Privacy.
Last updated 27 October 2026. This page describes opendub.app, the OpenDub browser extension for Chrome, Edge and Firefox, and the OpenDub app you run on your own computer.
OpenDub for Mac from the Mac App Store has no account and nothing for sale; its own, shorter policy is Privacy for the Mac App Store app.
What is never collected
Your videos. OpenDub dubs them on your own device — in this browser tab, in the browser extension, or in the app on your computer — and neither this website nor we have any way to receive one: there is no upload endpoint and no backend for dubbing. The website sets no cookie and loads no script from anyone else; it keeps anonymous counts of page views and of how far a task gets — never your files or their contents — described under This website. The voice you choose decides what else leaves your device: with the free voice, none of your video, audio or text; with a cloud voice and your own key, the speech and the text, as described under What leaves your device.
What is stored on your device
- This website remembers your language and voice choices in your browser's local storage. It never stores an API key.
- On this website, the video you choose to dub is also kept in this browser's own storage, so that it is still there if the page is reloaded (a phone can do that when it runs short of memory). It never leaves your device, and it is deleted when you remove or replace it, or after six hours.
- The browser extension stores nothing: it has no storage permission and keeps no settings, history or key. The browser caches the speech and translation models it downloads, as it would any download.
- A dub made in this tab exists only in the tab's memory until you download it; closing the tab discards it.
- One exception: if you press download and sign in first, the dub can be kept in this browser's storage meanwhile, so the download can continue when you are back. It stays on your device and is deleted once it has downloaded or you choose Not now.
- If you sign in, your session is kept in local storage too, so you stay signed in.
- The OpenDub app keeps the jobs you start in its own window — the video, the audio it separates, the transcript and the dub — in a
workfolder on your computer. Delete it whenever you like. Work it does for this website or the extension is not kept there; see The app on your computer. - If you save a Higgs Audio key in the OpenDub app, the app keeps it in a settings file on your computer that only your user account can read, and uses it for dubs you start in the app. Clearing the key in the app removes it.
What leaves your device when you dub
With the free voice (“On your computer”), your video and audio never leave your device: the audio goes from the page to the OpenDub app on the same computer, at 127.0.0.1, and never onto the internet. Only the anonymous counts described under This website are sent, and they carry none of your video, audio or text.
With a cloud voice, only to the provider you choose, with your own API key, straight from your browser or the app. The video file itself is never sent. On this website and in the extension, the key is first checked with a request that carries no audio, so a refused key sends nothing of yours.
- Higgs Audio (Boson AI,
api.boson.ai) — on this website and in the extension; the app, in its own window, sends the same kinds of data. Sent: the video's sound as 16 kHz mono audio (or only its voice, if you ask for the original voice to be removed), to be transcribed; the text of every line, to be translated, with each line's loudness and pace as numbers, to read its tone; and, with every line to be spoken, a voice sample of about 6 to 12 seconds cut from the video, which Higgs Audio imitates. No cloned voice is created or kept in your Boson AI account — the sample travels with each request — so there is nothing for OpenDub to delete afterwards. - ElevenLabs (
api.elevenlabs.io) — on this website only; the extension does not offer it. Sent: the video's sound, to be transcribed, and the text of every line, to be spoken. ElevenLabs needs a voice to speak with, so the page creates a temporary cloned voice in your own ElevenLabs account from a voice sample of about 6 to 12 seconds, named “OpenDub” with the date and time, and deletes it as soon as the dub ends, whether it finished or failed. If the tab is closed or the connection drops before then, that deletion cannot be sent, and the voice stays in your ElevenLabs account under Voices until you delete it there. ElevenLabs has no translation model, so without a Higgs Audio key the text is translated on your device.
On this website and in the extension, your key is held in the page's memory for that one dub, sent only to that provider in a request header, and never stored; closing the tab forgets it. The app keeps a key only if you save one in it, as described above.
Model downloads: the first time you dub, your browser (or the app) downloads the speech-timing model (Whisper), the translation models and, if you ask for the original voice to be removed, a voice separator, from Hugging Face. They are cached on your device; nothing of yours is sent to download them, though Hugging Face sees the request, like any website you visit. Chrome's own translator, when it is used, translates on your device.
The browser extension
- It reads a page only when you click it. When you open OpenDub from the toolbar, it looks once at the tab you opened it on, and at frames in that tab from the same site, for
<video>elements: each video's address, length and picture size, and the page's title. That is shown in the popup and goes nowhere else. Nothing runs on any page you have not clicked OpenDub on, and nothing stays behind on the page. A player embedded from another site is not looked into; the popup offers to open it in a tab of its own. - It asks for a video's site only when you dub a video from it. To download the video file into its own page, the extension needs access to the site the file comes from, and asks for that one site, on a button you press. Refuse, and nothing is read; you can choose the same video from your computer instead. You can take the access back at any time in your browser's extension settings.
- It downloads the video without your cookies, so the site sees an anonymous request, and keeps the file in the page's memory until you close the tab or download the dub.
- Its permissions:
activeTabandscripting, to look at the tab you clicked it on; access tohttp://127.0.0.1:8910, to reach the OpenDub app on your own computer; and, optionally, the sites you allow one at a time as above. It has no storage permission and no access to your browsing history, tabs or cookies. - It has no account and no advertising, and sends us nothing about what you dub.
- The use of information received from the extension adheres to the Chrome Web Store User Data Policy, including the Limited Use requirements. We do not sell it, do not use or transfer it for any purpose unrelated to dubbing the video you chose, and do not use it to determine creditworthiness or for lending.
The app on your computer
The OpenDub app answers only on your own computer, at 127.0.0.1:8910. When this website or the extension uses it, this is what each part receives and keeps:
- health receives nothing. It answers which voice engines are installed and ready, and which build of the app is running.
- separate receives the video's sound and returns the voice and the background as two tracks. It works in a temporary folder that is deleted as soon as it finishes; the result waits in memory until the page collects it; one that is never collected is dropped by the next separation after ten minutes, or when the app quits.
- transcribe and translate receive the speech as 16 kHz mono audio, or the lines as text, and return timed lines or translated lines. They work in memory and keep nothing.
- speak receives the voice sample and the text of a line, and returns the spoken line. The sample is written to a temporary folder for the voice model to read and deleted when the line is done.
- transcode receives a finished dub and returns a copy every video player opens, using a temporary folder that is deleted at once.
- models receives nothing. It hands the page the speech and voice-separator model files the installer brought, so they are not downloaded again.
- prepare receives nothing of yours. It starts the one-time download of the free voice's model from Hugging Face.
None of these is written to the app's work folder, and the app's log records which address was called, how it answered and, when something fails, the error — never the audio or the text.
Data used only if you sign in
Signing in is free. You can try OpenDub without an account; on this website, downloading a dubbed video or its audio takes a sign-in. Subtitle files, the browser extension and the app on your computer need none. When you sign in, our account server stores how you signed in (for example the email address Google shares, or a wallet address), your credit balance, your purchases, and what credits were spent on in our other apps.
Payments
Credits are bought through the payment provider you choose. We never see or store card numbers.
This website
This site counts page views with a self-hosted, privacy-preserving analytics service that runs on our own server and is served from this origin. It sets no cookie, records no personal data, and does not follow you to other sites. Beyond that, it keeps only anonymous counts of how far a task gets — for example whether a dub started, finished or failed, or whether a download went through — with general context such as the language pair, the type of device or the sign-in method, so we can see where dubbing gets stuck. Never your files or their contents: it never sees your video, your audio, your transcripts or your file names. The browser extension and the desktop app send us nothing about what you dub.
Messages you send us
If you use “Having trouble? Leave us a message” on this website, we receive what you write, your email address if you add one, and a few details that help us find the problem: which part of OpenDub you were in, the page address, your browser, the site's version and language, the type of error if there was one, and your account ID. You need to be signed in to send us a message. If you are not signed in, opening the box first asks our account server whether messages can be sent; that request carries nothing you typed. Your videos, their audio, transcripts and file names are never included. We use the message only to look into the problem and reply, and delete it after 180 days, or sooner if you ask at hello@opendub.app. Your message and the page it was sent from are also posted to our team's private Telegram group so we can reply quickly; your email address is never posted there.
Third parties
Google, a wallet or a Nostr signer, only if you choose them to sign in on this website. Boson AI or ElevenLabs, only if you dub with them using your own key. Hugging Face, for the one-time model downloads. Nobody else. OmniVoice's audio tokenizer, which the OpenDub app downloads from Hugging Face, is Higgs Materials under the Boson Higgs Audio 2 Community License.
What a cloud provider keeps is governed by its own policy and your account with it, not by us, and we cannot delete anything there for you. As of October 2026:
- Boson AI's privacy notice (updated 30 September 2026) says it does not use your content to train its AI models, and keeps the content it processes to provide its services until you ask it to delete it or delete your account; requests go to contact@boson.ai.
- ElevenLabs' privacy policy (updated 9 October 2026) says it keeps personal information for as long as it needs it to provide its services, for example while you have an account, and may use it to train its models unless you opt out under Data use in your ElevenLabs account.
Deleting your data
Your dubs are files on your computer; delete them there. To remove the extension's access to a site, or the extension itself, use your browser's extension settings. To delete your account and everything the account server holds, email us and we will remove it. For audio and text sent to Boson AI or ElevenLabs, ask that provider.
Children
OpenDub is not directed at children under 13, and we do not knowingly hold data about them.
Changes
If this changes, the date at the top changes with it.