An hour of audio transcribed in about 13s on MacBook Air

Re:An hour of audio transcribed in about 13s on MacBook Air by lukko on Tuesday October 13, 2026 @08:20AM (#50070520)

Interesting... is it overkill to use a multimodal model like this on just audio? Could the model size be decreased dramatically, or does the multimodality help the accuracy and speed in some way?

[ Parent | on HN ]

Re:An hour of audio transcribed in about 13s on MacBook Air by pjc50 on Tuesday October 13, 2026 @10:32AM (#50072398)

The benefit is that you can take input in one "mode" and extract output in another; that is, the input is audio mode and the output is text mode. The intermediate representation is the "embedding".

[ Parent | on HN ]

Re:An hour of audio transcribed in about 13s on MacBook Air by sd9 on Tuesday October 13, 2026 @08:45AM (#50070815)

Multimodality seems to just help with everything

[ Parent | on HN ]

Re:An hour of audio transcribed in about 13s on MacBook Air by coldtea on Tuesday October 13, 2026 @08:25AM (#50070572)

Correctly transcribing requires context too, no?

[ Parent | on HN ]

Re:An hour of audio transcribed in about 13s on MacBook Air by ranger_danger on Tuesday October 13, 2026 @09:13AM (#50071194)

I suppose it depends on the words in question... language knowledge and general world knowledge may not always both be necessary to transcribe accurately.

[ Parent | on HN ]