Profile cover photo
Profile photo
Josh Mandel
physician, programmer, and would-be pâtissier/poet
physician, programmer, and would-be pâtissier/poet
Josh's posts

Post has attachment
for a long time I've wanted speech recognition on my desktop that was as good as my android phone.

recently google's chrome browser added a hook for speech recognition in single-line text inputs like the google search bar. breaking my arm after falling off of my bike last night gave me the motivation to patch together a quick hack that allows any multi-line textarea to accept speech input as well.

it's an ugly 100 line userscript that relies on polling the document object model, but I wrote it one-handed and stopped as soon as it sort-of worked :)

of course this comes with the huge caveat that all speech runs through google's servers, so it may not be suitable for anything truly private. but since my most important use case is inside of gmail anyway, this it isn't a deal breaker for me.



1. In performing this exercise I came to realize that many of google's own sites don't use standard textareas, but instead implement their own pseudo-textareas based on javascripted divs, which has serious usability implications. For rich text entry fields there's no way around this behavior. But for plain text fields like the one I'm using right now, this is just bad behavior!

2. the speech recognition service offered through chrome seems inferior to the one offered through android. In particular, my brief experience has been that punctuation is poorly recognized or not recognized at all.
Wait while more posts are being loaded