View Single Post
Old 14th December 2016, 09:41   #41481  |  Link
webs0r
Registered User
 
Join Date: Jun 2007
Posts: 68
My 3c on the UI (I stayed on the v0.91.1 version for now as I was scared of losing the fastest doubler (s-xbr for me)+control - so haven't seen the new).

There are 2 main use cases:
(1) Upscaling
And then the progression is:
(a) upscaling -> (c) downscaling -> (d) refinements
These a-c could be on tabs going horizontally.

For the doubling, I feel a system where there is a blank array initially and the user can add items in corresponding to doubling instructions would be easier to use.
Each entry: trigger scaling factor, algorithm, (any modifers)
A modifier could be the s-xbr sharpness and bloating, or NNEDI32 neurons or NGU quality level.
So 2 entries would give the user quadrupling.

Final sizing would be similar to the 0.91.1 image upscaling but yeah, it could just use drop downs instead. In fact, this could be a mandatory line on the array - with a final resize trigger, and for that one the user can choose from the much wider array of scalers and options.

For chroma, maybe duplicate that?
The interesting thing here is, should the first chroma item be the initial chroma doubling? And then the final resize - could be anything, making it independent to the luma... don't know if its a good thing.
Another option is to combine the two - if user specifies a luma doubler they have to choose whether to double chroma or not at the same time. But then we'll need a separate initial chroma doubling option.

(2) Downscaling
Just the "final" scalers.

This makes more sense as the downscaling tab in the upscaling section only applies when supersampling, and you can have a different downscaler for when final size is <1.0.
I'm not sure this is where people/madshi want it to go - because it preserves the customisability of the old system and just improves the "flow". If simplification is required then options would have to go.
And I don't mean options as in algos - I mean things like - the ability to tailor that you want NGU-high for the first double and then NGU-low for the quadruple step. For example, a user could specify they want doubling, then input a supersampling ratio (1.0 for no supersampling or 2.0 to go to double the screen res) then just choose 1 option like NGU-med and that is applied for all doubling steps that brings it to (or over) the 2.0 ratio.

Not sure what to do about presets and such - I like the simple scripting system. Maybe a link to an external guide will do rather than baking in presets.

Otherwise, somewhere there could be a Fastest, Balanced and Quality. Or 5 options if preferred. These would just pre-populate the above settings dialogs -> allowing scripts to still run. The tricky bit is, like in my situation, when I have a high frame rate video I'll drop to lower power scaling algorithms, but also change the dithering method etc. Gets tricky. Probably not worth it yet?

Sorry, I should have probably installed the latest ver to take a look. Hope there is something useful in my rambling.
webs0r is offline   Reply With Quote