Portuguese speech-to-text API: 3.4% 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`	3.4%	1.1%	15 of 30	0 of 30	0.06× real time
Parakeet TDT 0.6B v3 `nvidia/parakeet-tdt-0.6b-v3`	4.4%	2.5%	15 of 30	0 of 30	0.047× 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	Concordamos com a declaração do USOC de que os interesses de nossos atletas e clubes, e seus esportes, podem ser melhor servidos ao avançarmos com mudanças significativas dentro de nossa organização, ao invés da descertificação.
Whisper Large v3	Concordamos com a declaração do USOC de que os interesses de nossos atletas e clubes e seus esportes podem ser melhor servidos ao avançarmos com mudanças significativas dentro de nossa organização, ao invés da decertificação.
Parakeet (its median clip)	Spoken: O scaffolding não é um método de aprendizado, mas sim uma ajuda para alunos que estão passando por uma nova experiência de aprendizado, como usar um novo computador ou começar um novo projeto. O scaffolding não é um método de aprendizado, mas sim uma ajuda para alunos que estão passando por uma nova experiência de aprendizado, como usar um computador ou começar um novo projeto.

Language detection

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

Which model for Portuguese

The two models are within the noise of a 30-clip test (3.4% and 4.4%). 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=pt" \
  -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="pt", 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 Portuguese test.

Notes for Portuguese

Brazilian and European Portuguese are both transcribed; the test clips here are Brazilian.

Text-to-speech in Portuguese

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

FAQ

How accurate is Whisper for Portuguese?

On 30 clean read-speech clips it made 3.4% word errors, about one word in 29. 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 Portuguese?

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

Whisper or Parakeet for Portuguese?

On the same clips Whisper Large v3 scored 3.4% and Parakeet 4.4%. 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 Portuguese speech to English?

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

Related languages: Spanish (2.7%) · Galician (12%) · Italian (2.7%)