Free Audio Translator (+ Guide)

What is an audio translator?

An audio translator turns speech in one language into text in another. This one writes English from any language Whisper knows, and keeps the time of every line.

OpenAI describes Whisper as "a multitasking model that can perform multilingual speech recognition, speech translation, and language identification."

How it translates

Whisper writes what's said in its own language first. Whisper's own translation is rougher: on our Spanish example, whisper-base wrote "the Sorra and the Uvas" for the fox and the grapes.

So the text goes to AI on our server, which reads the lines around each one and translates the meaning. Only the text is sent, never the recording.

English subtitles

SRT and VTT keep every line's time, so they work as English subtitles. Upload either to YouTube, Vimeo or your video editor.

Check names and places before you publish: a translation is only as clear as the recording.

Other languages

Translate to, under the file, takes the transcript into 1 of 30 languages the same way. To switch back, pick the original at the top of the list.

To keep the original language only, use the audio to text converter or Spanish transcription.

On your device or on our server

The page opens on your device. Whisper, OpenAI's speech recognition model, runs in your browser, and the file never leaves it.

For long files, pick On our server. The sound goes in 2-minute pieces to Groq's Whisper and comes back as text in seconds.

Groq, the company that runs the server's Whisper, says in its data policy that "by default, Groq does not retain customer data for inference requests." We don't keep the sound either.

Standard or Best

Standard is whisper-base, a 77 MB download that runs in any modern browser. On a 2021 MacBook Pro, it wrote a 23-second clip in 5 seconds.

Best is whisper-large-v3-turbo, 563 MB, on browsers with WebGPU. It wrote 100 seconds in 10, and a Spanish clip word for word, after its download.

Without WebGPU, Best is whisper-small, 249 MB, which is slower. Each model downloads once, and your browser keeps it.

Files it takes

Anything your browser can read: MP3, M4A, WAV, AAC, FLAC, OGG, Opus, MP4, MOV, WebM and MKV. A video's sound is read, and the picture stays on your screen to play along.

The file is read in parts, so an hour-long recording works on a computer. A phone can take longer on the device, and On our server is quicker there.

Try the example

The example is a 52-second Aesop fable in Spanish, "La zorra y los racimos de uvas," read by Paulino for LibriVox. It's in the public domain.

After you translate

Pull quotes and posts from the English with the asks in step 3. A dubbed voice that sounds like the speaker needs a paid voice clone, which this tool doesn't make.

Who is the audio translator for?

It's for anyone who needs a recording in another language:

• Spanish transcription – people working with Spanish audio
• Audio to text converter – anyone who wants the original words first
• Video caption generator – creators adding translated subtitles
• Podcast transcript generator – podcasters reaching new listeners

The rest of my tools are on the free tools page.