Audio to Text
Audio to Text, Whatever You Recorded
It looks at the file itself — its format, its length, its size — and never at what you recorded. Drop one in and you will know right away whether it is taken.
60 one-time signup credits · No credit card required · A completed recording required
Create a transcript
Add a supported link or upload a recording to get searchable text.
It Never Asks What You Recorded
Six things people arrive holding, and what happens to each.
- I have a voice memo I recorded on my phone.
- Taken. If the file ends in .m4a, .mp3, or .wav, it goes straight in.
- Someone sent me a voice message and the file ends in .ogg or .opus.
- Not taken. The upload box turns it away in your browser, before anything leaves your computer. Convert it to .mp3, .m4a, or .wav first.
- What I have is a video file, not an audio file.
- Taken — .mp4, .mov, and .webm all upload. Only the sound becomes text.
- I have a link, not a file.
- Taken, if the link points at the file itself — one ending in .mp3, .m4a, or .wav. The box beside this page also recognises episode links from the services listed under it. What it cannot use is a page that merely has a player on it.
- There are several people talking in my recording.
- Taken. The text comes back in the order it was spoken. It will not tell you which voice said which line.
- My recording is long, and the file is big.
- Taken up to 500 MB and 6 hours. Past either of those, it will not run.
Uploading Is Not Starting
When the upload finishes, nothing has started. The file is sitting there as a draft, and PodText stops to wait for you.
While it waits, the real length of the recording is measured from the file itself rather than guessed in your browser. You can name the spoken language or leave it blank and let it work that out. What this run will cost you is on screen before you agree to it.
Credits leave your balance when you press start, not when the upload finishes. PodText works from a recording that already exists: it does not transcribe live, and it cannot join something still being broadcast.
You Do Not Have to Trust It — You Can Check It
Speech recognition gets things wrong. Names, numbers, and the exact wording of a sentence you are about to quote are where it goes wrong most. PodText offers no accuracy guarantee.
So this page does not ask you to believe the text. Every segment carries a time reference back into the recording, which means anything you are about to rely on can be heard again in its own place. That check has a window: when a transcript was paid for with nothing but the one-time signup credits, its audio is removed seven days after the transcript finishes. The text, the time references, and the exports stay; the sound is what goes.
So if there is something you need to verify, do it while the recording is still there.
Start With the File You Already Have
Put it in, read what the confirmation tells you, and decide then. Nothing is charged until you press start.