Skip to content
ttsmaker

Transcribe audio to text

Drop an audio file into the box and ttsmaker transcribes it with Whisper, free and without an account. You get text with timestamps and can save it as TXT, SRT or VTT. The first 30 minutes each day run in the cloud; anything beyond that runs in your browser.

Drop an audio or video file here

MP3, WAV, M4A, OGG, WEBM, MP4, MOV

Transcript

Your transcript will appear here

How to transcribe an audio file

  1. 1.Open the recording

    Drag the file into the box or click to pick it from your device. Anything your browser can play, such as an MP3 or a voice memo, will usually work.

  2. 2.Check the language

    Auto is fine for most single-language recordings. Set the language yourself if the file starts with music or with a phrase in another language.

  3. 3.Wait for the pieces

    The progress counter shows how many pieces are done. Text appears as each piece comes back, so you can start reading before the end.

  4. 4.Save the result

    Copy the text or download it: TXT for reading and editing, SRT or VTT if the timestamps matter.

Key facts

  • Recognition: Whisper large-v3-turbo in the cloud, Whisper base in your browser after the free cloud minutes.
  • Free cloud time: 30 minutes of audio per IP address per day, reset at 00:00 UTC.
  • Cloud speed cap: about 10 minutes of audio per minute for each IP address.
  • The recording is converted to 16 kHz mono and split at pauses into pieces of 30 seconds or less.
  • Downloads: TXT, SRT and VTT with timestamps; speaker labels are not included.

What happens to your audio file

  1. Your browser decodes the file. The recording is opened on your own device, not on our server.
  2. The sound is simplified. It becomes 16 kHz mono, which is the format Whisper reads.
  3. The audio is cut at pauses. Whisper reads 30 seconds at a time, so the tool cuts at the quietest moment before each 30-second mark. That keeps words from being split in half. Silent pieces are skipped.
  4. Each piece is transcribed. In cloud mode the piece goes through our server to Cloudflare Workers AI, and the text comes back with start and end times.
  5. The pieces are joined. Timestamps are shifted to their place in the full recording, so the SRT and VTT files line up with the original audio.

How much can you transcribe for free?

Each IP address gets 30 minutes of cloud transcription per day, and the counter goes back to zero at 00:00 UTC. When the minutes run out, or the cloud does not answer, the remaining pieces are handed to Whisper running in your browser. It is slower and a little less accurate, but it has no daily cap and the audio never leaves your device.

For example, a 45-minute lecture uploaded with a fresh counter is split like this: the first 30 minutes are done in the cloud and the last 15 in your browser. The switch is automatic.

How long you wait depends on the length of the file, your connection and which model is working. The cloud is capped at about 10 minutes of audio per minute for each IP address, so a 30-minute recording cannot finish in less than about 3 minutes. Browser mode depends on your device: a recent laptop with WebGPU is usually faster than an old phone.

Which audio formats work?

MP3, M4A, WAV and most other audio files work, because your browser does the decoding; the list of supported formats shows the differences between browsers and what to do with a file that will not open.

How to get a cleaner transcript

  • Record close to the microphone and keep background music low. Whisper hears music as noise and may skip words under it.
  • If the speakers use names, brands or jargon, read those parts twice: Whisper spells unfamiliar words by ear.
  • More tips on noise, accents and wording are on how to improve transcription accuracy.

Curious which Whisper model runs where? See Whisper online.

What this audio transcriber does not do

  • It does not label speakers. An interview comes out as one block of text without “Speaker 1” and “Speaker 2”. For shows with several hosts, see podcast transcription.
  • It does not take links. Download the audio first, then upload the file.
  • It does not join live calls. Record the meeting, then transcribe the recording.
  • It does not translate. The text stays in the language that was spoken.
  • It is not proofread by a person. Check names, numbers and quotes before you publish.

Need text from a video file instead? Video to text uses the same engine but only reads the sound track. All modes are listed on speech to text.

Frequently asked questions

Is it really free to transcribe audio here?

Yes. No account, no watermark, no paid plan (November 2026). The one cap is 30 cloud minutes per day for each IP address, and after that the browser takes over at no cost.

How long can the audio file be?

There is no fixed length or size limit. Because the file is decoded in your browser, a recording of several hours can be slow or fail on a phone with little memory; cutting it into parts helps.

Does the transcript include punctuation?

Yes. Whisper usually writes commas, periods and capital letters on its own, so you do not need to say them. The tool also passes the end of each piece to the next one, which keeps sentences readable across the joins.

Why did the transcript come out in the wrong language?

Auto detection decides from the first piece with speech and keeps that language for the rest of the file. An intro in another language or a song can mislead it, so choose the language by hand and run the file again.

Is my recording stored?

No. Pieces of audio travel through our server to Cloudflare Workers AI and the text comes back to your browser; nothing is kept. In browser mode the audio stays on your device.

Updated: