How accurate is speech recognition in Romanian?

Romanian is solid but not top tier: about 10 to 12 wrong words per 100 on read speech. In our October 2026 benchmark on a Mac with an M2 Pro, Parakeet TDT 0.6B v3 had a word error rate of 10.00 percent ¹ and Whisper large-v3-turbo had 11.79 percent ², on 100 Romanian recordings from the FLEURS test set. We have no Apple SpeechTranscriber result for Romanian.

Parakeet is about two points better than Whisper in Romanian

Parakeet has the lower error rate and is also the faster engine, at about 20 times faster than real time ³, against about 12 times for Whisper ⁴. The gap in accuracy is small on 100 recordings, but it points the same way as the published figures below. The speed dataset covers every language.

Our measurement of the engines in Romanian

Lower is better in this table.

The transcripts behind each row are on the on-device benchmark page.

Published results put Romanian in the same band

NVIDIA reports 12.44 percent for Parakeet on FLEURS Romanian ⁵, a little worse than our 10.00 percent. OpenAI's paper gives 14.4 percent for the older Whisper large-v2 ⁶, worse than the turbo model we ran. Different runs and model versions move the figure by a couple of points, but Parakeet leads in every comparison.

All published rows are on the published accuracy page.

Romanian ranks fourteenth of 25 languages

Taking the best engine in each language, Romanian is fourteenth of the 25 in the benchmark. It is well behind Spanish (2.53 percent with Whisper ⁷) and Finnish (7.73 percent ⁸). It is level with Swedish (9.95 percent ⁹) and Czech (10.44 percent ¹⁰), and ahead of Slovenian (19.83 percent ¹¹).

Formatting adds about three points

Counting punctuation and capital letters as a reader sees them, Parakeet's text differed from the written sentences by 12.93 errors per 100 tokens ¹², and Whisper's by 13.99 ¹³. If you care about commas and capitals, expect a few more fixes than the word error rate alone suggests.

What to expect when you dictate in Romanian

In a quiet room, expect roughly 1 wrong word in 10 with Parakeet and a little more with Whisper, so proofread every dictated passage. Errors are most likely on names, numbers, rare words and diacritics, the letters with marks such as ă, â, î, ș and ț. Parakeet is the first engine to try and it runs entirely on the Mac. Whisper is a reasonable second choice if your app offers only that. Speak in full phrases, and keep a correction list if your app has one. We found no published fact that confirms Romanian as a built-in Mac or Windows dictation language, so check your system's language list.

Choosing between the two engines matters little in Romanian

With two points between them, the choice of engine is mostly about which one your app supports. Parakeet is both more accurate and faster in our test, so it is the default pick. Dictate one page of your usual text through each engine you can run and count the corrections, since names and technical words will behave differently from the read sentences used here. The published accuracy page shows how other models compare on similar tests.

What these numbers do not show

The recordings are people reading prepared sentences in a quiet room, scored on one Mac with 100 recordings. Spontaneous speech, noise, regional accents and domain vocabulary are not in the test, and we did not measure Apple's model. Whether a transcript uses the right diacritics is only partly captured by the word error rate. For the measure itself, see the guide on how accurate speech recognition is.

Our guide to European languages in speech to text helps you choose an engine, and the languages overview covers other languages. For comparison, read the same question for Hungarian, Czech and Bulgarian.