Auto dubbing is the machine production of a translated voice track for a video, made without a human translator or voice actor. Software transcribes what is said, translates the transcript, synthesises speech in the target language and attaches the result to the original upload as an alternative soundtrack. YouTube uses the term for a feature it switches on by default for eligible creators, labelling the output "auto-dubbed". The practice exists because subtitles demand reading, studio dubbing is priced for film and television rather than a weekly upload, and much of any video's potential audience does not speak the language it was recorded in.

How the pipeline works

Researchers at Amazon set out the architecture in a January 2020 paper, From Speech-to-Speech Translation to Automatic Dubbing. Their system chained four components, and commercial products still follow the same sequence.

Recognition comes first. The system identifies the source language and produces a transcript. On YouTube, the declared video language steers this step; the Help Centre warns that a wrong setting produces mistranslated dubs, and correcting it triggers regeneration.

Translation is constrained by time. A dubbed sentence has to fit roughly the duration of the original utterance, so the Amazon team built "neural machine translation generating output of preferred length" and aligned the translated text with the rhythm of the source speech.

Synthesis turns the translated text into audio, with each utterance stretched or compressed to fit its slot. Early YouTube dubs used stock synthetic voices, and the platform's own guidance in April 2025 stated that the tone and emotion of the original audio were not transferred. Expressive Speech, developed with Google DeepMind, now carries pitch, intonation and energy across from the source.

Rendering mixes the new voice with background sound separated from the original track. An optional fifth stage alters the picture. Buddhika Kottahachchi, YouTube's product lead for auto dubbing, told Digital Trends that its lip sync system makes "intricate pixel-level changes" to the speaker's mouth; as of October 2025 it was limited to 1080p video, according to Android Authority.

YouTube's eligibility rules, as of September 2026, exclude videos longer than 120 minutes, videos with no speech or only music, content whose language cannot be detected, material carrying copyright claims, and speech so fast that the dub would be, in the Help Centre's words, "unlistenable, sped-up". Creators can switch to manual publishing, review each track in YouTube Studio and publish, unpublish or delete individual languages, or disable the feature under advanced settings. Coverage is asymmetric: the Help Centre lists English dubbing into 20 languages, while most other supported languages dub only into English, with exceptions such as Portuguese to Spanish and Korean to Indonesian. Viewers change tracks from the player's settings menu.

Meta's equivalent, branded Meta AI translations, is opt-in. The dub uses a clone of the creator's own voice, lip sync is optional, and the system handles up to two speakers who do not talk over one another, according to TechCrunch. Facebook creators with at least 1,000 followers and public Instagram accounts in markets where Meta AI operates are eligible. Translated Reels carry a "Translated with Meta AI" label, and viewers can select "Don't translate".

Origin and evolution

Google's Area 120 incubator released Aloud on March 9, 2022. Its co-founders, Buddhika Kottahachchi and Sasakthi Abeysinghe, wrote at launch that "dubbing used to take weeks worth of effort and a large budget", a job their tool cut to minutes. Output was limited to Spanish and Portuguese, and creators had to state that the dubs were synthetic. YouTube began testing the tool with hundreds of creators at VidCon in June 2023, according to Engadget, promising voices closer to the creator's own and, eventually, lip sync.

Competitors moved in the same window: Spotify piloted voice translation for podcasts using OpenAI technology on September 25, 2023, and ElevenLabs released AI Dubbing, covering more than 20 languages, on October 10, 2023.

YouTube widened access to hundreds of thousands of creators at Made on YouTube on September 18, 2024, with French and Italian joining Spanish and Portuguese. The formal launch followed on December 10, 2024 for Partner Program channels focused on knowledge and information, dubbing English into eight languages including German, Hindi, Indonesian and Japanese and those languages back into English. All Partner Program creators could self-enrol from April 2025, according to Social Media Today. In August 2025, YouTube said creators would be able to edit auto-dubbed videos in Studio Editor, with tracks regenerated to match. On February 4, 2026, in a post by product manager Chandralekha Motati, YouTube declared the feature "available to everyone", expanded coverage to 27 languages and switched on Expressive Speech in eight.

Meta followed a parallel track. Mark Zuckerberg previewed lip-synced translation of Reels at Meta Connect on September 25, 2024. The product launched globally on August 19, 2025 in English and Spanish and added Hindi and Portuguese that October, according to TechCrunch.

Why marketers are paying attention

The adoption figures all come from YouTube and have not been independently audited. In a February 2025 letter, chief executive Neal Mohan said that more than 40% of watch time on videos with dubbed audio came from viewers choosing a dubbed language. A two-year pilot of human-recorded tracks produced more than 25% of watch time from non-primary languages and tripled views for Jamie Oliver's channel. In December 2025, more than 6 million viewers a day watched at least 10 minutes of auto-dubbed content, with dubbed views averaging 75% of the original language's average view duration.

For buyers, the consequence is that a video's language is no longer a fixed property. PPC Land's explainer on language targeting notes that a single upload may now carry audio in a language its publisher never produced. A placement list assembled by channel language can reach a Hindi-speaking audience hearing a synthetic voice. YouTube's automatic dubbing documentation does not address how advertising or revenue works on dubbed tracks.

The same machinery is moving into ad creative and premium streaming. Google began adding AI voice-overs to silent Performance Max videos in March 2026, with an opt-out deadline of March 20. Prime Video opened an AI-aided dubbing pilot on 12 licensed titles in March 2025, pairing machine output in English and Latin American Spanish with human localisation staff.

Limitations and disputes

Quality is the first problem. YouTube's Help Centre says the system struggles with "proper nouns, idioms, and jargon" and that quality "may not be the same across all languages". On a Creator Insider episode published on May 15, 2026, Kottahachchi was blunt: "YouTube is in the bleeding edge of culture, our models are usually playing catch-up."

Consent is the second. The default-on design provoked sustained complaints. Viewers objected in April 2025 that no global switch existed to turn dubs off, with SEO consultant Gagan Ghotra reporting that his feed showed European videos "dubbed as English". Machine-translated titles and descriptions drew similar anger in July 2025. The Preferred Language setting, introduced in February 2026, was YouTube's response.

Rights and labour form the third. Amazon withdrew English "AI beta" dubs of Banana Fish and other anime in early December 2025 after the National Association of Voice Actors called them "AI slop", according to Engadget. Kadokawa said it had not approved the dub of No Game, No Life Zero "in any form", according to Popverse. In Germany, the Berlin Regional Court ruled on August 20, 2025 that an AI clone of a dubbing artist's voice violated personality rights, awarding 4,000 euros. Detection lags behind: YouTube's likeness detection matches faces, with audio promised during 2026.

Regulation adds pressure. Article 50 of the EU AI Act has applied since August 2, 2026, and its definition of a deep fake covers AI-generated audio or video that "resembles existing persons" and would falsely appear authentic, according to the European Commission. Machine-readable marking for systems already on the market must be in place by December 2, 2026. YouTube moved its generative AI labels to more visible positions in May 2026, yet the Help Centre still places the auto-dubbed marker in the video description.

Not the same as

Multi-language audio is the delivery container rather than the generator. Opened to millions of creators in September 2025, it accepts tracks recorded by people or vendors and, in YouTube's words, "does not automatically create these tracks".

AI voice-over synthesises new narration from written ad copy for videos with no speech. Nothing is translated.

Live speech translation, such as Gemini 3.5 Live Translate, streams speech-to-speech output across more than 70 languages for conversations rather than published media.

Deepfakes present a real person saying or doing something they did not. Voice cloning and facial manipulation are shared techniques; lip-synced dubs apply them to a speaker's own footage, with a platform label attached.

Recent developments

Development has moved to faces and live video. On September 11, 2026, Prime Video applied AI lip sync to the English dubs of Maxton Hall, its most-watched international Original, inverting YouTube's approach: human actors record the voices, and software reshapes the mouths. Seasons one and two carry the feature, with season three due on December 9, 2026. At Made on YouTube on September 23, 2026, Barbara Macdonald, a product manager for YouTube Live, announced live auto dubbing from early 2027, noting that more than 40% of live watch time comes from viewers outside a creator's home country. The pilot will run in English and Spanish, according to Social Media Today. Meta, meanwhile, keeps adding languages; French, German, Italian, Japanese and Korean reached Instagram in July 2026, according to Meta.

Timeline

  • January 2020: Amazon researchers publish From Speech-to-Speech Translation to Automatic Dubbing
  • March 9, 2022: Google's Area 120 launches Aloud with Spanish and Portuguese dubbing
  • June 2023: YouTube tests Aloud with hundreds of creators at VidCon
  • September 25, 2023: Spotify pilots AI voice translation for podcasts
  • October 10, 2023: ElevenLabs releases AI Dubbing
  • September 18, 2024: YouTube expands auto dubbing to hundreds of thousands of creators
  • September 25, 2024: Meta previews lip-synced Reels translation at Connect
  • December 10, 2024: YouTube launches auto dubbing for knowledge-focused Partner Program channels in eight languages
  • February 11, 2025: Neal Mohan reports that more than 40% of watch time on dubbed videos comes from dubbed languages
  • March 5, 2025: Prime Video starts its AI-aided dubbing pilot on 12 titles
  • April 2025: All Partner Program creators gain early access to auto dubbing
  • August 19, 2025: Meta AI translations launch globally in English and Spanish
  • August 20, 2025: Berlin Regional Court rules against an unauthorised AI voice clone of a dubbing artist
  • August 2025: YouTube announces Studio Editor support for auto-dubbed videos
  • September 10, 2025: YouTube opens multi-language audio to millions of creators
  • September 16, 2025: YouTube announces lip sync testing across 20 languages
  • October 9, 2025: Meta adds Hindi and Portuguese
  • December 2025: Amazon withdraws AI beta anime dubs; YouTube records 6 million daily viewers of auto-dubbed content
  • February 4, 2026: YouTube opens auto dubbing to everyone in 27 languages with Expressive Speech
  • March 2026: Google tells advertisers AI voice-overs will be added to Performance Max videos unless they opt out by March 20
  • May 15, 2026: YouTube details auto dubbing performance on Creator Insider
  • July 2026: Meta AI translations reach French, German, Italian, Japanese and Korean on Instagram
  • August 2, 2026: Article 50 of the EU AI Act applies
  • September 11, 2026: Prime Video launches AI lip sync on Maxton Hall
  • September 23, 2026: YouTube announces live auto dubbing
  • December 2, 2026: Marking deadline for AI systems already on the EU market
  • Early 2027: Live auto dubbing pilot scheduled to begin

Summary

Who. YouTube operates the most widely used auto dubbing system, which grew out of Aloud and was developed with Google DeepMind and Google Translate. Meta runs Meta AI translations for Facebook and Instagram, Prime Video uses hybrid and lip sync approaches, and vendors such as ElevenLabs sell dubbing as a service. Voice actors, rights holders and regulators are the principal counterweights.

What. Automated transcription, translation and speech synthesis that produce an alternative soundtrack in another language, sometimes with the speaker's cloned voice and with lip movements altered to match.

When. Described in research in January 2020, launched as Aloud on March 9, 2022, rolled out by YouTube from December 10, 2024 and opened to all YouTube creators on February 4, 2026, with live dubbing due in early 2027.

Where. YouTube videos in 27 languages, Facebook and Instagram Reels, Prime Video titles, and third-party tools used across platforms.

Why. It removes the cost and time that kept most online video in a single language, reshapes who watches a given upload, and makes content language an unstable signal for ad targeting, while raising unresolved questions about consent, quality and disclosure.