Step by step instructions to turn off AI dubbing on YouTube as a viewer or a creator, on desktop and mobile, plus the settings that stop dubbed audio from returning. See when a dub is worth keeping and how to build better ones with dubbing and lip sync models.
You click a video from a creator you like, and the voice coming out of your speakers is wrong. The rhythm is a little too even, the jokes land flat, and the mouth on screen is clearly shaping words in another language. YouTube has swapped the original audio for an AI dubbed track, and now you are hunting through menus to get the real voice back.
The fix takes about ten seconds on any device, and creators can stop dubbing on their own uploads altogether. Below you will find the exact steps for viewers and for channel owners, the reasons the dub keeps returning on new videos, and an honest look at the moments when a dubbed track is the better choice. If you make videos yourself, there is also a section on building dubs you control with models on PicassoIA.
YouTube's automatic dubbing takes the speech in a creator's video, translates it, and generates a synthetic voice track in another language. That track is attached to the same video, so one URL can carry the original audio plus several dubbed versions. Nothing is re-uploaded and the picture never changes. Only the sound does.
The feature supports dozens of languages, and some of them use an expressive speech mode that tries to copy the pitch and intonation of the original speaker so the result sounds less robotic. Videos longer than 120 minutes are not eligible, and creators decide whether their channel takes part at all.
YouTube marks videos that carry a synthetic track as auto-dubbed in the description. If you open a video and the narration feels oddly smooth, expand the description first. That label tells you the audio you are hearing may not be the creator's own voice, and it is a faster check than guessing from the sound alone.
Two things have to line up. The creator's channel has to be eligible and have dubbing switched on, and your own language settings have to point somewhere other than the video's original language. When both are true, YouTube may start the video on the dubbed track instead of the original. Availability and supported languages change over time, so two people watching the same video can hear different audio.
There are two layers here. The first is a per video switch that works instantly. The second is a language preference that changes what plays by default. Use the switch to fix the video in front of you, and the preference to stop the problem from repeating.
The sound changes without restarting the video, so you can do it in the middle of a sentence and keep watching.
💡 No "Audio track" row in the menu? Then that video only has one track, and there is nothing to switch.
On a phone, the menu hides behind the player controls.
Menu names shift a little between app versions, but the route is always the same: player settings, then audio track, then the original option. Tablets and smart TV apps follow the same path through the player's settings menu, just with a remote or a touch screen instead of a mouse.
This is the fix that sticks. On the web, click your profile picture, open Settings, go to Playback and performance, and find Preferred languages. YouTube describes it as the place to choose the languages you prefer for audio, titles and descriptions.
Add every language you genuinely speak. If the original language of a video is on that list, YouTube is far more likely to start on the original track instead of a dub. Bilingual viewers should list both languages, otherwise half of what they watch will arrive dubbed into the other one.
Here is how the three viewer options compare:
| Method | Scope | Effort | Best for |
|---|---|---|---|
| Audio track menu | One video | A few seconds | Fixing a dub right now |
| Preferred languages | Your account | One time | Stopping repeats |
| Browser extension | Desktop browser | Install once | Forcing original audio everywhere |
Many viewers fix one video, open the next, and hear the dub again. Three reasons explain almost every case:
Some viewers install browser extensions that force the original track on every video. They work for many people, but they also request permissions inside your browser, so read the reviews and the permission list before adding one.
If the videos are yours, you hold the controls, and you have three levels to choose from: switch dubbing off for the whole channel, review each dub before it goes public, or remove a single language from a single video.
In the YouTube Studio app, tap your profile picture, open Settings, and under Content tap Automatic dubbing. Switch it off.
On desktop, the same control sits under Settings, then Upload defaults, on the second tab of that page. There you untick Allow automatic dubbing and save. New uploads stop receiving synthetic tracks from that point on.
💡 Old videos need a separate check. The channel setting applies to uploads you publish after you change it. Open the Languages tab of older videos to see whether dubs already exist.
You do not have to choose between everything and nothing. YouTube lets creators pick Publish manually, so each dub waits for your approval instead of going public on its own.
On desktop, open the video's Languages tab, play the dub, and publish it only if the voice matches your tone and the translation is accurate. This is the safest setting for tutorials with technical terms, brand names, or jokes that do not survive translation.
Sometimes only one dub is the problem, for example a language where the translation mangles your product name. In YouTube Studio on desktop:
💡 Deleting is permanent. YouTube states that a deleted dub is gone and cannot be republished. Pick Unpublish unless you are certain.
Switching the dub off is the right call for personality driven videos. It is not always the right call, and treating every dub as a defect means giving up something useful.
A repair walkthrough, a recipe, or a travel documentary in a language you do not know is a good example. Subtitles pin your eyes to the bottom of the screen while your hands and attention are busy elsewhere. A dub lets you keep watching while you cook or fold laundry. Viewers with low vision or reading difficulties also benefit from spoken audio in their own language.
For creators, one upload can reach viewers in other languages without hiring voice actors or re-recording anything. Evergreen how-to content benefits most, because the viewer wants the steps and not the personality. A sensible approach is to review a few dubs in a language you can check, keep the ones that sound right, and unpublish the rest.
Here is a quick way to decide which audio to use:
| Situation | Better choice | Why |
|---|---|---|
| Comedy, vlogs, storytelling | Original audio | Timing and tone carry the content |
| Step by step tutorials | AI dub | Information matters more than delivery |
| Interviews and podcasts | Original audio | Voice identity and emotion matter |
| Music and live performance | Original audio | A dub cannot reproduce it |
| Lectures in an unfamiliar language | AI dub or subtitles | Depends on whether you listen or read |
| Background listening while busy | AI dub | Your eyes stay free |
YouTube's automatic dub is a one size fits all option. When you want to control the voice, the language variant and even the lip movement, a dedicated dubbing model gives you more say. PicassoIA hosts several of them, plus the voice and image models you need around a dub. You can publish the result as a separate localized video, or upload it as an extra audio track if your channel has multi-language audio available.
ElevenLabs Dubbing re-records a video or audio file in a new language while keeping the original speaker's voice, emotion and timing. It supports more than 90 languages and regional dialects, including Mexican Spanish and Brazilian Portuguese.
es-MX, pt-BR or en-GB.auto. The model detects the original language by itself.💡 Test on a short clip first. Cut thirty seconds that include a proper noun, a number and a joke. If the dub handles all three, the rest of the video is usually safe.
An audio only dub leaves the speaker's mouth saying something else. HeyGen Video Translate takes an .mp4 and returns a dubbed video with the lip movements aligned to the new audio, in more than 150 output languages.
The settings are simple. Upload the video, choose the output language, and pick a mode. Speed is built for quick turnarounds and precision, the default, is built for higher quality output. Use speed for drafts and precision for the version you publish.
HeyGen also lists Lipsync Precision and Lipsync Speed as separate dubbing models. If you already have the translated voice and only need the mouth to match, Lipsync 2 Pro is built for that job.
| Model | Output | Best for |
|---|---|---|
| ElevenLabs Dubbing | Dubbed audio | Keeping the original voice in 90+ languages |
| HeyGen Video Translate | Dubbed video with lip sync | Speakers who are visible on camera |
| Lipsync 2 Pro | Video matched to your audio | Your own recorded translation |
| Speech 2.8 HD | Narration audio | A voiceover written from a script |
If you only need a narration track in another language, a text to speech model is faster than a full dub. Speech 2.8 HD, ElevenLabs v3 and Gemini 3.1 Flash TTS all turn a script into spoken audio, and Gemini 3.1 Flash TTS offers 30 voices across more than 70 languages. To keep one recognizable voice across your videos, Voice Cloning builds a custom voice from a sample.
Translations are only as good as the script they start from. Run the original through GPT 4o Transcribe to get an accurate transcript, fix the names and numbers, and then translate.
Localized videos also deserve localized thumbnails. GPT Image 2, Seedream 4.5 and P Image can produce a regional version of your thumbnail in minutes. Keep any on image text to one or two short words so it stays readable on a phone.
Most problems come from the same handful of causes. Check these before you assume something is broken.
Two more mistakes show up on the creator side. Deleting a dub when unpublishing would do the job is the expensive one, because deletion cannot be undone. Judging every dub from one test video is the other. Check at least two or three before you decide how a language performs.
Whether you want the original voice back or a better dub than the automatic one, the quickest way to settle the question is to hear both. Take a thirty second clip from one of your own videos, run it through ElevenLabs Dubbing with Spanish as the target, then send the same clip through HeyGen Video Translate and compare the two side by side.
From there, generate a matching thumbnail, record a narration with a text to speech model, or try a different language entirely. Every model mentioned above lives on Picasso IA, so you can experiment with voices, lip sync and images in one place and keep whichever result sounds like you.
