Audio translator: translate a recording and voice it

MP3, WAV, M4A, OGG, FLAC, AAC and OPUS all go in. The more useful part of the answer is when not to use this one: if what you have is a video, the dubbing tool is the right door, because there the translated track goes back into the video file already lined up with the picture — here you would get an audio file and then have to sync it yourself. Likewise, if you only want the words rather than a new voice, transcription and translation of the text is a shorter route. This tool is for the case where you need the recording spoken again in another language.

Far enough to understand a recording, not far enough to quote from it. Two steps can each go wrong: the speech has to be recognized, and the recognized text has to be translated. Recognition is the weaker link — noise, a strong accent, and two people talking at once all damage it, and a mis-heard word becomes a confidently translated wrong word further down the chain. Everything you would act on is worth checking against the transcript, which is shown next to the translation for exactly that reason. Names, numbers and negations deserve a second look in particular: they change the meaning most and survive recognition worst.

No — the translation is voiced with ready-made voices, not with a copy of the speaker. When several people talk, each gets a different voice, so an interview does not turn into one narrator.

With “Keep music and background” on, a separate model splits speech from the backing track and the translated voice is laid over the original music.

Up to an hour per run, with a five-minute trial before you commit and the full length available with Premium. The reason to think about length at all is that a long recording is a long wait during which the tab has to stay open, so an hour-long interview is better split at a natural break than fed in whole — you also get to check the first part before spending the rest of the time. If your recording is longer than an hour, cut it at a topic change rather than at the exact hour mark: a sentence split across two runs comes back as two half-translated sentences, and no later editing recovers what was said across the seam.

Yes. The timestamped transcript is shown on the page, and you can download it as TXT with the translation, SRT in the target language and SRT in the original.

Yes. The transcript appears on the page: copy it with one button or take it as TXT and SRT, and never download the audio. If you need no translation at all and just want the recording written out in its own language, that is Audio to text.

No — here “translate” means the language, not the file format. Changing the format is File converter, and pulling the soundtrack out of a video is Extract audio from video.