Random "work in progress" shots
category: general [glöplog]
You should definitely post that in the paintovers thread!
@visy: extremely cool! Not sure, what it is, but it looks very promising.
@pixtur: it's some voxel demostuff we've been working on lately, hopefully releasing at Payback in late Janjuary :)
Faking video compression artifacts
that is, simulating the youtube capture
Very nice. I wonder how THAT looks on youtube crap compression :D
no WIP anymore but this is the very professional concept draft of streams of gaia I found yesterday cleaning my messy desk :D
@Xetick: that is a lovely.
Xetick: ooh nice. Just a suggestion, it seems the "city lights" are not lit? I mean they're in the texture?
Xetick, I think what leGend is trying to say is that your image looks a little unnatural, because at the exposure levels neccesary to see stars, your city lights would be MUCH brighter... not some brownish yellow :-)
Beautiful! The only unnatural thing is the extreme luminosity of Nile river.
revival: thanks for the explanation :D
ham: That's just bioluminescence.
leGend: Correct, textures are from Nasa's blue marble project.
revival: Couldn't agree with you more. Going to see if I can't fix that somehow. Thanks for the reason for the nile river. Never question why since I though Nasa knows their shit. But now we know they do.
Fun facts. The scene is Oculus Rift compatible, texture size om disk is 1.8Mb and the stars/background is procedurally generated.
revival: Couldn't agree with you more. Going to see if I can't fix that somehow. Thanks for the reason for the nile river. Never question why since I though Nasa knows their shit. But now we know they do.
Fun facts. The scene is Oculus Rift compatible, texture size om disk is 1.8Mb and the stars/background is procedurally generated.
Quote:
Faking video compression artifacts
Are you really faking it ? Or do you use a compression algorithm ?
That's very interesting though
Buanderie: Completely faked :), based on feedback, perlin (simplex) noise, simple 2d sharpening, and a lot of fine-tuned parameters.
Wow...thats quite clever.
Noby: is it one quad or several?
good, another idea is to do the 8x8 dct, then remove random components/ quantize, then do the inverse. This will look a lot like bad jpg.
blur out the 8x8 coeff block towards the lower right, gets right rid off those nasty details :)
gargaj: One quad, render-to-texture feedback where on each pass you do a composite of the current frame, previous frame with distorted coordinates (blockiness) and previous frame sharpened (artifacts).
navis: Yes i've been thinking about that too :-), could be made even better by separating the chrome and luminance similar to composite video and mangling each of those separately.
navis: Yes i've been thinking about that too :-), could be made even better by separating the chrome and luminance similar to composite video and mangling each of those separately.
revival: or simply a lot of nocturnal human activity! :D
@noby: if you want to do all that true to form you'll end up splitting luma to a half size buffer and chroma (uv) to quarters, then dct'ing that, then fucking up the coeffs with a filter of sorts, decompressing (dct->yuv->rgb) and recompositing *and then* do some sort of blend with your current result
for mpeg2
all in all that seems like a lot of work and potentially worse looking than just taking a shortcut here and there
(luma/chroma separation for filtering in itself is fun though)
for mpeg2
all in all that seems like a lot of work and potentially worse looking than just taking a shortcut here and there
(luma/chroma separation for filtering in itself is fun though)