The problem with the Web Speech Recognition API is that for now it more or less requires either:
- including proprietary code (open source browsers are reluctant to do that (widevine is usually the only exception and it's opt in on Firefox and some Chromium forks)) (Google does this for other parts of Chrome at least, idk if it's used for WebSpeech or just live captions)
- interfacing with OS-provided proprietary code (doesn't exist on all platforms)
- streaming everything to a cloud server (likely prohibitively expensive unless you're Google) (Google does/did this too)
Safari not allowing requests to localhost from HTTPS is weird and I'm not sure why it's like that but it works fine in Firefox. Same for OffscreenCanvas which is the only canvas feature I've missed, Firefox finally added it 2 months ago.
Anyway, just don't block specific browsers, feature detect and if possible use a polyfill or ponyfill.
- including proprietary code (open source browsers are reluctant to do that (widevine is usually the only exception and it's opt in on Firefox and some Chromium forks)) (Google does this for other parts of Chrome at least, idk if it's used for WebSpeech or just live captions)
- interfacing with OS-provided proprietary code (doesn't exist on all platforms)
- streaming everything to a cloud server (likely prohibitively expensive unless you're Google) (Google does/did this too)
Safari not allowing requests to localhost from HTTPS is weird and I'm not sure why it's like that but it works fine in Firefox. Same for OffscreenCanvas which is the only canvas feature I've missed, Firefox finally added it 2 months ago.
Anyway, just don't block specific browsers, feature detect and if possible use a polyfill or ponyfill.