Provisional fares. Beta pricing. Your fare is fixed when you submit a batch. View pricing details

Whisper large-v3

Speech to text. Sold per minute of audio, by the second, 30 second minimum per file, up to 240 minutes.

List fare, per audio minute: $0.00044
Launch fare, from 30 audio minutes: $0.0004
Minimum, per submission: $0.0075
1,000 audio minutes, one batch: $0.44; $0.40 at the launch fare

The minimum stops mattering at 18 audio minutes. Below that, a submission pays $0.0075 whatever its size. A file under 30 seconds bills as 30 seconds.

Cold start: expect 10 to 30 s, occasionally minutes. This is a queue, and later items in a batch ride on the same start.

Using it.

Set language if you know it. Left empty, the model detects it.
A transcription that finishes bills its units even if the transcript comes back empty.
Audio is measured from decoded samples, so size max_units from the real length, and no lower than the smallest in Table 1.

Table 1 Limits
Model ID	`whisper-large-v3`
Input types	`audio/mpegaudio/mp4audio/wavaudio/flacaudio/oggaudio/webmvideo/mp4`
Largest file	500 MB
Largest item	240 audio minutes
Default `max_units`	60
Smallest `max_units`	0.5
Licence	MIT (OpenAI), faster-whisper conversion MIT

Table 2 Parameters, all optional
Name	Type	Allowed	Default
`language`	string	ISO language code, two or three lowercase letters	detected from the audio
`format`	string	json, text, srt, vtt	json

Dry run, which writes nothing

curl -s https://www.y-pad.com/v1/batches \
  -H "Authorization: Bearer $YPAD_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "whisper-large-v3",
    "items": [{"file": "FILE_ID", "params": {"language":"en"}}],
    "dry_run": true
  }'

The first-run guide covers the upload before this and the rest of the batch after it.

Real output.

From the same worker you would pay for, on a public-domain input, and cached here.

Transcribed

chapter i the beginning of things things. They were not railway children to begin with. I don't suppose they had ever thought about railways except as a means of getting to Maskelyne and Cook's, the Pantomime, Zoological Gardens, and Madame Tussauds. They were just ordinary suburban children, and they lived with their father and mother in an ordinary, red-brick-fronted villa, with coloured glass in the front door, a tiled passage that was called a hall, a bathroom room with hot and cold water electric bells french windows and a good deal of white paint and

The Railway Children, chapter 1, the first 52 seconds. Read by Kara Shallenberg for LibriVox. Text by E. Nesbit. Public domain. Public Domain Mark 1.0. LibriVox states its recordings are public domain in the USA. Source.

Who charges less.

DeepInfra beats us on a short Whisper job. It charges $0.00045 per minute, so up to 16 audio minutes it costs less than our $0.0075 minimum.
Runware beats us on a lone image. It lists BiRefNet at $0.0006 per image and shows Real-ESRGAN at $0.0013 at 4x, varies with input, while a lone job here pays the $0.0075 minimum.
A lone job here costs more than on Replicate. Ours pays the $0.0075 minimum. Replicate charges $0.002 per image for Real-ESRGAN and about $0.0017 per run for BiRefNet.

Table 3 Whisper large-v3, our launch fare first, then lowest first, per 1,000 audio minutes
Who	Published price	Per 1,000	Against our fares	Checked	Source
Y-PAD, launch fare	$0.0004 per audio minute$0.0075 minimum per submission, 30 second minimum per file	$0.40	Ours	2 Nov 2026	Fares
Y-PAD, list fare	$0.00044 per audio minute$0.0075 minimum per submission, 30 second minimum per file	$0.44	Ours	2 Nov 2026	Fares
DeepInfra	$0.00045 per minutepublished price	$0.45	More than both our fares	2 Nov 2026	deepinfra.com
Together	$0.0015 per minutepublished price	$1.50	More than both our fares	2 Nov 2026	together.ai
Groq	$0.111 per hour, 10 s minimumpublished price, converted	$1.85	More than both our fares	2 Nov 2026	console.groq.com

Whisper large-v3

Using it.

Real output.

Transcribed

Who charges less.

Join the private beta

You're on the list