French speech-to-text API: 6.6% 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`	6.6%	3.0%	6 of 30	0 of 30	0.072× real time
Parakeet TDT 0.6B v3 `nvidia/parakeet-tdt-0.6b-v3`	5.2%	2.2%	8 of 30	0 of 30	0.049× 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	La plupart des petites îles sont des nations indépendantes, ou associées à la France, et sont connues pour être des stations balnéaires de luxe.
Whisper Large v3	La plupart des petites îles sont des nations indépendantes ou associées à la France et sont connues pour être des stations bannères de luxe.
Parakeet (its median clip)	Spoken: La tempête subtropicale Jerry, 10ème tempête de la saison des ouragans à porter un nom, est apparue ce jour dans l'océan Atlantique. La tempête subtropicale Jerry, dixième tempête de la saison des ouragans à porter un nom, est apparue ce jour dans l'océan Atlantique.

Language detection

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

Which model for French

Parakeet TDT 0.6B v3 was the more accurate model on these clips (5.2% against 6.6%) and it is also the faster one. Start with nvidia/parakeet-tdt-0.6b-v3. 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=nvidia/parakeet-tdt-0.6b-v3" \
  -F "language=fr" \
  -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="nvidia/parakeet-tdt-0.6b-v3", file=f, language="fr", 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 French test.

Text-to-speech in French

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

FAQ

How accurate is Whisper for French?

On 30 clean read-speech clips it made 6.6% 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 French?

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

Whisper or Parakeet for French?

On the same clips Whisper Large v3 scored 6.6% and Parakeet 5.2%. Parakeet was more accurate here and is also the faster model, so start with it. Both cost $0.002 per minute.

Can it translate French speech to English?

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

Related languages: Italian (2.7%) · Spanish (2.7%) · Catalan (6.8%) · Occitan (71%)