Upload a recording or video in Spanish, French, Japanese, Hindi or any of 99 languages and get English text with timestamps. The Whisper speech model runs on your own device, so the audio is never uploaded. Copy the translation or save it as TXT, SRT or VTT subtitles.
Last updated:
Preparing…
English translation
TimeEnglishOriginal
What you get (example)
0:00Good morning everyone, and thank you for coming.
0:04Today we are going to talk about the new train schedule.
Add a recording in any language above and press Translate to English. The first run downloads the Whisper model once; after that it is cached by your browser.
Click a timestamp to play from there. Your edits are used in every download.
Advertisement
Advertisement
How to use the Audio Translator to English
Add the audio: drop an audio or video file on the box, choose one, or press Record from microphone.
Pick a model: Tiny is quickest, Small gives the most natural English but downloads about 250 MB once.
Set the spoken language if you know it (better than Auto-detect for short or noisy clips).
Press Translate to English. The English text appears part by part as the audio is processed.
Fix any lines by clicking into them, then copy the translation or download TXT, SRT or VTT.
How Whisper translates speech to English
This page runs Whisper, the open speech model OpenAI released in 2022, inside your browser with WebAssembly. Whisper was trained not only to transcribe speech but also on a large set of recordings paired with English translations, so it can listen to Spanish, Korean or Arabic and write English directly, without first producing a transcript in the original language.
The browser decodes your file and resamples it to 16 kHz mono, the format Whisper expects.
The audio is split into windows of up to 30 seconds, cut at a quiet moment so words are not chopped in half. Silent windows are skipped.
Each window is decoded with the translate task, which outputs English sentences with start and end times. Times are shifted back to the window's position so the SRT lines up with the original video.
If you ask for the original transcript too, the same windows are decoded a second time with the transcribe task and lined up by time.
Which model to use
Model
Parameters
Download (once)
Translation quality
Best for
Tiny
39 million
about 40 MB
Rough, gets the gist
Checking what a clip is about, phones
Base
74 million
about 80 MB
Usable for clear speech
Interviews, vlogs, voice messages
Small
244 million
about 250 MB
Noticeably more fluent and accurate
Subtitles you will publish, lectures, accents
Translation leans on a model's general language knowledge more than transcription does, so the jump from Tiny to Small matters more here than in plain speech-to-text. The "English-only" Whisper variants and the newer "turbo" model were not trained for translation and are not used.
Tips for better translations
Choose the spoken language instead of Auto-detect when you know it. Detection looks at the start of each 30-second part and can be fooled by music, a greeting in another language or mixed-language speech.
Cut music and long silences first with the Audio Cutter. In silent or music-only parts Whisper may invent a short phrase.
Turn on the original transcript for bilingual work: you can check names and numbers in the source language next to each English line.
Proofread names, numbers and idioms. Machine translation can be confidently wrong; treat the output as a draft for anything important, legal or official.
How do I translate an audio file to English for free?
Add the file on this page, choose the spoken language (or leave Auto-detect) and press Translate to English. The open-source Whisper model runs in your browser and writes English text with timestamps. There is no account, upload or usage limit.
Which languages can be translated to English?
Whisper was trained on 99 languages, including Spanish, French, German, Portuguese, Italian, Russian, Arabic, Hindi, Chinese, Japanese, Korean, Turkish, Vietnamese and many smaller languages. Translation quality is best for widely spoken languages with lots of training data and noticeably weaker for rare ones.
Can it translate from English into another language?
No. Whisper's built-in translation only goes from other languages into English. For the opposite direction, transcribe the English audio with Audio to Text and translate the text with a text translator.
Is my recording uploaded?
No. The file is decoded and translated on your own device. The only download is the Whisper model itself (about 40 MB for Tiny, 80 MB for Base or 250 MB for Small), which your browser caches after the first use.
Which model should I choose?
Tiny is fine for clear speech and quick checks. Base handles accents and noise better. Small is much slower and a larger one-time download, but its English is clearly more fluent and accurate, so use it for anything you plan to publish or share.
Can I make English subtitles for a foreign-language video?
Yes. Drop the video file in, translate it and download the SRT or VTT file. Most video players and editors can load it as an English subtitle track, or you can burn it in with the Add Subtitles to Video tool.
Why are some lines wrong or repeated?
Speech translation is harder than transcription: music, crosstalk, slang and names cause mistakes, and Whisper sometimes repeats or invents a short phrase in silent or music-only parts. Choose the correct source language, try a larger model and proofread the result.
Are all the tools on FreeToolCabin tested for quality and accuracy?
Yes! Every tool on our site undergoes automated Q/A testing in real browser environments to ensure high precision, correct output, and zero errors. If you happen to spot a bug or have a suggestion, please contact us at [email protected] and our team will review it right away.