Lithuanian speech-to-text
FairWe sent 30 Lithuanian test clips through this API. Whisper Large v3 made 28% word errors, about one word in 4; Parakeet made 23%. $0.002 per audio minute, word timestamps, SRT and VTT subtitles, files up to 100 MB.
Measured accuracy
| Model | Word error rate | Character error rate | Clips with no errors | Empty transcripts | Speed |
|---|---|---|---|---|---|
Whisper Large v3openai/whisper-large-v3 | 28% | 8.4% | 0 of 30 | 0 of 30 | 0.076× real time |
Parakeet TDT 0.6B v3nvidia/parakeet-tdt-0.6b-v3 | 23% | 7.1% | 1 of 30 | 0 of 30 | 0.047× real time |
Fair. Gets the content across, but needs a human pass before it is published or quoted. 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 | Senovės tautos ir gentys pradėjo jas laikyti, kad lengvai gautų pieno, vilnos, mėsos ir odų. |
| Whisper Large v3 | Senovis tautos irgentys pradėjo jas laikyti, kad lengvai gautų pieno, vilnos, mėsos ir odu. |
| Parakeet (its median clip) | Spoken: Gali būti naudinga įsigyti kortelę „Wild Card“, su kuria galima įeiti arba į tam tikrus Pietų Afrikos parkus, arba į visus Pietų Afrikos nacionalinius parkus. Gal būti naudingais įgyti kortelę Wild Cards, kuria galima įeiti į tam tikrus Pietų Afrikos parkus arba į visus Pietų Afrikos nacionalinius parkus. |
Language detection
All 10 clips we sent without a language field were identified as Lithuanian. Auto-detection is safe here; sending language=lt still skips the detection step.
Which model for Lithuanian
Parakeet TDT 0.6B v3 was the more accurate model on these clips (23% against 28%) 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 https://api.inferenceapis.com/v1/audio/transcriptions \
-H "Authorization: Bearer $INFERENCE_API_KEY" \
-F "model=nvidia/parakeet-tdt-0.6b-v3" \
-F "language=lt" \
-F "file=@recording.mp3" \
-F "response_format=verbose_json"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="lt", 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 Lithuanian test.
FAQ
How accurate is Whisper for Lithuanian?
On 30 clean read-speech clips it made 28% word errors, about one word in 4. Gets the content across, but needs a human pass before it is published or quoted. Real recordings with noise, accents or several speakers will score worse.
Does language auto-detection work for Lithuanian?
In our test, yes: all 10 clips sent without a language were identified as Lithuanian. Sending language=lt still saves the detection step on every request.
Whisper or Parakeet for Lithuanian?
On the same clips Whisper Large v3 scored 28% and Parakeet 23%. Parakeet was more accurate here and is also the faster model, so start with it. Both cost $0.002 per minute.
Can it translate Lithuanian speech to English?
The API returns Lithuanian text. For English, transcribe first and translate the text with a chat model on the same key.