OpenAI Whisper 1 | EmpirioLabs AI Docs

OpenAI · Transcription

POST /v1/audio/transcriptions

Whisper-1 speech-to-text transcription trained on multilingual supervised audio, with a 25 MB upload limit per file.

This model is deprecated and will be retired on 2027-02-25. After that date, requests to this model will fail. Migrate to a successor model before then.

At a glance

Field	Value
Model id	`openai-whisper-1`
Model release date	2022-09-21
Input modalities	Audio
Output modalities	Text
Context window	-
Weight precision	-
Features	transcription, speech_to_text, realtime, streaming_asr
Native inference	No
New	No
Supported endpoints	`POST /v1/audio/transcriptions`, `WEBSOCKET /v1/realtime`
Alternate model ids	`openai-whisper`, `whisper-1`
Deprecation date	2027-02-25

Pricing

Charge	Spec	Rate
Per Minute of Audio	per minute	$0.030

Example request

curl https://api.empiriolabs.ai/v1/audio/transcriptions \
  -H 'Authorization: Bearer $EMPIRIOLABS_API_KEY' \
  -F model=openai-whisper-1 \
  -F file=@meeting.mp3

Parameters

Parameter	Type	Required	Default	Description
`file`	string	yes	-	Audio file (multipart upload) OR use file_url for the JSON path.
`file_url`	string	no	-	Public URL to fetch audio from (alternative to file upload).
`translate`	boolean	no	false	If true, route to /audio/translations and translate to English instead of transcribing in source language.
`timestamps`	boolean	no	false	Convenience toggle. If true, sets response_format=verbose_json and includes word-level timestamp_granularities.
`language`	string	no	-	Optional ISO-639-1 language code. Auto-detected if omitted. Ignored when translate=true.
`prompt`	string	no	-	Glossary or prior context to bias the model.
`response_format`	enum	no	`"verbose_json"`	Defaults to verbose_json, which timestamps=true always uses. · Allowed: `json`, `text`, `srt`, `verbose_json`, `vtt`
`temperature`	number	no	`0.0`	Sampling temperature. · Range: 0 – 1
`timestamp_granularities`	string	no	-	Comma-separated list: word, segment. Used when response_format=verbose_json.

Notes

Upload a file to POST /v1/audio/transcriptions, or transcribe live over a WebSocket at wss://api.empiriolabs.ai/v1/realtime?model=openai-whisper-1. The socket takes 24 kHz mono 16-bit PCM as base64 in input_audio_buffer.append and returns transcript events; voice activity detection decides when each phrase ends, so you do not send a commit. Both surfaces bill per minute of audio at the same rate.

Machine-readable schema: GET https://api.empiriolabs.ai/v1/models/openai-whisper-1.