[HN Gopher] OpenAI releases Whisper v3, new generation open sour...
       ___________________________________________________________________
        
       OpenAI releases Whisper v3, new generation open source ASR model
        
       Author : crakenzak
       Score  : 44 points
       Date   : 2023-11-06 18:50 UTC (2 hours ago)
        
 (HTM) web link (github.com)
 (TXT) w3m dump (github.com)
        
       | nshm wrote:
       | Good improvements for many languages, numbers here
       | 
       | https://github.com/openai/whisper/blob/main/language-breakdo...
        
         | WXLCKNO wrote:
         | Curious as to how dutch has the lowest error rate
        
           | crucialfelix wrote:
           | They enunciate.
        
         | tekacs wrote:
         | From the WER numbers alone it looks like a very small
         | difference for English itself, but I've found WER to be a
         | misleading assessment mechanism.
         | 
         | Having extensively tested Whisper v2 large against other 'lower
         | WER' models and found them wanting (because of differences in
         | their methodology for generating output), I'm super curious to
         | get a feel for how v3 holistically behaves.
         | 
         | Will probably test it right now. :)
        
         | Void_ wrote:
         | I don't understand how a pop. 10M country - Czech Republic is
         | among the best.
         | 
         | And I can confirm - my app Whisper Memos
         | (https://whispermemos.com) is very popular in Czech Republic.
         | 
         | It makes perfect sense. Whisper is almost as good as
         | transcribing Czech as English!
        
           | GaggiX wrote:
           | I don't know Czech, but Italian is extremely consistent in
           | the way it's written, so it's at the top of the list with
           | about one or two orders of magnitude less data.
        
           | godelski wrote:
           | I'm more impressed about Korean! I didn't even realize it was
           | that good in V2. But I've just seen a lot of systems perform
           | really poorly (judged by my Korean gf not me) and Korea is
           | only a country of 52M (between Spain and Italy).
           | 
           | A funny note, if Siri is set in Korean mode and reads your
           | texts that come in as English, they sound like a racist
           | imitation of a Korean accent. It is absolutely hilarious.
        
         | ComputerGuru wrote:
         | It looks like it's basically whisper-2 with extra training
         | against datasets for specific languages that brought incidental
         | improvements to the rest. Support for some of the languages is
         | still really bad (from real-world experience).
        
       | tomrod wrote:
       | Word from my GenAI contact is that this (or similar announcement)
       | replaces the need for RAG.
        
         | danielbln wrote:
         | An ASR model replaces RAG?!
        
           | tekacs wrote:
           | Pretty sure they're confusing it with the Retrieval feature
           | added to the Assistants API: https://platform.openai.com/docs
           | /assistants/tools/knowledge-...
        
       | alex_young wrote:
       | Still doesn't look like it can do real-time unfortunately.
       | 
       | Edit: I understand that you can use small samples and approximate
       | something like streaming, but the limitation here is you wind up
       | without context for the samples, increasing WER. It would be nice
       | if there was some streaming option.
        
         | thot_experiment wrote:
         | Hm? I've never had trouble getting faster than realtime perf
         | out of the old Whisper? What sort of hardware are you running?
        
           | manmal wrote:
           | Isn't the problem that you can't stream live audio data in?
        
             | tekacs wrote:
             | So... you have to operate in chunks, 30s at a time
             | generally, although you get reduced time-per-chunk if there
             | are zeroes in the chunk and there are variants.
             | 
             | The zeroes + faster model is how Gladia (mentioned in my
             | other comment here) achieves live transcription by simply
             | transcribing really short chunks one after the other, I
             | believe.
             | 
             | For more advanced stuff you kinda have to get your hands
             | dirty, which I've done for my own product (not linked).
        
               | modeless wrote:
               | Interesting teaser. I thought there must be some way to
               | better optimize the model for real time, but haven't dug
               | in because it's decently fast as is and there's so much
               | other stuff to work on. So many models, so little time!
        
         | tekacs wrote:
         | A few providers have done a variant of live transcription
         | that's similar to how old-school providers do it, where they
         | transcribe a short window (i.e. XXXms at a time) and this is
         | definitely the easiest path. One such provider is Gladia:
         | https://www.gladia.io/
         | 
         | There are other ways too with different trade-offs, can e-mail
         | me at the link in my profile if you'd like to talk about how.
        
         | modeless wrote:
         | I have Whisper working in real time here, with TTS too for a
         | real time voice AI that's _much_ faster than ChatGPT 's voice
         | mode: https://www.microsoft.com/store/apps/9NC624PBFGB7
         | 
         | It's inefficient, but even older gaming GPUs are fast enough
         | for real time performance, and accuracy is good. If you were
         | going to train a model from scratch for real time you could do
         | something more efficient, but it works as is.
         | 
         | Edit: I'm not sure what you mean by "you wind up without
         | context for the samples". You can supply context to Whisper.
        
       | dang wrote:
       | Related ongoing threads:
       | 
       |  _New models and developer products_ -
       | https://news.ycombinator.com/item?id=38166420
       | 
       |  _OpenAI DevDay, Opening Keynote Livestream [video]_ -
       | https://news.ycombinator.com/item?id=38165090
        
       | ComputerGuru wrote:
       | They say whisper-3 will be available via the api soon. Does
       | anyone know why only whisper-1 was ever made available via the
       | api (no whisper-2)?
        
         | Topfi wrote:
         | whisper-large-v2 was/is available via API from OpenAI, simply
         | wasn't as much of a focus in their recent communications.
        
         | Void_ wrote:
         | > Developers can now use our open-source Whisper large-v2 model
         | in the API with much faster and cost-effective results.
         | 
         | https://openai.com/blog/introducing-chatgpt-and-whisper-apis
        
         | bakkoting wrote:
         | The endpoint was just confusingly named:
         | 
         | > The Whisper v2-large model is currently available through our
         | API with the whisper-1 model name.
         | 
         | https://platform.openai.com/docs/models/whisper
        
       | singularity2001 wrote:
       | did they break the api?
       | 
       | from openai import OpenAI
       | 
       | Traceback (most recent call last): File "<stdin>", line 1, in
       | <module> ImportError: cannot import name 'OpenAI' from 'openai'
       | 
       | If so where is the current documentation?
        
       | jsight wrote:
       | This seems like the best free voice recognition in general.
       | 
       | Is there a model that is the best at wake word detection? The
       | last that I looked, it seemed like this was fairly lacking.
        
       | Nitrolo wrote:
       | Does anyone know of a nice UI wrapper for something like
       | whisper.cpp?
       | 
       | I need to write a lot of long texts for work and some good
       | dictation software would be great. I know there's Dragon, but
       | somehow I have not been able to find something that fits my need
       | and is free.
        
         | conradev wrote:
         | for macOS: https://goodsnooze.gumroad.com/l/macwhisper
        
       ___________________________________________________________________
       (page generated 2023-11-06 21:01 UTC)