Four virtual bunnies were harmed in the making of this film
June 16, 2018 10:11 AM   Subscribe

 
Funny thing, I've watched these before but this is the first time it dawned on me that these new simulation methods aren't just for entertainment or image manipulation, the dress simulation creates the dress pattern on the fly too. Maybe this is meant for texture artists, but those patterns could just as easily be used to make real life custom tailor-fitted garments from a photograph. Likewise, being able to automatically calculate the optimal parting line for complex molded parts is hugely useful in manufacturing.
posted by Mr.Encyclopedia at 10:58 AM on June 16, 2018


It's all amazing but the fold-aware fabric sim, zippables and the wind-tunnel are just wow.
posted by bz at 10:59 AM on June 16, 2018 [1 favorite]


As usual, Ke-Sen Huang's site has the full list of papers, most of which are available as open-access PDFs. I enjoy skimming through them every year, even though almost all of the details are way over my head.

And yeah, it's really cool to see how the definition of "computer graphics" keeps expanding to fill new kinds of applications beyond just gaming. I think my favorite project that didn't get highlighted is "Looking to Listen at the Cocktail Party: A Speaker-Independent Audio-Visual Model for Speech Separation". It might seem like a gimmick at first glance, but it has the potential to drastically improve the quality of automatic video captioning and video chat.

"Automatic Machine Knitting of 3D Meshes" is also pretty neat, and includes a knitted bunny as a palate-cleanser after all the other bunny mutilations.
posted by teraflop at 11:53 AM on June 16, 2018 [3 favorites]


Pity little Stanford bunny-- the Job of the 3D graphics world. "Bunny that is born of a rabbit is of few days and full of trouble..."
posted by gwint at 12:02 PM on June 16, 2018 [1 favorite]


Before the weaponization of social media and the ubiquity of massive data hacks and ransomeware, the yearly SIGGRAPH Preview was always my go to "this is so amazing / but but this could also be terrible"
posted by gwint at 12:06 PM on June 16, 2018


That modeling of mirror and reflective surfaces while the camera is in motion was pretty cool. That’s going to liberate a lot of scene planning and blocking.
posted by Thorzdad at 12:08 PM on June 16, 2018


This year's was significantly less terrifying than last year's, which had a couple 'now we can synthesize speech and video of anyone* saying anything' entries.

* - (whom we have enough training data for.)
posted by kaibutsu at 12:25 PM on June 16, 2018


Oh don't worry, there's still plenty to be terrified about.
posted by teraflop at 12:32 PM on June 16, 2018 [6 favorites]


That’s a lot of labor being automated right there
posted by The Whelk at 12:57 PM on June 16, 2018




The Whelk: That’s a lot of labor being automated right there

If the history of computer-generated imagery is any indication, these advances will be turned into flaky software packages which require weeks of work by multiple sysadmins to get installed and configured properly at each studio where they're used, and then they'll create years of additional work for render wranglers trying to figure out what went wrong in each of thousands of batches of renders.

So they do create some labour, too.
posted by clawsoon at 1:30 PM on June 16, 2018 [5 favorites]


That’s a lot of labor being automated right there

Having worked in the gaming industry and seen what tools do for game artists and developers... I don't think so. It's just going to complicate the final result.

Think about how big the special FX and animation teams are on movies today vs in, say, 1980...
posted by Foosnark at 3:38 PM on June 16, 2018 [7 favorites]


Bunnies are OK, I suppose, but I'm old-school, and as such I was gratified to see at least one teapot.
posted by Nerd of the North at 9:17 PM on June 16, 2018 [3 favorites]


That bunny is fine. Just got inverted, not folded. No worries. Just like turning a sphere inside out.

If it were a two dimensional bunny outline, then that would be much worse. Bunny would have to get folded.
posted by nat at 10:58 PM on June 16, 2018


"Looking to Listen at the Cocktail Party: A Speaker-Independent Audio-Visual Model for Speech Separation" yt . It might seem like a gimmick at first glance, but it has the potential to drastically improve the quality of automatic video captioning and video chat.

This basically enhances automatic lip-reading; how many of these SIGGRAPH techniques aim squarely at bettering surveillance tech?

Brings to mind this other 2018 conference: Artistic Research Will Eat Itself.
posted by progosk at 2:34 AM on June 18, 2018


« Older Dress for Success - 2018 Edition   |   "Canned beans and ramen noodles night after night" Newer »


This thread has been archived and is closed to new comments