Swedish speech-to-text API: 8.8% 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`	8.8%	3.1%	12 of 30	0 of 30	0.057× real time
Parakeet TDT 0.6B v3 `nvidia/parakeet-tdt-0.6b-v3`	18%	6.1%	3 of 30	0 of 30	0.054× real time

Good. Usable with light review. Expect to correct a word here and there, mostly names and numbers. 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	Resenärer rekommenderas starkt att vara medvetna om risken för hårt väder som påverkar deras område, eftersom det kan påverka resplanerna.
Whisper Large v3	Resenärer rekommenderar starkt att vara medvetna om risken för hårt väder som påverkar deras område eftersom det kan påverka resplanen.
Parakeet (its median clip)	Spoken: Därför var blyertspennan en trogen följeslagare för många människor när den kom ut. Därför var byrspännen en trugen följeslagare för många människor när den kom ut.

Language detection

Send language=sv. Of 10 clips sent without it, only 8 were identified as Swedish; the others were labelled Nynorsk (2). When detection is wrong, Whisper writes the transcript in the language it guessed.

Which model for Swedish

Whisper Large v3 was more accurate on these clips (8.8% against 18%). Parakeet is faster, which matters mainly on long recordings. Both cost the same. Side by side.

Quickstart

The request is the OpenAI transcription request with a different base URL. language is optional in the API but, as shown above, not optional in practice for Swedish.

cURL

curl https://api.inferenceapis.com/v1/audio/transcriptions \
  -H "Authorization: Bearer $INFERENCE_API_KEY" \
  -F "model=openai/whisper-large-v3" \
  -F "language=sv" \
  -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="sv", 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 Swedish test.

FAQ

How accurate is Whisper for Swedish?

On 30 clean read-speech clips it made 8.8% word errors, about one word in 11. Usable with light review. Expect to correct a word here and there, mostly names and numbers. Real recordings with noise, accents or several speakers will score worse.

Does language auto-detection work for Swedish?

Not reliably. Of 10 clips sent without a language, 8 were identified as Swedish; the rest were labelled Nynorsk (2). A wrong label means the transcript is written in the wrong language, so always send language=sv.

Whisper or Parakeet for Swedish?

On the same clips Whisper Large v3 scored 8.8% and Parakeet 18%. Whisper was more accurate here; Parakeet is faster, which matters mainly for long recordings. Both cost $0.002 per minute.

Can it translate Swedish speech to English?

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

Related languages: Norwegian (9.7%) · Danish (15%) · Nynorsk