read-aloud
read-aloud copied to clipboard
ken107
Azure Cognitive Services voices support
Support for Microsoft Azure Cognitive Services voices - ~~closes #168~~ . [WIP]
- [ ] implement support for different voices
- [ ] implement periodic refresh of voice list
- [ ] clean up
Questions:
- should I target firefox or master branch? (I branched off firefox since I use Firefox, but I'm sure what's the difference with master)
- I included the Speech SDK minified bundle, since it makes things easier, but I noticed that Amazon's has been removed in the firefox branch. Is there a preference for not including SDKs?
Great work, thank you. However, there's a problem with Firefox. If you check the FF branch, there's no aws-sdk.js. I had to remove it because Mozilla rejects minified third-party code. I'll have to prove where I get the code from, has to be an official source, and how to build it, etc. Basically they make it real difficult.
I'm thinking about creating a separate branch for this. This feature won't be used by most regular users. Only more advanced users, i.e. developers who know how to set up Azure account, will use it. In that case they can clone the branch and install manually. Also it will keep the size of the package small.
#168 actually has different intention, to call the web API on that demo page and provide these voices for free, to the extent Microsoft would allow it. Would have to try it out first, may not work.
I had to remove it because Mozilla rejects minified third-party code. I'll have to prove where I get the code from, has to be an official source, and how to build it, etc. Basically they make it real difficult.
That's unfortunate :/ Actually this article says they banned obfuscated code, but still accept third party libraries (minified or not). But there might be more scrutiny indeed. The extension policies also rule out loading remote code for execution, which would have solved the size issue.
I'm thinking about creating a separate branch for this.
That can work, but I'd recommend using a single branch and some feature flags that can include or exclude files/features when packaging, as it would be easier to maintain.
Anyway, I'll try to polish this and make it as independent as currently possible from the rest.
#168 actually has different intention, to call the web API on that demo page and provide these voices for free, to the extent Microsoft would allow it. Would have to try it out first, may not work.
Ack. I suppose it can work: there's a 5h tts free tier, which should be ok for a demo page (if using a subscription key for public demo is fine).
Yes, let me do a little more thinking, let's keep this open as work in progress