Playground

Immediate and queued transcription

What the Speech to Text page is doing behind each state of its status line, how long it waits, what happens when you leave, and when the API answers in one call instead.

Every transcription started from the Speech to Text page is a job that runs after your file has been accepted; the page waits for it and shows the transcript when it arrives. Knowing what each state of the status line means, and how long the page will wait, tells you when to sit tight, and when to reach for the API's one-call form instead.

Every transcription from the page is a job

The page never returns a transcript in the same breath as the upload. It sends the file, the file is accepted and charged, and a job runs until the transcript is ready. The page holds on to the job the whole time and shows the result in place, so from your seat it looks like one wait. The status line next to the Transcribe button shows the parts of that wait:

Status line What is happening Charged yet
Preparing audio…, with a moving bar The browser converts an MP3, M4A, AAC or WEBM file to WAV before it sends it. Other files skip this state. No
Uploading, with a percentage The browser sends the file. No
Transcribing, followed by the file's name The file has been received. Once it passes the checks that run after upload, the credits are taken and the transcript is produced. Yes, once the checks pass

The page has no time estimate. A short clip is often back within seconds; an hour-long file is not.

The API offers the other shape as well: POST /v1/speech-to-text/sync returns the transcript in the response for files of up to 10 minutes, and POST /v1/speech-to-text behaves like the page, accepting the same files and letting you poll or receive a webhook. The page always uses the second shape. See Synchronous and asynchronous calls for choosing between them in code.

How long the page waits

The page waits for the transcript for up to 65 minutes from the moment the file is accepted. Long before that, a job that has produced nothing is closed:

  • A transcription that is still running 10 minutes after the file was accepted is marked failed within the next 5 minutes, so at most 15 minutes after acceptance.
  • The status line then shows the reason Job timed out — AI server did not respond, followed by an error code in square brackets, and a Try again button appears next to Transcribe.
  • The credits for that job are refunded.

Try again sends the same file again as a new job with a new charge. Before pressing it on a long recording, split the file: a shorter file finishes sooner and stays inside the 10-minute window.

If the page reaches its own 65-minute limit first, the status line shows the reason Timeout has occurred. The job it was waiting for is already closed by then.

Leaving the page

While a file is uploading or transcribing, the upload box, Record and the Transcribe button are locked. The page stays open, so you can read an earlier transcript from Transcripts on the right while you wait. When the transcript arrives it appears on the Transcript card.

Leaving the page, or reloading it, is different. The job keeps running and its credits stay charged, but this page no longer waits for it: the transcript goes to Assets, and the Speech to Text page starts with an empty card on your next visit. To keep a transcript on the page, stay on it until the card shows it.

Worked example

You drop a 40-minute lecture as MP3 at 10:00 and press Transcribe. The status line reads Preparing audio… for a moment while the browser converts the file, then Uploading climbs to 100%, and at 10:01 it switches to Transcribing with the file's name. You open an earlier transcript from Transcripts and read it on the card while you wait.

If the transcript is ready at 10:07, it appears on the Transcript card and the credit counter in the sidebar refreshes. If nothing has arrived by 10:11, the job is marked failed by 10:16 at the latest, the status line shows Job timed out — AI server did not respond, and the charge is refunded. Split the lecture into two files and transcribe each.

FAQ

Is a short clip transcribed immediately?

It is transcribed quickly, but it still goes through the same accept-then-wait path as a long file. Only the API's POST /v1/speech-to-text/sync returns the transcript in one call.

Can I upload a second file while the first is still processing?

No. The upload box, Record and Transcribe are locked until the current transcription ends. Once it ends, choose the next file.

Am I charged while I wait?

The credits are taken once the file has been received and has passed the checks that run after upload, early in the Transcribing state. They are refunded if the job fails, including when it times out.

What if I close the browser tab?

The job finishes on its own and the transcript is kept in Assets. The credits are not charged twice.

Can I see how far along the transcription is?

No. The status line has the states above and a percentage only during the upload. The transcript arrives whole.