Catalan speech-to-text API: 6.8% 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`	6.8%	2.8%	10 of 30	0 of 30	0.057× 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	No cal dir que si coneixeu una llengua romanç us serà més fàcil d'aprendre portuguès.
Whisper Large v3	No cal dir que si coneixeu una llengua romans us serà més fàcil d'aprendre portuguès.

Language detection

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

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=ca" \
  -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="ca", 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 Catalan test.

FAQ

How accurate is Whisper for Catalan?

On 30 clean read-speech clips it made 6.8% word errors, about one word in 15. 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 Catalan?

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

Can it translate Catalan speech to English?

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

Related languages: Spanish (2.7%) · Occitan (71%) · French (6.6%)