search

NOtherDev / whatwebcando

An overview of the device integration HTML5 APIs
https://whatwebcando.today/
MIT License
836 stars 61 forks source link

Speech recognition is convenience not a capability #12

Closed petkaantonov closed 8 years ago

petkaantonov commented 8 years ago

If you can access microphone you can do speech recognition.

NOtherDev commented 8 years ago

Well if you mean reading the <audio> stream, then sure, technically you can run the recognition yourself. I've found a library that is doing exactly this - pocketsphinx.js. Are there any other solutions you're thinking of? Let me know please and I'll put a note in the Speech Recognition page.

petkaantonov commented 8 years ago

I am not sure of the specifics but I imagine having microphone API yet inability to actually access the input from microphone would be pretty unreasonable. I guess MediaStream can be used to access the input?

NOtherDev commented 8 years ago

Yes. Added an appropriate note in Speech Recognition. Thanks!

petkaantonov commented 8 years ago

I still wonder if it should be mentioned that doing it yourself directly has some advantages as you have far more control over the algorithm and limitations. For example webkitSpeechRecognition seems to have a maximum duration of 1 minute.