Where can the NST speech synthesis audio be downloaded now?
Hi, and thanks for publishing nb_samtale — we're using it for Norwegian TTS reference audio.
A question about a different resource: the NST speech synthesis dataset (oai-nb-no-sbr-15, the ~8 hour single-speaker recording). We'd like to use it to train an open Norwegian voice.
The metadata and documentation for the new release still serve fine:
https://www.nb.no/sbfil/talesyntese/nst_tts_dataset.jsonl
https://www.nb.no/sbfil/talesyntese/NST-tts-dataset-documentation.pdf
But the audio doesn't. The archive named in the documentation, sbfil/talesyntese/no.ibm.talesyntese.tar.gz, returns 404, and so do the Swedish and Danish equivalents at the same path — while sbfil/talegjenkjenning/ still downloads normally. So it looks like the talesyntese directory was cleared during the re-release rather than a single file going missing.
Is the audio available somewhere else, or is it expected back?
Also, if you happen to know: the Tuva Speech Database (sbr-44) would suit us well too, and its catalogue page 404s as well.
Thanks!
Glad to hear you find nb_samtale useful!
Unfortunately, we can no longer share the speech recordings from the NST speech synthesis dataset, because we no longer have the rights to share the voice data. So you're right, that data has been cleared. We are currently developing a new tts dataset to fill the gap. It will hopefully be published in a few months.
The Tuva Speech Database should be available in our resource catalogue at this path "nb.no/sprakbanken/ressurskatalog/oai-nb-no-sbr-44/", so hopefully the 404 was just a temporary lapse.
Cheers!
Thank you, and I look forward to the new dataset