GPT Transcribe

Upload a Recording, Get Editable Text in Seconds — OpenAI Speech-to-Text

GPT Transcribe,
ready for real recordings.

Upload audio or video and turn it into editable text. Start with GPT Transcribe; switch when the job needs speaker labels or a different model.

25 MB
Maximum upload size
Live quote
Model and duration pricing
3
OpenAI transcription routes
TRANSCRIPTION / 01

Source recording

Your selected recording stays on this device while you preview it. When you start transcription, it is uploaded to the transcription provider for processing. Privacy details.

Transcription quality
Need a different route or speaker labels?Choose an option for a specific job.
Add a recording for a quote2 credits / 10 min · the exact quote appears after the duration check

MODEL ROUTES / CHECKED JULY 30, 2026

One default.
Two deliberate switches.

Use gpt-transcribe for the default accuracy-first workflow, gpt-4o-mini-transcribe for lightweight draft or high-volume work, or gpt-4o-transcribe-diarize when the transcript needs speaker turns and timestamps.

01
gpt-transcribe

Current general transcription route

Default
02
gpt-4o-mini-transcribe

Lightweight token-priced route

Drafts
03
gpt-4o-transcribe-diarize

Speaker labels and timestamps

Speakers
Official referencesSpeech-to-text guide API pricing

WORKFLOW / FOUR MOVES

From recording to working text.

  1. 01

    Upload your recording

    Drop in an MP3, MP4, M4A, WAV, MPEG, MPGA, or WebM file up to 25 MB.

  2. 02

    Pick the right route

    Keep gpt-transcribe for the best general-purpose result, choose mini for lightweight draft work, or pick diarize when the transcript needs speaker names and timestamps.

  3. 03

    Add context for better accuracy

    Drop in a short note — speaker names, product names, acronyms — before the job starts. The model uses it for spelling, not as a script.

  4. 04

    Review and export

    Read the transcript in the browser, copy it to the clipboard, or download a TXT file for editing, captions, or notes.

PRICING / MODEL + DURATION

Know the charge before you start.

Credit cost follows the selected model and source duration, not transcript length. Reserved credits return only after OpenAI confirms failure; uncertain network outcomes are held for reconciliation.

Pay month to month, receive a fresh credit grant after every successful renewal, and cancel anytime.

Start here

Free

Transcribe one short recording.

Freeto start
2 welcome credits
Estimated capacityUp to 10 transcription minutesUp to 5 minutes per credit on the lowest-cost model
  • 2 welcome credits
  • Model-based quote before transcription
  • All transcription models
  • TXT export
Monthly plan

Studio

For campaigns and higher-volume production.

$79/ month
330 credits / month
$0.239 / credit
Estimated capacityUp to 1,650 transcription minutesUp to 5 minutes per credit on the lowest-cost model
  • 330 credits every month
  • Credits roll over
  • Generation history
  • Email support
Monthly plan

Max

For larger workloads, with adjustable credit capacity.

$158/ month
660 credits / month
$0.239 / credit
Estimated capacityUp to 3,300 transcription minutesUp to 5 minutes per credit on the lowest-cost model
  • Monthly credit grant
  • Scale up to 5x credits
  • Credits roll over
  • Generation history
  • Email support
  • Cancel anytime

Capacity uses the lowest-cost live setup for this site. Your exact credit cost changes with the selected model and settings and is shown before generation.

Plan prices are workspace prices, not provider charges. Every route quotes from its own credit rate, and token-priced routes keep a separate 2× variability budget inside that quote. The displayed credit charge will not increase after submission.

Billing details

HELP / PRACTICAL ANSWERS

Before you upload.

Which model does GPT Transcribe use?

The workspace defaults to gpt-transcribe, the model OpenAI currently recommends starting with for recorded speech. You can also select GPT-4o Mini Transcribe or GPT-4o Transcribe Diarize.

How are credits calculated?

Credits are calculated from the selected model and recording length. The exact quote is shown before transcription starts, so you always see the credit cost for that recording before you submit it.

Which files can I upload?

MP3, MP4, MPEG, MPGA, M4A, WAV, and WebM files are supported up to 25 MB.

Can the transcript identify speakers?

Yes. Choose GPT-4o Transcribe Diarize for speaker-labelled segments and timestamps.

Can I add names or technical terms?

Yes. Add short context before transcription to improve how names, acronyms, and specialist vocabulary are represented.