Skip to main content

openai/gpt-4o-mini-transcribe/speech-to-text

GPT-4o Mini Transcribe is OpenAI's compact speech-to-text model: send one recording of up to 25 MB or two hours, add a language code or spelling hints, and receive a written transcript.

From $0.00005/s

Workbench

Input

Speech to text
0/2048
Spelling and context in the spoken language at most 2048 characters
1/1
1 / 1 files
Voice memo to convert to text, up to 25 MB at least 1 characters
Two-letter language of the recording at least 2 characters · at most 2 characters
Sampling temperature between 0 and 1 no lower than 0 · no higher than 1

Checking your account before enabling submission and pricing.

Output

Ready for a new run
Example result
Model response

Related models and next tasks

Missing the model you need?

LIGHTWEIGHT VOICE-NOTE TRANSCRIPTION

Turn everyday voice memos into separate, editable text drafts.

Key numbers

Context / structure
prompt
Keep each note's identifier beside its task so a completed transcript returns to the correct reminder or observation.
File limit
25 MB
Each Mini memo must stay within 25 MB and 7200 seconds; split oversized recordings before submission.
Audio limit
2 h
Each Mini memo must stay within 25 MB and 7200 seconds; split oversized recordings before submission.
Minimum billing
1 s
Measured audio seconds

Before you spend

All modes
  • Each Mini memo must stay within 25 MB and 7200 seconds; split oversized recordings before submission.
  • Mini's file workflow begins after recording ends; it is not a live dictation stream.
  • Short vocabulary hints can help, but Mini does not guarantee exact spelling of codes or personal names.
  • There is no promised completion time for a queued memo, and this endpoint does not return speaker attribution.
  • OpenAI has marked GPT-4o Mini Transcribe as deprecated, with service ending on 2027-02-26; Mini stays usable and billed as usual until that date. If you are starting a new voice-note transcription project, choose GPT Transcribe instead of integrating Mini.

What it does

Small lime sound capsules sit beside blank porcelain tiles connected by golden light.

Voice-note drafts

One memo, one text entry

Keep each note's identifier beside its task so a completed transcript returns to the correct reminder or observation. For three separately dictated reminders, keep three recording-to-result associations. Your application can offer each reminder as an editable draft without guessing where to split one long transcript.

Voice-note drafts

Hints for everyday names

A shallow lime-green glass resonator on a black pedestal, outlined by a fine golden arc.

Use a brief prompt for a product name or place mentioned in the memo, without replacing the words actually recorded. Provide only vocabulary that belongs to that recording. A spelling hint is context for recognition, so the person reviewing the draft still needs access to the original note.

Miniature lime acoustic sculptures sit in separate dark trays crossed by a gold arc.

Voice-note drafts

Plan a queue by audio length

Mini is the lower-cost, speed-focused option; billing follows recorded seconds rather than how many words a note contains. This makes the total length of a note queue useful for estimating usage. The speed-focused positioning is not a fixed completion deadline for an individual task.

Where it fits

  • Field observations

    Save a technician's dictated finding next to the equipment record.

  • Personal reminders

    Convert an individual spoken reminder into a draft you can edit.

  • Voice journals

    Let the writer review each recorded thought as its own paragraph.

  • Mobile note inboxes

    Collect completed memo tasks independently as recordings arrive.

Model reference

GPT-4o Mini Transcribe API reference

GPT-4o Mini Transcribe API specifications

Input
GPT-4o Mini Transcribe · Audio URL · MP3 / MP4 / MPEG / MPGA / M4A / WAV / WEBM
Language
Optional ISO 639-1
Temperature
0–1; default 0
Billing
Measured audio seconds
Delivery
Asynchronous task

How to call the GPT-4o Mini Transcribe API

  1. Record a clear note

    Choose a short saved memo with audible speech for your first Mini task.

  2. Keep notes distinct

    Submit each recording separately with its own identifier and optional vocabulary hint.

  3. Return the draft to its owner

    Retrieve the completed text and let the author correct names before using it.

The short version

Submit each recording separately with its own identifier and optional vocabulary hint. Retrieve the completed text and let the author correct names before using it.

GPT-4o Mini Transcribe API questions

All GPT models18