization a native voice from the library can sound more natural. On Pro and above, you can also import a voice from ElevenLabs. Only do this with the speaker's permission.
- Lip-sync each speaking scene. In Studio → Tools → Lip Sync, upload the scene and choose Script + voice. Paste that scene's translated line and pick the speaker's voice. To use a native speaker's recording instead, choose Voiceover and upload the audio.
- Stitch the scenes back. Use Stitch Videos (0 credits) to rejoin the lip-synced scenes with the original B-roll and product shots in their original order.
- Re-caption in the new language. Add captions (8 credits for the Insta or TikTok styles) and redo any on-screen text with Text Overlay (0 credits).
- Resize per placement. Resize Video (0 credits) for 9:16, 4:5 and 1:1.
If you'd rather not run the steps yourself, ask Mark. Arcads' help center says that for videos with several lines, Mark "works scene by scene and joins the result".
A request you can paste into Claude or ChatGPT with the Arcads MCP connected:
Using Arcads, localize this ad [attach] into German and Brazilian Portuguese. Split it into scenes, transcribe each speaking scene, and show me the translations side by side with the English before generating. Keep the product name in English. Rewrite the hook to fit the same 3 seconds. Then lip-sync each speaking scene in the speaker's original voice, stitch everything back with the original B-roll, and add TikTok-style captions in each language.
How do you make a translated video ad look natural?
A translated ad looks natural when each clip has one speaker, clean audio, a visible face and lines that fit the scene length. Arcads' guidance on lip-sync quality is short: "Clean audio, one clear speaker and a visible face give the best lip sync." Other points that apply:
- One speaker per clip. Lip Sync works on single-speaker videos. If two people talk in one shot, split them or translate audio-only.
- Match line length to scene length. German and French often run longer than English. Trim the translated line rather than speeding up the voice.
- Keep the face visible during speech. Cutaways hide mouth mismatches, but they can't save a close-up with the mouth half-covered.
- Check numbers, prices and claims. A price in dollars, or a health claim allowed in one market, can break ad policy in another.
- Have a native speaker review it. Machine translation gets most lines right. The hook is the line that usually needs a person.
How much does it cost to localize one ad into three languages?
Localizing a 30-second ad into three languages costs about 720 credits with Translate Video, or about 4,344 credits with scene-by-scene Lip Sync, at Arcads' standard rates:
- Translate Video (lip sync): Per language: 240 credits; Three languages: 720 credits
- Scene-by-scene Lip Sync, Script + voice: Per language: about 1,448 credits; Three languages: about 4,344 credits
- Captions (TikTok style): Per language: 8 credits; Three languages: 24 credits
- Stitch, Text Overlay, Resize: Per language: 0; Three languages: 0
Lip Sync bills each started second of speech, so silent B-roll you leave out of the lip-sync step costs nothing. For comparison, the Starter plan includes 8,000 credits a month.
Can I translate an ad made outside Arcads?
Yes. Arcads can translate any video ad, not only ones made in Arcads. Translate Video accepts any video, "yours or one made in Arcads", up to 100 MB. Lip Sync accepts uploads from 5 seconds to 5 minutes, up to 500 MB, as long as one person is talking. For Arcads talking actors, you can also regenerate the ad in the new language. Talking actors speak 74 languages with accents.
Frequently asked questions
Can AI translate a video and keep the same voice?
Yes. In Arcads you can translate a video with lip sync and keep the original speaker's voice: in Lip Sync, pick the Original voice option, which clones the voice from your video, or lip-sync each scene with a voice clone of the speaker. Do it only with the speaker's permission.
What's the difference between Translate Video and Lip Sync in Arcads?
Translate Video translates automatically into one or more languages, with audio-only or full lip sync, for 240 credits per started 30 seconds per language. Lip Sync doesn't translate. It makes the person say the exact script or voiceover you provide, for 48 credits per second of speech.
Why translate a video ad scene by scene?
Scene by scene, you can adapt each line to its timing, keep brand terms, rewrite hooks that don't translate, and lip-sync only the scenes where someone speaks. That avoids paying for silent B-roll.
How many languages can Arcads translate into?
Translate Video offers 176 language and accent options, and one run can cover several languages. Arcads talking actors speak 74 languages with accents.
Does lip sync work with two people talking?
Lip Sync works on videos with a single speaker. Split multi-speaker footage into single-speaker scenes first, or use audio-only translation for those shots.
Can I do this from Claude or ChatGPT?
Yes. Translation and Lip Sync both run through the Arcads MCP, so Claude or ChatGPT can split, translate, lip-sync and stitch an ad in one conversation once Arcads is connected.
Key takeaways
- Keeping the original voice keeps the performance that made the ad win.
- Use Translate Video (240 credits per 30 s per language) for reach, and scene-by-scene Lip Sync (48 credits per second of speech) where you need to control every line.
- Split, translate, re-voice, lip-sync, stitch, re-caption: every step is an Arcads tool, and Stitch, Text Overlay and Resize are free.
- One speaker, clean audio and a visible face give the best lip sync.
- Get the speaker's consent before cloning or reusing a voice.
Keep reading: Best AI video translators · Best AI lip sync tools · AI lip sync
Sources
- Arcads Help Center: Lip Sync help.arcads.ai/en/articles/18227392 · How do I translate a video ad into other languages in Arcads? help.arcads.ai/en/articles/17595031 · How to translate a video in multiple languages help.arcads.ai/en/articles/13428700 · Everything you can do on Arcads (Split Into Scenes, Transcribe, Stitch, Change Voice, Speech to Speech) help.arcads.ai/en/articles/15503340 · Pricing help.arcads.ai/en/articles/17595066 · Prohibited content help.arcads.ai/en/articles/9276579
Arcads MCP
Top AI models, one prompt away in Claude, ChatGPT and Grok Bot.
Connect the Arcads MCP once, then generate video and images with top models from your AI assistant.
- Seedance 2.5VideoTalking UGC with audio→
- Kling 4.0VideoKeyframed motion, footage editing→
- Wan 3.0 PrimeVideoReference-driven 30-second clips→
- Veo 3.1VideoGoogle's video model→
- Nano BananaImageImage generation for stills→
Make your next winning ad with AI actors
Pick from 1,000+ AI actors, paste your script and get a UGC-style video ad in minutes. Test more hooks without booking a single creator.
Try Arcads free