Question 1

Is it actually free?

Accepted Answer

3 transcripts a day are free, with no account and no card. The count resets at midnight IST. What bounds the cost is length, not file size — 15 minutes per recording on the free tier — so one person can't run up the bill for everyone else.

Question 2

How long can the file be?

Accepted Answer

About 15 minutes, or 500 MB — whichever comes first. Most speech recordings hit the time limit long before the size one. On the paid plan a single file can run to three hours.

Question 3

What does the paid plan cost?

Accepted Answer

₹9 for the first 3 days, then ₹99 a month, on UPI AutoPay or card. It removes the daily limit and raises the per-file cap to three hours. Fair use is 8 hours of audio during the trial and 20 hours a month after that. Cancel any time — access runs to the end of the period you paid for.

Question 4

What happens to my file?

Accepted Answer

It's uploaded to storage we control and transcribed there. Nothing is used to train a model, and no one outside decibl can reach it. There's no account here, so nothing shows you a list of past uploads: if you want a recording removed, email team@decibl.ai with roughly when you uploaded it and we'll find it.

Question 5

Can I record instead of uploading?

Accepted Answer

Yes. Hit the mic icon and record up to 90 seconds in the browser. There's a timer so you can see the limit coming. Anything longer needs to be a file.

Question 6

How accurate is it?

Accepted Answer

Clean, single-speaker English comes back close to verbatim. Accuracy falls with crosstalk, background noise and low bitrates. We'd rather you tested it on one of your own recordings than took a percentage from us.

Question 7

Which formats work?

Accepted Answer

.mp3, .m4a, .wav, .ogg, .opus, .flac, .aac, .webm, .mp4, .mov. Video files work too — we read the audio track. WMV, AVI and RMVB don't work; convert to MP3 or MP4 first.

Question 8

Does it label who's speaking?

Accepted Answer

Best effort. Tick "Label speakers" and the model prefixes each turn with Speaker 1, Speaker 2. It's prompt-based rather than true diarization, so it can merge two voices when people talk over each other. Check it before you rely on it.

Question 9

Can it handle a recording that isn't in English?

Accepted Answer

Yes. Leave "Audio language" on auto and it works out what's being spoken. If the audio is noisy or code-mixed, setting it explicitly usually helps.

Question 10

Can it translate what was said into another language?

Accepted Answer

Yes — that's what "Transcript language" does. Set it to any of the 33 languages and the transcript comes back in that language regardless of what was spoken. Leave it on "Same as spoken" for a verbatim transcript.

Question 11

Can I export subtitles or a Word file?

Accepted Answer

Not today. You get plain text to copy or download as .txt. If SRT or DOCX would decide it for you, tell us on a demo call and we'll know it's worth building.

WAV to Text

Pricing

Transcription

Get unlimited transcription

Questions about this tool

Free Tools