|
“I’m giving it to you in this form now, so that you can give me feedback. There are decisions to be made which we should make together.” The source-code for the deformer can be found here, along with a video which I’ll embed below.
20 Responses to “Qarl’s parametric mesh deformer on display, feedback sought”Leave a ReplyCommenters are to be civil, courteous and respectful to others, insofar as it is possible to do so. Beyond that, you're not required to agree with the opinions expressed by me or by others. Think for yourselves! First time commenters will wind-up in the moderation queue and your comment won't appear right away. Ditto for anything that gets flagged by the anti-spam rules. Got a news tip or a press-release? Send it to news@taterunino.net. | Support us...or use Flattr  | ||||||||||||
Dwell On It is proudly powered by WordPress
Entries (RSS) and Comments (RSS).
© Tateru Nino. All Rights reserved.
Tateru Nino and Dwell On It are trademarks of Tateru Nino. Your local laws provide the conditions for your fair and reasonable use of these marks.
All other trademarks are properties of their respective owners.
Page optimized by WP Minify WordPress Plugin
Yesterday (and I nearly missed it), Karl Stiefvater put up a preliminary look at the crowd-funded parametric mesh deformer. “I should specify immediately that it’s not done,” said Stiefvater, “but the heavy-lifting part of it is.”
The video could have been planned a bit more, so we didn’t get the slightly puzzled search for Avatar Physics…
It’s impressive what is possible. But what is it doing to frame rates?
I noticed a couple of times when the body-shape changes led to odd-looking distortions. Well, they do that even for textures. But there’s an area of that jacket that was used where the edge of the front opening doesn’t hang straight. There’s a sideways movement, more or less parallel to the body surface. I’ve seen the same happen with textures, and I suppose it’s a side effect of the movements of the underlying figure-mesh.
I can’t see an easy way of fixing that.
I don’t see an easy way of fixing that either – but then, I’m also not Karl.
As for the frame-rates issue, I don’t believe the deformer has any impact on frame rates at all.
Wow, looks great so far!
As for the jacket edge moving sideways, I don’t think that’s an “issue”. Its just the way those particular vertices are weighed and influenced on that jacket.
Now, hopefully Linden Lab will hastily get the code out for testing. I believe there’s enough eyes on this that it’d be deserving of its own Project Viewer even. Ideally they won’t lag too far behind TPVs. Seems so far they’re playing pretty nice with Qarl so that’s great.
it’ll be fun seeing what it does with my clothes. I’ll try to find a test viewer once my head is clear of this cold virus i have.
Niran’s viewer has already published a version that includes this code. Downloading as I type! lol
Ezra, if 100 out of 40,000 concurrent logged in eyes are enough for the Lindens…i´d strongly suggest them to change their business into something like “How to please the geek crowd and ignore user demand”.
Well, Tateru, it might not have effects on fps, but considering the fact that Blue Mars never went to anywhere except sudden exit to desktop it might have some different side-effects, apart from oversized jackets remaining oversized jackets. Maybe the commercial SL mesh textile industry should get used to the standards of the RL textile industry instead – and offer some more regular sizes than the simple S/M/L scheme they try to make money with. But hey, this would be WORK -zomg!
@Vivienne Not sure what connection you’re seeing to Blue Mars here. BM used fixed shape/size bodies so that mesh clothing was a one-size-fits-all deal (because everyone was the same size).
I have been avoiding Mesh clothing, so now I need to find something to test with, and the search has been showing up some of the problems of the search engine. The word “mesh” gets attached to such things as fishnet stockings, and other less respectable garments.
Incidentally, if this works with Avatar Physics, there are going to be certain attachments which will take obvious advantage. Some people do want to be able to use 3D nipples in their activities.
Somebody is going to have their eye out over this…
Working fine in the Cool VL Viewer v1.26.3.0 (to be published later this week)…
Yes this puts Second Life well ahead of Blue Mars in a few ways when it comes to being able to be creative with mesh. There’s still a lot Second Life could use from the likes of Blue Mars like texture map channels other than color/diffuse, but I’m sure Second Life will continue improving in that direction.
Exciting times.
There obviously is nothing at all Second Life could “use” from the likes of “Blue Mars”, never was, and never will be.
[...] it looks like we will be getting a parametric mesh deformer in due course. i’m indebted to Tateru Nino’s blog for breaking the news, which is probably the most positive thing yet to emerge on the mesh [...]
@Wolf “The video could have been planned a bit more…”
Video work takes time. Time better spent on mesh. We, as the target audience, are willing to suck it up and take the hit?
In other matters: SL content is improved not via Linden Labs. They really do drink from our milkshake.
“Video work takes time. Time better spent on mesh.”
Of course, a person with the skills for the video work would probably not be a useful resource for work on mesh. The two areas involve specialities that are not easily bridged, so that work on one does not take time from another.
I have been saying that I would wait for the deformer to buy a new jacket, but today a bought new hair and spent half an hour trying to get it to fit just right. The deformer should make mesh hair (and perhaps even attached flex parts) FIT!
Good job on getting heavy part done but it’s not flawless yet, however.
Need flag or checkbox to disable the use of deformer on assigned (rigged) Mesh. Otherwise, won’t be very friendly for robotic-like avatars.
I also noticed (in other videos) some part of the shape sliders seem to affect in wrong area of the “assumed” part of the mesh. Like changing the size of your nose have affect on hands and feets.
We still need Avatar 2.0 support anyhow.
http://wiki.secondlife.com/wiki/Avatar_2.0
On video production…
It’s wasting everyone’s time if you’re trying to find a menu option, or a particular information display, as Karl was doing a couple of times. It’s worth outlining what you want to show people, and making some clear notes, so that you avoid that.
That isn’t fancy video production stuff. It doesn’t need any editing, or a fancy script.
@Wolf: “It’s wasting everyone’s time if you’re trying to find a menu option”
I think you should send Qarl an email. Really lay in to him. Tell him its not good enough. Demand compensation.
Re the video planning: detail his short comings, along with suggestions for improvements. Reiterate to Qarl, that he wastes everyones time.
*Apologies for taking up your time if you read this.
[...] was “our” Qarl in that BoingBoing post earlier today. Gee, he’s not busy enough deforming mesh for fashion-forward avatars, he has to create a tiny model of his house as a snowglobe holiday gift [...]
@Tateru – Actually you could customize Blue Mars skins and shapes, I still have a custom shape on sale there. There just were not all that many content creators (or players), so only a few got made.
To the people irrationally afraid of Mesh – Second Life has always been 100% triangulated mesh. Turn on “wireframe” mode in any viewer and you can see that the sky, terrain, avatars, and prims are made of mesh and always have been. That’s because graphics cards only do one thing very very fast: shade triangles. Shade means apply a texture and adjust the color to whatever lights are in the scene. Triangles because that is the simplest shape with an area. Graphics cards have multiple “shader pipelines” that do this job in parallel on different parts of the frame (mine has 216 of them), which is how they do it fast.
A mesh is just a set of triangles that share edges and form a more complicated shape, but your graphics card does not care where they came from. All it sees are triangles to color in. Some meshes like the avatar body and prims are pre-defined in the Viewer, others, like sculpts and the new custom mesh feature are user-defined. The three things that matter are how many triangles total there are to color, how complex the shading process is, and how fast your graphics card is at doing the job.
Sculpts default to 2048 triangles each, which is quite a lot for real-time graphics like Second Life. To the extent a custom mesh can look nice with fewer triangles than a sculpt, it will render faster. To the extent it uses multiple textures (you can have up to 8, vs 1 for a sculpt) it will render slower, because it takes time for your graphics card to load each texture to color with. So the net effect on total frames per second depends entirely on how well designed the items are.
The parametric deformer will have no effect on ongoing rendering speed because it only changes where the triangles of the attachment are located, not how many of them there are. It will have a one-time effect when an item is first worn/comes into view/appearance changes to calculate where the triangles are based on the avatar shape. After that’s done it should have no effect, so Qarl and the Lab just have to make sure it does that step efficiently.