News for hackers. Stuff that matters.
Re:Show HN: State-of-the-art German speech recognition in 28... by fxtentacle on Wednesday August 10, 2022 @09:46AM (#32411826)
"This takes 2 weeks of A100 GPU time to train?"

Yes, but only for me in this specific case. That's because this German model was derived based on a model which was pre-trained for months on hundreds of thousands of hours of YouTube audio in various languages. So if I now train an English recognizer, I don't start from scratch. Instead, I start from a checkpoint which can already almost flawlessly recognize any human utterance. (Or all phonemes in the IPA alphabet, to be more precise)

The "training" then only learns the mapping between IPA phonemes and text notation.

Re:Show HN: State-of-the-art German speech recognition in 28... by mciancia on Wednesday August 10, 2022 @06:15PM (#32419098)
What other languages can you "easily" do like that?
Re:Show HN: State-of-the-art German speech recognition in 28... by fxtentacle on Wednesday August 10, 2022 @06:23PM (#32419169)
I believe for English, German, Dutch, Spanish, French, Italian, Portuguese, Polish there's enough data in CommonVoice to get good results.
Re:Show HN: State-of-the-art German speech recognition in 28... by gogusrl on Thursday August 11, 2022 @01:05PM (#32428733)
Any chance for Romanian ?
Re:Show HN: State-of-the-art German speech recognition in 28... by fxtentacle on Thursday August 11, 2022 @01:39PM (#32429233)
For Romanian, I believe someone would first have to collect a large dataset of speech recordings together with groundtruth text. Even if you find cheap narrators working for $10 per hour, that's still $100k for 10k hours of data.