Edge Render Thread

Artwork Blender Tests
201

@Ace
What you proposed as alternative isn’t very convincing, you know.
There are a lot of tricks, involving matcap based solutions. Zbrush offers a lot of them.
In a much better and easier way.
There are more alternatives, all leading to a complicated use of the blender compositor.

I spent some time learning freeStyle.
It is really complicated and obviously useless for some fast presentations, exactly because it is complicated and occasionally unpredictable. It is also related to the BI render engine, which is also “poorly developed” and a real conflict to a cycles based blender UI. (another story, right?)

No Ace, this new announcement makes use of this “poorly developed” statement.
Poorly developed?
Not convinced.
Because too many serious parts of blender are poorly developed.
As multires and the whole sculpting, texturing, Vpaint, baking workaround is.
Noticed, I’m using this “workaround”
It is important to show what “poorly developed” means.

202

I thought I’d test out the developer’s suggestion of using Freestyle as a replacement for this technique. Here is my first draft attempt at creating a Line Set to emulate the Edge render.

Edge on the left and Freestyle on the right.

heart_freestyle.jpg1920×1040 537 KB


While this line set does approximate the effect, the time it takes to render is greatly increased. The Edge render took 37 seconds the Freestyle took 2:32 seconds.

Perhaps someone with better Freestyle skills can give us a better/faster recipe?

203

@Atom: Yeah, Freestyle is much slower, and I’ve managed to freeze it up and crash it a few times as well. It’s got many more options, but I don’t see me using it until a later date when the calculations are sped up.

@Ace: I agree with michalis, those nodes aren’t convincing to tell the truth, and they really make no sense. :slight_smile: I doubt they would hold up with some real lighting, you’d either have to take some nodes away, or add about 50 more XD. Nodeland is not where I want to be as an artist, I feel I am getting away from my actual picture that I had envisioned by hooking crap up for two hours instead of focusing on the art as I should be.

Anyway, sucks they are getting rid of Edge, but there’s not much we can do about it apparently. :frowning:

204

Here’s another example of an alternative method done through the compositor, the following node setup and .blend file simply makes use of the RGB information in the final image so that the effect is visible to raytracing.

OutlineEx.jpg1666×991 508 KB


OutlineExp.blend (687 KB)

Yes I know, it uses those icky nodes ;), but that’s why the file has it all set up for you. I also think this would be much better than the original attempt through materials since I do think that the old edge render itself was just an NPR post-process trick necessitated by the fact that Blender lacked a compositor until version 2.40.

Also, I haven’t tried, but this should also allow you to get your line information from an alternative version of the scene in another render layer where the only purpose is to create contrasting lighting to use as a map for the edges. For this anyway, it should be much faster than Freestyle even though I use 16x FSA here (use a lower value to speed up compositing).

205

I just do not understand this pressing need to clean up, when there are people who make use profitably, even in production, of these feature.

paolo

206

For some fast presentations, as an alternative, I suggest you to export a collada (selected objects only) and realtime render it in SketchUp free. LOL
If you have the Pro version, you can render the NPR as PDF (vector).
All these in a very fast and clean way.

Yeah, it’s about time for a good clean up.

207

? Wait they are/have removed edge rendering?! It was so much faster than freestyle! :frowning: Well it is fun to look back over the thread anyways. :wink:

208

I’m asking the same question! I would love to mess with this style some :confused:

209

@PhoenixSmith & Nickersf: Yeah it sucks, I agree. They should remove other things instead, like Cycles LMAO!

210

Or, if everyone defaults to pessimism every time a change is made, just forget about the whole 2.5x/6x stuff and use 2.49b for eternity (not like the foundation’s going to make changes to it. You can even get your own developers and make an expanded 2.49 under a new name (call it, Ye Olde Blender 3D with the tag Thou shall make art, creation with no simulation (meaning, not only no GI, but very few simulation options in general. physical plausibility = big no no) ;))

211

I would call this pessimism at all Ace, we are just sad to see it go because it is fun to use. Getting rid of Edge isn’t exactly progress, nor would keeping it be going backwards. I do like the sounds of “Ye Olde Blender” though…hmmmmmmmmmmmm :eyebrowlift2:

212

Well I tried for a few minutes to comment Ace Dragon’s post.
I deleted everything.
Didn’t find the right words.

The english language barrier.
A barrier anyway.

213

I’m asking the same question! I would love to mess with this style some :confused:

@Nickersf: Just keep an old version of 2.67 around to play with this feature.

214

@Atom
the last official blender is the 2.69 and edges are still there.
So keep a 2.69 official.
Ace’s suggestion is to keep a 2.49.
well…

215

My suggestion was for one to keep a 2.49 and fork Blender from that version, if they’re among the few who still believe that a lot of the stuff added in 2.5 onward is a bunch of professional mumbo-jumbo at the expense of the (legacy) Blender user (the new UI, Cycles, Bmesh, ect…), there have been a lot of threads here in the past where people tried to argue against the inclusion of the new UI, GI, and Ngons for instance. (and yet the inclusion of those ultimately increased the total net quality of the work being made in Blender).

216

@Ace: Who is arguing against global illumination?

217

Well, you stated before of your strong disdain for Cycles, Cycles does GI out of the box. Brecht tried to hook a GI engine onto the BI code when Sintel was being made, but it was more hassle than it was worth and what we ended up with was a 1-bounce solution with a number of limitations (which is what led to Cycles to begin with). So, you’d rather the devs. struggle with the BI code and see about cleaning it up, which might lead to more or less a new engine anyway because much of it would’ve needed to be rewritten (even if all of the legacy shading methods were still there)…

Heck, from my knowledge, there were at least four attempts to add GI to the old Internal Renderer going back to version 2.36, listed as follows…

-An old user known as Blackmage started writing a photon mapper and distributed rendering system for BI soon after Blender 2.36 was released, he got as far as material support, soft shadows, SSS, and Cache functionality before it was decided that the project was getting too big and thus was scrapped.

-Farsthary came onto the scene and started writing a photon-mapper for Blender 2.4x that was supposed to slot right into the BI code (that was after his work in prototyping a volumetric rendering system), this had many limitations such as not supporting textures in indirect light and ultimately didn’t make it far).

-Another user known as Matt-Ebb did some tests with Photon Mapping in BI, it would also not get very far and he would later abandon BI development in part due to the messy code.

-A GSoC student known as Uncle Ziev decided to try to add GI using an alternative method known as Lightcuts (that is basically the engine placing tens of thousands of basic lamps around the scene to give the impression of GI and make legacy specular shading resemble glossy reflections). He actually got pretty far on the project, but ultimately didn’t finish it and the project died.

-The render25 saga

-CURRENT: Sgraham’s GI implementation, seems fairly complete, but people are noting that it’s slow for simple scenes compared to Cycles (and this with no caustic bounces), its fate remains to be seen.

So I hope you’re starting to see the difficulty that multiple developers had when trying to add GI to the Internal Renderer, in terms of how it actually took the writing of a new path-based engine to bring it to reality in Blender.

218

My first (uploaded) Cycles render was August 2011, and I was really excited about it at the time, and truthfullly, still am (just need a bit of a faster PC, but that’s another story, and no fault of Brecht’s of course). I also understand the difficulties that people have apparently had trying to integrate it into BI, but I’m coming from C4D that has it in their regular engine. Sure, without being knowledgeable about code, I could (and did) easily say, “why not Blender?” Obviously it’s harder than it seems, and probably will never happen, and I have accepted that fact. I joke about Cycles a lot, but I use it more than most, I just wish the materials worked like BI in the Properties panel, just like every other engine in the universe. I know they do, but it’s a lot easier to use the nodes, although I am using the word “easily” quite loosely. I mean come on, in YafaRay for instance, I can add a color, bump, and spec, right on top of each other right in the Properties, and then adjust the mix, in no time at all. My problem is with the node-based materials, not Cycles, and certainly not GI, but hey that’s just my personal preference. I know you love playing with them, and that’s all fine and good, but I don’t.
My latest dining room, I spent more time test-rendering simple materials than I did setting up the entire scene. I make things often, and it just seems I spend more time on stuff that should be easy rather than concentrating on the picture itself, especially when I am lost in a spider web of nodes. If you ever make a complicated scene, you will see what I am talking about. You got a little beaten up in your SSS thread, and I felt for you, but you had a billion nodes that essentially did absolutely nothing, and that was just for the skin on your dragon, and not the eyes, tongue, teeth, etc. or any other part that you would have to construst nodes for. If this is the way of the future, I guess I’ll have to get more into them, but Cycles isn’t the only engine out there, and certainly not the fastest at the moment, far from it. Hope you see what I’m getting at here, and no offense is meant towards Brecht, Cycles, you, or anyone else, it’s just my preference as a highly active artist.

219

I actually dumped the entire attempt to do fake SSS when Brecht implemented the improved implementation for real SSS that we have now, and I have to say that the node group I have now that makes good use of it delivers much better results. Of course that was before we saw some of the new ray-length based group nodes that utilize a lot of math knowledge to deliver a very convincing effect.

Heck, my current avatar is a minimized version of this image here.

Aimage6.jpg1000×1000 270 KB

You can easily see the difference in basing the SSS shading off of the use of Brecht’s official implementation. Still took a bit of tweaking, but that was mainly in trying to find the right balance between scattering and glossiness as well as making limited use of add shader nodes to balance light energy.

220

I don’t see it, sorry. It might be your lighting, it’s extremely bright, but I also think you are making it more complicated than it is. It’s probably a combination of both. Judging from your nodes I have looked at, and not just for SSS, it seems you are putting the cart before the horse as it were. Start simple, GI on, one emitter plane, and an SSS shader, then go from there. I think these Add shaders aren’t doing you any good.

Anyway, haven’t tossed any Edge stuff in here for a while, so I just whipped this one up:

Edge_RIP.jpg1400×1400 1.56 MB