English speech-to-text API: 5.2% word error, measured

Measured accuracy

Model	Word error rate	Character error rate	Clips with no errors	Empty transcripts	Speed
Whisper Large v3 `openai/whisper-large-v3`	5.2%	2.6%	9 of 30	0 of 30	0.065× real time
Parakeet TDT 0.6B v3 `nvidia/parakeet-tdt-0.6b-v3`	5.8%	2.2%	7 of 30	0 of 30	0.062× real time

Excellent. Production quality. Transcripts can be used as they are for notes, search and subtitles. The clips are clean, single-speaker read speech, so treat this as a best case for your own recordings. How we measured, and all 83 languages.

What a transcript looks like

The clip with the median error rate from the test, reference text against what the API returned.

Spoken	However, due to the slow communication channels, styles in the west could lag behind by 25 to 30 year.
Whisper Large v3	However, due to the slow communication channels, styles in the West could lag behind by 25 to 30 years.
Parakeet (its median clip)	Spoken: Danielle Lantagne, a UN expert on the disease, stated the outbreak was likely caused by the peacekeepers. Danielle Landing, a UN expert on the disease, stated the outbreak was likely caused by the peacekeepers.

Language detection

All 10 clips we sent without a language field were identified as English. Auto-detection is safe here; sending language=en still skips the detection step.

Which model for English

The two models are within the noise of a 30-clip test (5.2% and 5.8%). Parakeet is faster; Whisper handles mixed-language audio and has automatic detection across 100 languages. Both cost the same, so try both on a sample of your own recordings. Side by side.

Quickstart

The request is the OpenAI transcription request with a different base URL. language is optional.

cURL

curl https://api.inferenceapis.com/v1/audio/transcriptions \
  -H "Authorization: Bearer $INFERENCE_API_KEY" \
  -F "model=openai/whisper-large-v3" \
  -F "language=en" \
  -F "file=@recording.mp3" \
  -F "response_format=verbose_json"

Python · openai SDK

import os
from openai import OpenAI

client = OpenAI(base_url="https://api.inferenceapis.com/v1", api_key=os.environ["INFERENCE_API_KEY"])
with open("recording.mp3", "rb") as f:
    r = client.audio.transcriptions.create(model="openai/whisper-large-v3", file=f, language="en", response_format="verbose_json")
print(r.text)

Subtitles and timestamps

response_format=srt or vtt returns a subtitle file with cues of up to two lines, broken at sentence ends and pauses. verbose_json with timestamp_granularities[]=word returns a start and end time for every word if you want to build your own.

Cost

$0.002 per minute of audio on either model: $0.12 for an hour, $12 for a hundred hours. Duration is measured from the file you upload. There is no daily cap on minutes and no subscription; enough to run your own English test.

Notes for English

For heavy accents or domain terms, pass a prompt containing the names and jargon spelled the way you want them.

Text-to-speech in English

The reverse direction is covered too: hear the English voices on the English text-to-speech page.

FAQ

How accurate is Whisper for English?

On 30 clean read-speech clips it made 5.2% word errors, about one word in 19. Production quality. Transcripts can be used as they are for notes, search and subtitles. Real recordings with noise, accents or several speakers will score worse.

Does language auto-detection work for English?

In our test, yes: all 10 clips sent without a language were identified as English. Sending language=en still saves the detection step on every request.

Whisper or Parakeet for English?

On the same clips Whisper Large v3 scored 5.2% and Parakeet 5.8%. That is within the noise of a 30-clip test; Parakeet is faster, so try it first on long recordings. Both cost $0.002 per minute.

Can it translate English speech to English?

The API returns English text. For English, transcribe first and translate the text with a chat model on the same key.

Related languages: German (3.4%) · Dutch (7.1%) · Swedish (8.8%)