Re: [mediacapture-main] Add a mechanism to inform a video track about its type of content

I don't think we necessarily want to expose all encoder settings to 
users to optimize for this case, this is a hard thing to get right. 
That's why I do think that a content type hint makes sense. Saying 
"Firefox, this thing contains text, do the best you can with it for 
me!" seems like a very reasonable request to me.

What we want for the text-content case might be to expose the max QP 
and set it significantly lower. Specifying that "this is text content"
 doesn't require expertise in the field, but understanding QP as a 
concept and limiting it accurately requires significant expertise. 
Then being responsible for setting thresholds for all codecs and all 
future supported codecs is not something that I think a user should be
 responsible for (thresholds are different per codec for raw QP).

I think having browser vendors figuring out thresholds scales better 
than every application doing it on their own, because if you're not a 
person working on encoders they're hard to set thresholds for (that's 
why encoders have presets, right?).

-- 
GitHub Notification of comment by pbos
Please view or discuss this issue at 
https://github.com/w3c/mediacapture-main/issues/391#issuecomment-250783081
 using your GitHub account

Received on Friday, 30 September 2016 16:00:01 UTC