> For the full page index, fetch https://echonova.ai/llms.txt

# Speech to Text limits and credits

Every limit on the Speech to Text page with the message it shows, what a transcription costs in credits and when the credits are taken, and what you see when the balance cannot cover a file.

This page is the lookup table for the **Speech to Text** page: every limit
and the message you see when a file crosses it, the price of a transcription
in credits and the moment it is charged, and what the page shows when the
workspace balance cannot cover a file. The steps themselves are on
[Transcribe a file](/docs/playground/transcribe-a-file).

## Limits

| Limit              | Value                                                            | Applies to                                                      | Message when crossed                                           |
| ------------------ | ---------------------------------------------------------------- | --------------------------------------------------------------- | -------------------------------------------------------------- |
| Audio length       | 1 hour                                                           | Uploaded files, checked in the browser                          | **Audio must be 1 hour or shorter.**                           |
| Audio length       | 1 hour                                                           | Uploaded files, checked again after upload on the decoded audio | **Audio exceeds the 1-hour limit.**                            |
| Readable length    | The server can measure the audio                                 | The file as sent                                                | **Could not read the audio length. Try another file.**         |
| Microphone take    | 1 hour                                                           | Recording stops on its own and the take is held on the page     | No message                                                     |
| File size          | 150 MB                                                           | Files chosen on the drop area, before upload                    | **This file is larger than 150 MB.**                           |
| File size          | 150 MB                                                           | The file as sent, after any conversion in the browser           | **File too large**                                             |
| Format             | MP3, WAV, FLAC, OGG, M4A, AAC, WEBM                              | Uploaded files, by extension                                    | **Please upload one of: MP3, WAV, FLAC, OGG, M4A, AAC, WEBM.** |
| File type          | Must be audio                                                    | The file as sent                                                | **Unsupported file type: … Please upload an audio file.**      |
| Shortest audio     | 50 milliseconds                                                  | The audio inside the file                                       | A reason starting **Audio is too short**                       |
| Files at a time    | One                                                              | The page sends one file per upload or take                      | No message                                                     |
| Transcription time | 10 minutes after acceptance, then marked failed within 5 minutes | Every transcription                                             | **Job timed out — AI server did not respond**                  |
| Page wait          | 65 minutes after acceptance                                      | The page, if no result has arrived                              | **Timeout has occurred**                                       |

Every message appears in the status line under the box on the page, followed
by a short code in square brackets. Once a file has passed the browser
checks, a **Try again** button appears beside **Transcribe**.

Everyone in the workspace can use the page.

## Get it right

The 150 MB size limit is checked on the file as you chose it, and the
10-minute transcription window applies to every job. Both favour shorter,
compressed files.

**Recommended**

- A one-hour recording exported as MP3 and split into two or three files,
  each transcribed on its own.

**Less effective**

- The same hour as one stereo 44.1 kHz WAV, which passes 150 MB long before
  the hour ends and is refused before upload.

## Credits

| Fact             | Value                                                                                                                                                                                                                                                                                                                      |
| ---------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Rate             | 100 credits per minute of audio, charged per second.                                                                                                                                                                                                                                                                       |
| What is measured | The length of the audio as the server decodes it, not as the file's header claims. A file whose length cannot be measured is refused with **Could not read the audio length. Try another file.**                                                                                                                           |
| Rounding         | Per second, with a minimum of 1 second. The charge is rounded to a whole credit, at least 1.                                                                                                                                                                                                                               |
| When it is taken | After the file is received and passes the checks after upload, before transcription starts.                                                                                                                                                                                                                                |
| Refunds          | The whole charge is refunded if the transcription fails for any reason, including the 10-minute timeout.                                                                                                                                                                                                                   |
| Adjustment       | None. The charge is computed from the length before transcription and does not change when the transcript arrives.                                                                                                                                                                                                         |
| Where you see it | When the transcript arrives, **Used** followed by the credits charged appears under the box. The credit counter in the sidebar, shown as used out of total with the label **Used**, refreshes too. **Credits remaining** in the user menu shows the same balance. The page shows no price before you press **Transcribe**. |

The rate is set by hand and can change; a transcription is priced at the rate in force when it
runs. The rates for every service are on [Credits](/docs/developers/credits).

### Worked examples

At today's rates:

| Audio                 | Charged                     |
| --------------------- | --------------------------- |
| A 10-second note      | 17 credits (16.67, rounded) |
| A 12-minute interview | 1,200 credits               |
| A 15-minute recording | 1,500 credits               |
| A 45-minute lecture   | 4,500 credits               |
| A full hour           | 6,000 credits               |

Every second costs the same; nothing is rounded up to a minute.

## When the balance cannot cover a file

The charge is checked against the workspace balance before anything runs.
If the balance is short, nothing is transcribed and nothing is charged. You
see the message twice: as an error notice titled **Error**, and in the
status line under the box, with a **Try again** button.

| Message                               | Meaning                                                                           | What to do                                                                                        |
| ------------------------------------- | --------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------- |
| **Insufficient subscription balance** | Your credit balance is less than the charge for this file.                        | Open **Subscription & billing** from the user menu and add credits, or transcribe a shorter file. |
| **Subscription has expired**          | No paid plan and no extra credits left. Basic limits apply, with no plan credits. | Choose a plan or add credits under **Subscription & billing**.                                    |

**Try again** sends the same file again. Press it only after the balance
has changed; the check runs again and fails the same way otherwise.

## FAQ

### What does a short clip cost?

Its seconds at 100 credits a minute: a 3-second clip costs 5 credits. Each charge is a whole number of credits.

### Is a refused file charged?

No. Every check before and after upload runs before the charge is taken. A
file refused in the browser never leaves your computer.

### Is a failed transcription charged?

No. The charge is taken when the file is accepted and refunded when the
transcription fails, including when it times out.

### Where can I see what one transcription cost?

The page shows **Used** and the credits charged under the box once the
transcript arrives. Before that, work it out from the length at
100 credits per minute, or watch the sidebar counter before and
after. For a transcription made over the API, the job carries the exact
figure; see [Credits](/docs/developers/credits).

### Does a microphone take cost the same as an upload?

Yes. It is priced by its length at the same rate once it is sent.

### Can I raise the 1-hour limit?

No. Split the recording and transcribe each part.

## Related

- [Transcribe a file](/docs/playground/transcribe-a-file): the steps, and where each message appears.
- [Speech to Text formats](/docs/playground/stt-formats): what the browser does to each format before the size check after upload.
- [Immediate and queued transcription](/docs/playground/sync-versus-async): the 10-minute window and the 65-minute wait in detail.
- [Credits](/docs/developers/credits): rates for every service, and how a key's credit limit works over the API.
