Italian speech-to-text API: 2.7% 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`	2.7%	0.8%	19 of 30	0 of 30	0.055× real time
Parakeet TDT 0.6B v3 `nvidia/parakeet-tdt-0.6b-v3`	2.4%	0.9%	21 of 30	0 of 30	0.039× 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	Le squadre antincendio alla fine sono riuscite a domare il fuoco alle 23:35.
Whisper Large v3	Le squadre antincendio alla fine sono riuscite a domare il fuoco alle 23.35.
Parakeet (its median clip)	Spoken: Condividere una gita scolastica virtuale è inoltre un ottimo pretesto per condividere esperienze e riflessioni di viaggio con le classi future. Condividere una gita scolastica virtuale è inoltre un ottimo pretesto per condividere esperienze e riflessioni di viaggio con le classi future.

Language detection

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

Which model for Italian

Parakeet TDT 0.6B v3 was the more accurate model on these clips (2.4% against 2.7%) 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=it" \
  -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="it", 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 Italian test.

Text-to-speech in Italian

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

FAQ

How accurate is Whisper for Italian?

On 30 clean read-speech clips it made 2.7% word errors, about one word in 37. 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 Italian?

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

Whisper or Parakeet for Italian?

On the same clips Whisper Large v3 scored 2.7% and Parakeet 2.4%. Parakeet was more accurate here and is also the faster model, so start with it. Both cost $0.002 per minute.

Can it translate Italian speech to English?

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

Related languages: Spanish (2.7%) · French (6.6%) · Portuguese (3.4%) · Romanian (10%)