New Appleseed release.

Hi All.

I am following the work of Franz and Jon on the render engine Appleseed.
About a week ago they released the newest version of the engine. 1.1.0 Alpha 17.
Do not be scared of the Alpha tag. It is a very capable app and very easy to use.

I feel that many might overlook this awesome piece of software even though the team behind actually have a great blog and are very good at telling about their animation project Masculine, made to test the engine.
It is kinda like Mitsuba. A pearl in open source rendering with a world of niche strengths compared to other projects but quite overlooked in the blender world for some wierd reason.

I have followed the blog from the very start and while this project is relatively young it is already showing much muscle especially for animations, including unbeaten motion blur, interactive rendering and studio style rendering.

And apart from Maya integration it integrates great with Blender through a simple export addon.

Interested in animations and havent tried this? Then go do it already! You may be surprised.

http://appleseedhq.net/

I haven’t really followed development of Appleseed in the last few months (more like a year) so I feel like I have missed out on quite a few things. First, I can’t really find Blenderseed anymore, did development on the exporter stop?

Appleseed is a great render engine and I would like to use it more but where can I find the exporter for Blender?

@NinthJake, the exporter actually comes bundled with the download, I believe, in the extras directory. Just copy it over like any other addon.

@ejnaren - yeah, I really like appleseed too, but the exporter is only an exporter at the moment, not any kind of integrated plugin (unless est777 has done some major work on that?). I think that’s why it’s gotten overlooked - the integration just isn’t (or at least wasn’t) there, it took too long to even set up textures correctly inside the appleseed GUI after exporting the scene.

Is it any better now? I’ll check out the new release anyway.

@^ No you are right some things are rather more difficult than should be when we talk integration with Blender. The guys behind it are very resourceful though and hopefully they are up to it anyway. I actually dont know if the pynodes project will benefit this or is there still being worked on the renderAPI of blender?

It looks like they’ve been making steady development, I almost forgot about this engine as well.

However, it does look like they could use a new server, the site is running quite slow compared to most other websites.

They also mention the possibility of bringing in OSL after they release the final version of 1.10, that could mean the possibility of easily moving shaders between it and Cycles.

But there osl shader dont work in blender they work with that render engine there osl shader are not working in blender
dont work.

brentison: Could you maybe rephrase that again please? :slight_smile: I am not sure what you mean.

OSL shaders should be pretty much transferable between applications.

They didnt compile on my side or didnt have the correct input and outputs.

Another photorealistic engine?

I wish at least one of these devs would stop working on these “physically based” rendererers for once and give us a nice real-time openGL engine.

Not everyone works with photorealism, but more essentially I think most people don’t have a renderfarm in their basement and would like to see immediate results, which with current graphics advancements are more than acceptable in a lot of cases.

I never get why everyone is so happy to wait 30 minutes to render a sphere.

highly exaggerated example.
Agree on realtime improvements, though. You can of course use blender internal with GLSL, but it’s limited and performance easily gets far from interactive in complex scenes.
My wishes on that:
-Cycles compatible real time preview
-Individual Matcaps for objects, available next to viewport colour in cycles. Gives greater sense of interaction when working on an unrendered scene
-better game engine display (There are problems with the licenses, making it impossible to make commercial games with BGE, if that’d be changed we could expect some improvements made by smaller game studios.)

Realistic/announced improvements in that area:
pixar opensubdiv: Awesome realtime display, check the last part of it’s demo
currently incompatible with GPL, Ton Roosendaal already asked someone at pixar to clear the licensing stuff
Opensubdiv inside of blender would be a really big step towards a good realtime engine.

btw. what kind of realtime engine do you want, if it’s not supposed to be Photorealistic?

you get the point, people doing 15 minute renders of a patch of grass and all that. Put a madcap on something and do an openGL render at 4000 by 4000 and you get it in a split second. I basically work exclusively in the viewport when I can now, anything else seems stupid.

essentially and up to date, modern video game engine, but in the viewport for easy rendering instead of confined to the BGE.

Look at any page in this thread on NEOGAF to get what I mean: http://www.neogaf.com/forum/showthread.php?t=506297&page=82

I simply mean it does’t have to be physically-based. Billboard grass is plenty enough for most cases.
Seriously though keep an eye on that thread and be amazed at the possibilities of properly optimised realtime engines.

The problem I have is that they have abandoned the internal engine and then dubbed it “obsolete” in favour of what is, in my opinion, a highly impractical one. So people now diss the old engine because they think it cannot be pushed further, when in reality all the rendering and compositing could be done in realtime, like it’s been done in games for years now.

What’s sad is that most of the pictures in there are horrible examples, which barely make the point. The future of CG (or one big aspect of it) is in GPU rendering for the most part alongside realtime (and near realtime) rendering. I also agree not everyone goes for photorealism.

If making game environments or low poly pieces of art is overlooked by many artist…then its always good to remind them that the video game industry brings in more annual revenue than film, tv and music industries combined for their primary market. In short, it is not wise to overlook the real time rendering arena, especially how it pertains to visualization and entertainment.

Here are some vids to really help sell your point:

the thread is almost 90 pages, I wouldn’t have posted the link if I thought it didn’t have good examples. The pics of modded Skyrim alone are worth it, and plenty of pics of Cryengine and such. Far Cry 3 also looks pretty good. Anyway people should give it a look.

Another good blog: http://deadendthrills.com

This guy does exclusively art-shots from games, with all settings maxed. Most of the stuff looks good to me :).

Thanks for posting those videos!

One thing you might be forgetting though is that no matter how high you crank the settings up for most of those games pictured in the thread, the meshwork is still designed to perform well on consoles which means the topology is usually very basic. The future of real time is essentially taking that basic boring mesh work and tessellating it.

Real time rendering doesnt need to look the way it does in most games if its for a beauty shot or cinematic…or smart smaller high density scenes.

Simpler topology aimed at performance is not necessarily a bad thing. A good model can be decimated quite a bit provided it has a good texture. The topology is more obvious to us in those shots because we are trained to see it, and also because they are shots from a work that is meant to be experienced in motion, and often at high speeds.

I personally am happy to have more conservative polycounts as long as I can work with the scene and navigate in real time with no stuttering. This is a must for me, which is also why I work exclusively in the openGL viewport render.

The point is simply that low poly count and realtime graphics are more than capable of telling a story, and should not in my opinion be discounted.

We would love to provide a solid integration of appleseed into Blender, with live rendering inside the viewports, shader graphs, etc. We actually made an important step in that direction with the recent introduction of Python bindings for appleseed. However we haven’t started work on an actual integration into Blender.

Unfortunately our resources are limited and our current focus in on supporting Maya, finishing Mescaline and bringing important new features to appleseed (OSL, hair, displacement). As usual, we welcome all forms of contribution and will do all we can to help anyone interested in improving the Blender+appleseed scenario.

Franz

I’d tell anyone looking to integrate Appleseed, or any external renderer, to wait for pyNodes to hit trunk. It will open all kinds of doors.

For what it’s worth, Appleseed isn’t a ‘new’ render engine, it’s been around a long time, you just hadn’t heard of it until now :wink:

Also, from the home page: “We are not building a strictly unbiased or strictly physically-based renderer”

Also, this thread is in danger of getting way off-topic. :slight_smile:

Quick heads up. New appleseed release, v. alpha-18:

http://appleseedhq.net/downloads

This is it if anyone wants to see it in action:


I can’t really speak for it as half the stuff was grayed out/not functioning when I used it