Apple Music Hear your music the way it was mastered. Get up to 3 months of Apple Music free, with lossless audio and Dolby Atmos. Try it free Ad · New subscribers only
SoundBench فارسی
Guides

How to Transcribe Audio to Text for Free

Whether you have a recorded interview, a lecture or a voice memo, you can turn the speech into editable text in a few minutes. Here is how to do it, and how to get the most accurate result.

Typing out a recording by hand can take several times the length of the audio. Automatic speech recognition does the first draft for you, and you only need to read through it and fix the odd word. This works for interviews, meetings, classes, podcasts, voice notes and even the soundtrack of a video.

Many transcription websites upload your file to a server and limit the free minutes. The SoundBench Audio to Text tool works differently: it runs an AI speech model on your own device, so your recording stays private and there are no credits or accounts.

What you need before you start

You need an audio file or a way to record one. Common formats such as MP3, WAV, M4A and FLAC all work. If your recording is a video, extract its audio first. If you have nothing recorded yet, you can capture the speech directly in the browser.

Accuracy depends mostly on the recording, not on the tool. Clear speech, one person talking at a time and a microphone close to the speaker make the biggest difference.

Step by step: transcribe with SoundBench

The SoundBench Audio to Text tool at soundbench.net/audio-to-text/ uses OpenAI's open Whisper model, which understands more than 90 languages. The model is downloaded the first time you use it and then runs locally.

  • Open soundbench.net/audio-to-text/ and drop your audio file onto the page, or record with your microphone.
  • Choose the language, or let the tool detect it. For languages other than English, pick the Accurate setting.
  • Pick Fast, Balanced or Accurate, then click Transcribe and watch the text appear.
  • Copy the text, or download it as TXT. Download SRT or VTT if you want subtitles with timestamps.

Prepare the audio for better accuracy

A few small steps before transcribing can reduce mistakes. If a recording has steady hiss, hum or fan noise, run it through the SoundBench Noise Remover at soundbench.net/remove-background-noise/ first. If the voice is very quiet, raise it with Change Volume at soundbench.net/change-volume/.

If only part of a long file matters, cut it with Trim Audio at soundbench.net/trim-audio/ so you do not wait for the whole recording. Starting a new voice memo? The online recorder at soundbench.net/online-voice-recorder/ saves your take in full quality and can pass it on to the other tools.

Transcribing a video or subtitles

Videos need one extra step. Use Video to MP3 at soundbench.net/video-to-mp3/ to pull the soundtrack out of the clip, then transcribe the resulting audio. The SRT file you download can be loaded into most video players and editors as subtitles.

Choosing the accuracy level

The three accuracy settings trade speed for quality.

Always read the result once before you use it. Names, technical terms and overlapping voices are the most common trouble spots.

SettingBest forSpeed
FastClear English, quick draftsQuickest
BalancedEveryday recordingsMedium
AccurateOther languages, hard audioSlowest

Tips for a cleaner transcript

When you record an interview, place the microphone near the speaker, choose a quiet room and ask people not to talk over each other. Test your setup first with the SoundBench Mic Test at soundbench.net/mic-test/.

Keep sentences of the final text short and check them against the audio wherever the result looks odd.

Frequently asked questions

Is the transcription really free?

Yes. SoundBench has no minutes, credits or accounts for this tool. The AI runs on your device, so there is nothing to charge for.

Is my recording uploaded anywhere?

No. The speech model is downloaded once and the transcription runs on your own device, so private interviews and meetings stay with you.

How long does it take?

It depends on your device and the accuracy setting. On a modern computer the Fast setting handles an hour of audio in a few minutes, while Accurate takes several times longer. Phones are slower.

Related tools