Fracture Modifier branch of Blender


#1

As one of the FM devs, I’ve begun to integrate Appleseed/blenderseed in to my Fracture Modifier workflow and will be promoting it along side FM and doing all my renders in Appleseed. I’ve got some demo files coming soon.

I just wanted to post this to get a thread started here. You can search Fracture Modifier documentation to find our blender wiki page with a list of resources to get jump started now.

I should have some demo files posted in the next couple days. If you have any questions you can ask me here or I and the rest of the team on our BA thread. It’s listed on the blender wiki docs page.

Fracture ON!


#2

To give a taste of things to come!
Here is the stylish Rendering Caustics in Blender with appleseed demo scene of Gottfried Hofmann @BlenderDiplom (https://t.co/B3e5I53Fte) put through @JTa’s powerful fracture build of Blender.


#3

Cool! Would love to see an animation of this.


#4

My laptop is not really beefy enough to render out a high quality animation but if someone else wants to give it a go the Blender file is attached. One can just move the time slider and see the break up unfolding.Simply pressing “Export animation” should do the trick then.
as_glass.caustics_fracture.blend.zip (6.1 MB)

Note, to use it the Blender Fracture Modifier build together with appleseed is necessary. Somehow in Windows it is not working with the latest appleseed version so I used Blender FM for linux.
Enjoy!


#5

I’ll render an anim out sometime in the next day or so. Hey Mango3, check out our support thread and please put some more info on they issue you had there. I’ll check it out on a windoze setup when I have some extra time to see what happens with Appleseed/blenderseed. Thanks for trying out the FM! Fracture ON!


#6

I am working on the animation and it looks like this run I only have time for 3 passes. It should be a good teaser. Original glass image by Mango3 was 20 passes. I’ll see if I can find time to do one at 10 passes @ 1280x720. So far things are looking good and I should have it posted in a few hours. It’s interesting to do an FM animation with SPPM.

UPDATE:
The frame times seem to be increasing more than expected as the shards fall away from each other and create more objects to bounce off of…so it doesn’t look like I will get more than a second of animation. I’ll have to make a mother pass at this when I have more render computer time…I might have something to post tomorrow.


#7

@Mango3 after looking closer I see you changed some of the settings for the lighting and rendering compared to GH’s original file. Since I am new at this type of lighting and rendering I was curious as to what your thoughts were in changing from his settings?

UPDATE:
Ah, looks like he just used the default settings and then raised the passes and lowered the samples. You adjusted more of the settings. I’m definitely interested if you experimented and eyeballed it or if you had specific special settings.


#8

Does anybody have any recommendations on their favorite settings for this scene?
Engine: SPPM
Light Sampler: CDF
Direct Lighting: SPPM Direct Lighting
with Caustics

Sampling:
Tile Size: 64/64
FIlter: Blackman-Harris @ 2.00
Pixel Sampler: Uniform
[Anti-aliasing] Passes: ???
[Anti-aliasing] Samples: ???
Which number combinations of the two do you like?
More passes or more samples?

Decorrelate Pixels: on
Denoising: off

Photon Tracing settings???
Radiance Estimation settings???

Thanks and Fracture ON!


#9

Good observation @JTa . I changed the scene slightly from Gottfried Hofmann’s unfractured original to put the emphasis on the fracturing. I increased the spot light power and changed camera position. Once it is more broken up in the animation the light is probably too intense and would have to be decreased.
Also, I increased the SPPM bounces and filter size. Denoising is off, which was less an issue for my still than it would be for an animation.

The Anti-alising Samples vs. Passes setting depends on your preferences. For the image quality the total number of samples matter, so samples/pass * passes, whichever way you split it. My laptop (Haswell Core i7) renders more efficient with a sample size/pass of >16 than with very low ones. The advantage of low samples/pass is that you get more of an immediate, progressive style feedback from the renderer, as each tile is rendered faster.


#10

First test animation done for the Hoffman/Mango3/JTa SPPM caustics demo.
1280x720/24fps/tiles 64/passes 10/samples 6
~4hrs 45min render time on i7/3+GHz/4 CPU cores
See the included file for more info

Save the file by taking off the blend extension and then unzip. Resulting folder is a blend file and mp4 video linked in the VSE. The render setup is for Fracture Modifier branch of blender and Appleseed/Blenderseed.

Appleseed-Hoffman-SPPM-demo.zip.blend @ pasteall.org

Below is frame #61. You can see a cluster of optical dots that may or may not be fireflies. Watch the video and you can see them move…

This is a first experiment to animate SPPM in Blenderseed. Comments and suggestions are welcome.

Fracture ON!


#11

Wow, thanks for rendering out the animation! The breakup of the pieces look really great.

  • One thing I would change is to increase the roughness settings for the inside glass material. The inside of the shards edges look to smooth which is unrealistic. This was not a problem for my still render where the edges were not fully visible but in the animation it is clearly the case.

  • Another thing would be to compare it with denoising on and see how well it works in that SPPM scene.


#12

Your settings look good.

Regarding passes and samples: for SPPM you will need to use many passes anyway (as many as you can afford) so it’s best to keep the number of samples per pass low, 2 or 3 should be enough.

Regarding photon tracing settings: the more photons you use, the less blurry each pass will be. I’d say at least 1 million photons but you can try higher value. The limit here will be RAM usage, mostly.

Regarding radiance estimation settings: the number of photons is a noise vs blurriness tradeoff. The initial radius should be set such that the first pass isn’t overly blurry. If it’s too blurry, decrease the radius. In theory SPPM converges to the correct solution given enough time, but the timeframe can be gigantic so it’s a good idea to make sure it starts reasonably sharp already by increase the number of photons shoot per pass and decreasing the initial radius.


appleseed 1.9.0-beta won't open any file
#13

I have literally no idea how to use Blender. How do I view the video?


#14

When you unzip the file it usually results in a folder. That should have 2 files. The blender blend file and the mp4 video file, which is linked externally. You can just play the mp4 file standalone or you can launch blender, load the blend file and the Video Sequence Editor/VSE should come right up with preview window and playback controls. So just click on the play button in the timeline window on the bottom of the screen and it will run the video.

There’s notes in that screen with additional information. Then you can go into Screens at the top and change the view to the Default screen and see the 3D View and shader settings.


#15

Excellent information, thanks for the feedback… @ Mango3 also.


#16

The 16p/8s version is done. It took 9 1/2 hours. I’ll post the animation later today. Here’s frame #61 again for comparision:

Next version I’ll use Franz and Mango3’s recomendations, different settings and more realistic fracturing.

This is not the best test scene but it’s helping me understand Appleseed’s SPPM. I’ll run some more reference renders and then post a blog article or something eventually to share the info. Thanks for everyone’s help.


#17

I skipped ahead to take some of the suggestions and it ended up in a longer render. So I’m waiting until that is done to post the most current animation.

As suggested I went to 32 passes w/3 samples, changed the inner/edge texture so it’s not so reflective and is more translucent as well as more rough.

Renders are taking about 9 minutes a frame so I will have the animation posted tomorrow, Sunday. It’s looking really good.

Here’s frame #30:


#18

Doesn’t proper DoF results interpretation come into play as one of SPPM’s benefits? I’ll have to check but I don’t think DoF is turned on in this sample file so we can see more of the signature features of SPPM.

I’m not sure since I don’t do much photo realistic work these days. I do more quick and dirty for live events. But I used it extensively in my early days so I have a workflow and understanding of basic concepts. These new concepts though seem to have very fragmented documentation and reference material aside from just reading white papers.

Can anyone point me to a good SPPM reference webpage? Tanks : )


#19

Second test upload of the animation done for the Hoffman/Mango3/JTa SPPM caustics demo.
1280x720, 24fps, tiles 64, passes 32, samples 3
~12hrs 30min render time on i7/3+GHz/4 CPU cores
See the included blend file for more info

Save the file by taking off the blend extension and then unzip. Resulting folder is a blend file and mp4 video linked in the VSE. You can use official release or FM branch of blender to view the video and notes in the blend file. The MP4 video is linked externally so you can view that separately if you like.
Appleseed-Blenderseed-Hoffman-SPPM-causitics-animation v2.zip.blend @ PasteAll

The render setup is for Fracture Modifier branch of blender and Appleseed/Blenderseed.
Here’s the render blend file:
Appleseed-Blenderseed-Hoffman-SPPM-causitics-animation v2- Fracture Modifier Source.blend @ PasteAll

I’m done with this test for the time being so comments and suggestions are welcome for an eventual v3.
Thanks for everyone’s input.


#20

There’s nothing special about DOF with SPPM. DOF works exactly as in path tracing.