More from Pascual Labs
Three ways to understand a video in a language you don't speak — and an honest account of where each one stops working.
Yandex renders a natural-sounding dub of the whole video and AnyDub plays it over the original, ducking the source audio so both stay audible.
Best for: videos you want to actually enjoy rather than merely follow. The voice sounds human, not synthetic.
Trade-off: the first request for a given video has to be rendered from scratch, which can take up to a couple of minutes. Source-language support is limited (see below).
Takes the subtitles the video already has, translates them, and speaks them with a voice installed in your system.
Best for: starting immediately. There is nothing to render, so playback begins within a second or two, and it works on far more sites and languages than the AI voice-over.
Trade-off: the voice is synthetic, and the video must have subtitles in the first place.
Captures the tab's audio and transcribes it locally with a speech-recognition model, then translates and speaks the result. Nothing about the audio leaves your computer for recognition.
Best for: live streams, and any video with no usable subtitles at all. It works with any spoken language.
Trade-off: it runs on your hardware, so it needs a moment to load the model on first use and follows speech with a short delay.
By default the on-screen subtitles mirror whatever the active engine produced. Turn on Yandex subtitles in the CC tab and AnyDub instead requests Yandex's own subtitle track, which carries its own timings.
You can tell them apart at a glance: language badges turn red when the lines come from Yandex, and stay violet otherwise.
With the checkbox on, CC speaks the Yandex lines rather than its own. That matters because the two tracks are split differently — using Yandex timings keeps the speech aligned with what is actually being said.
The subtitle overlay can show the translation, the original, both stacked, or a teleprompter view. Drag it anywhere over the video; the position is remembered separately for fullscreen and windowed playback.
The subtitles tab has a Download subtitles section at the bottom. SRT keeps the timings and loads into any player or editor; TXT is plain text for reading or pasting into a document.
You can save the translation on its own, the original on its own, or both stacked in one file. Start the translation first — the file contains exactly the lines that are on screen, whether they came from Yandex, the CC engine or live recognition.
Yandex officially supports eight source languages: English, German, French, Spanish, Italian, Chinese, Japanese and Korean. On these, both the AI voice-over and the subtitles are dependable.
A few more — Arabic, Lithuanian, Latvian — are accepted but undocumented. They often work, yet subtitle coverage can be thin: on some Arabic videos only about a third of the runtime comes back with lines, leaving long silent gaps. That is a limit of the source, not a fault in the extension. Switch to LIVE for those languages — local recognition does not care which language is being spoken.
Not every video has subtitles, and no amount of retrying will conjure them. When AnyDub reports that none are available, LIVE is the mode that still works — it listens to the audio directly.
The first time anyone requests a dub for a particular video, it is rendered from scratch. Afterwards the same video starts instantly for everyone, which is why a popular video feels immediate and an obscure one does not.
The overlay shows what is happening rather than a frozen line: a pulsing indicator with an attempt counter while the request is in flight, and a progress counter while subtitles are being translated.
A stream has no finished audio track to render, so the AI voice-over cannot apply. LIVE is built for exactly this case.
AnyDub re-syncs automatically after seeking, pausing or changing playback speed. If it still drifts, stop and start the dub again — that reloads the track against the current position.
The panel has separate volume controls for the dub and for the original audio. Lower the dub or raise the original until the balance suits you; the setting is remembered.
That is the CC mode speaking through a system voice. VIDEO mode uses a rendered AI voice and sounds markedly more natural — if it is available for your video, prefer it.
AnyDub attaches to standard video players, so most sites work. On unusual players the control may sit as a floating pill rather than inside the player bar. If a site still shows nothing, tell us and we will look at it.
For the AI voice-over and for Yandex subtitles, the video's address is sent so the track can be produced. For CC, subtitle text is sent for translation. In LIVE mode speech recognition runs entirely on your computer.
AnyDub does not upload your video, your microphone or your browsing history. The full statement is on the privacy page.
