Skip to content
PegaFalaPortuguês

About PegaFala

Updated on October 25, 2026

PegaFala is a free site that turns audio into text. No sign-up, it works on phones and computers, and it deletes the audio as soon as the transcript is ready.

What you can do

The result comes as text to copy and in three files: TXT, SRT and VTT.

How it works

The audio you pick goes to our server, runs through a speech recognition AI (Whisper, which is open source) right there, and the text comes back to your screen. Transcripts are made one at a time: at peak times there may be a short line, and the screen shows your turn. Only when the line is full can a backup service do the transcript.

The audio is deleted right after

A voice message is a conversation, so the site is built to keep nothing:

  • the file sits in a temporary folder, in the server’s memory, only while it’s transcribed;
  • it’s deleted right after, whether it worked or not;
  • the text isn’t stored on the server: it goes to your device and disappears when you close the tab;
  • file names and transcripts never go into the site’s logs.

Details in the privacy policy.

Who runs it

The site is run by Isabele Morais, an individual based in Brazil, who is responsible for the service, the little data handled here and removal requests.

PegaFala is independent. It is not affiliated with, sponsored or endorsed by WhatsApp or Meta. WhatsApp is a registered trademark of WhatsApp LLC. Instagram and TikTok are trademarks of their owners.

How the site pays for itself

With Google AdSense ads. They pay for the server, which does the transcription, and the domain. You pay nothing, and we don’t sell data about the people who use the site.

Rules we follow with ads:

  • no ads inside the tool card, on the “Uploading” and “Transcribing” screens or over the text;
  • no countdowns, pop-ups, pop-unders or requests to turn on notifications;
  • no button in the tool is an ad. If something looks like a fake button, let us know.

Limits and known issues

5 minutes per file. Longer files get the first minute as a preview. Every minute uses the server’s processor, which serves everyone; the limit keeps the site free and fast.

4 transcripts every 15 minutes per connection, with no daily cap. Audio with no speech and failed attempts don’t count.

25 MB per file. Plenty for a voice message; HD video and WAV go over quickly.

No speaker labels. The text comes out as one block, with punctuation.

Not perfect. Noise, music and overlapping voices get in the way. It isn’t a certified transcript for court.

Get in touch