Partial Render

[wayne_m]

Hi guys,

There is a render I am attempting to do. It covers 500 frames and is a simple scene of a curve object (drawn to create letters of a name) that has an ICE particle animation that traces the curve object. In the viewport, the animation works as expected but when I try to render it, there appear to be gaps in the curve object as if it is broken. This gap also shows up in the simple viewport test render. I am not sure what to make of this error. I have gone inside the "Render" > "Renderer Otions" > "Optimization" & clicked "Optimize" under "For Heavy Scenes" but it did not make a difference. Also, the strange thing is after rendering all 500 frames, I can see there is not a match between what is rendered in the viewport (using the region tool) & what is rendered in the pass. The pass has more than half of the animation missing from the rendered region. Are there any ideas about what might be wrong?

thanks

[sirdavid32]

you need to cache your particles first. Then render.
Also, your curve segments might be too distant for the particle to reach next UV segment.
Please attach the comparison pictures before/after as well as viewport/render to be able to help you. Thanks.

[wayne_m]

Okay so I made the adjustments as shown in "iceRender-bad04" and played the simulation to the end. (One small change, it is 1000 frames and not 500 frames)



Then for the render, I switched the setting to "Read Cache" hoping to see a difference (as Pierre suggested - thanks for the tip Pierre) but it doesn't appear to have helped.
As shown in "iceRender-bad01" the yellow text is the particle component I am trying to render. The 'CV curve' is a single curve.



"iceRender-bad02" is what I get when I use the render region tool for a quick preview.


"Default_Pass_Main.200" shows what is rendered at frame 200 which is less than what really should be rendered at that point. In other words, the particles rendered (Render > Render Current Pass) at frame 1000 does not match the particles that are visible in the render region area that I draw at the end of playing the simulation.

Even though this is my first attempt at using ICE, I doubt this behavior has to do with the ICE tree itself.

The particle simulation is a simple one - create particles that flow along the CV curve in 1000 frames. It behaves as expected in the 3d environment but does not render properly.

"Also, your curve segments might be too distant for the particle to reach next UV segment."
>>take a look at "iceRender-bad05", is that what you are talking about? Based on what is there, do you suggest I increase the amount of points on the curve?


thanks for your help

[sirdavid32]

Fantastic. Thank you for the captures.
2 things:
1. your sim needs more time to complete filling the frames, so you used 1000 and still get these "empty segments", correct? - You even mention particles at frame 1000 are not the expected particle numbers in the render than the ones you see in viewport. Here we have a situation with an old cache. Have you totally deleted your old cache folder? also: try a new cache folder name (new cache name and make it SHORTER. like: "sim" that's all.
2. Duplicate the "y" segment (it will inherit the ICE properties) and shape a "T" just in top of the original 1 segment-word "T". Clear caches, resim to 1000 frames. And SEE if now with the "y" segment duplicated as "T" you still have the problem. If YOU DO, this is definitely a problem with UV curve orientations. Also, please, attach the ICE sim tree to see what's computing there. This should not happen: On the contrary: on frame 1000 you should see a lot of particles cumulated on "N" and it's not happening. So I think it's definitely a segment problem with UV orientation coordinates in the ICE tree. Duplicate the "y" and we'll know.

[wayne_m]

"1. your sim needs more time to complete filling the frames, so you used 1000 and still get these "empty segments", correct?"
>>Actually no. When I run the ICE simulation, I watch the particles fill in from the beginning and run all the way to the end completely as I expect AND I also see what appears to be the particles accumulating at the end when they get to the leter 'N' (you spoke about this in your second point).
The screen capture "iceRender-bad01" that shows the yellow text is the actual complete particle simulation as I expect. So the simulation is okay, it is the rendering that is showing gaps - the rendering is not okay.

iceRender-bad01




"You even mention particles at frame 1000 are not the expected particle numbers in the render than the ones you see in viewport."
>>You are probably referring to to my sentence -->"Default_Pass_Main.200" shows what is rendered at frame 200 which is less than what really should be rendered at that point. In other words, the particles rendered (Render > Render Current Pass) at frame 1000 does not match the particles that are visible in the render region area that I draw at the end of playing the simulation."
What I was trying to say is... what you are seeing in that screenshot is what the renderer produces at frame 200 but the renderer should have produced more so it tells me the renderer 'forgot' to produce or technically is out of memory. That is why I thought the cache idea should really work but apparently it did not.

"Here we have a situation with an old cache. Have you totally deleted your old cache folder? also: try a new cache folder name (new cache name and make it SHORTER. like: "sim" that's all."
>>When I used your suggestion of creating a cache folder, I did not have a cache folder previously - I did not see one to delete.
If there is a default cache folder then I need to look for it & delete it. The cache folder I have now is a new folder I created in the project space.

"So I think it's definitely a segment problem with UV orientation coordinates in the ICE tree. Duplicate the "y" and we'll know."
>>I duplicated the entire curve and after some thought I decided to make an adjustment. Looking at "iceRender-bad07" you can see the yellow curve (particles) at the lower left side of the 'A' and you can also see a black curve. After I copied the original curve, I moved the cv points on the lower left side of the 'A' because those cv points appeared to be overlapping and in the path of the cv points on the lower right side of the 'A'. So the gap you are seeing in that screenshot is showing the adjustment (translate left) of the cv points.




Looking at screenshot "Default_Pass_Main.350" you can see the results of making that adjustment - the letter 'A' completely renders. Comparing that screenshot to the screenshot in my previous post, you can see that section of the letter 'A' has never rendered before however looking at "Default_Pass_Main.1000" you can see the particle render also did not move past that point, in fact, you can see it actually reversed - as if those frames were lost or forgotten.
For each time I did a new render, I ran/played the simulation to the end with a new cache folder I created.

Default_Pass_Main.350


Default_Pass_Main.1000







"Duplicate the "y" segment (it will inherit the ICE properties) and shape a "T" just in top of the original 1 segment-word "T". Clear caches, resim to 1000 frames. And SEE if now with the "y" segment duplicated as "T" you still have the problem. If YOU DO, this is definitely a problem with UV curve orientations"
>>I drew the curve in the "Front" viewport but not all at once. Do you think it would make a difference if I drew the curve (formed all the letters) all at once before closing the 'Draw curve Tool'?
If the particles in the simulation are colliding, will that affect the rendering?


"Also, please, attach the ICE sim tree to see what's computing there."
>>ICE Tree below




Again, thanks for your help

[sirdavid32]

Thank you for your details. This is good.
Ok here's what I'll do in your situation:
If you render letter "a" draw a curve with the letter and extra space curve like "_ _a_ _"
That way the caching and sim will precisely "fill" the middle area.
I'd recommend to do this on the letter you're seeing empty.
Copy paste ICE tree. Create new "_ _a_ _ " curve( just an example) and place it over.

I would also not rely on the 3d scene at all at this point, I'll deal with this also in post production.

[wayne_m]

Okay, so I think I understood what you (Pierre) were trying to get at and what I did was to make 5 copies of the single curve and then I would delete some points to make a part of the word. So at the end I had "hi" & "Ka" & "iT" & "ly" & "n" - 5 shorter curves. Next I would modify my ICE tree to be able to render two separate curves in the same single simulation. By adjusting the "Limit by Time Range" node I was able to chain the simulation using the "Use start frame" and the "Use End frame" values. I figured if I could do that for 2 curves in the same simulation, then I would just need to do the same for the remaining 3 curves and render the whole thing. I chained the simulation and it went according to what I thought however there was still a bottleneck and I still had gaps.

So I went back to a previous version of the scene and I realized I did not have this issue before I decided on the color of the particles when they are rendered. The default color is grey (as shown in the screenshot below).

iceRender-bad08


iceRender-bad09


As you can see, the area at the 'h' and at the 'n' are dark grey because that is where the particles would accumulate as expected. The entire curve rendered completely with no problems. So it was after the color change that this render issue appeared. I am now thinking I need to make adjustments to the material I used for the particle cloud. See below for the material I created for the particle cloud

iceRender-bad10.png


I suppose my particle cloud material is faulty somewhere. Any suggestions for a simpler/more effective material?


Thanks for your help again.

[sirdavid32]

Yeah, well, The original problem with particle sim, was killed.
Particle shading is another thing. Maybe someone else may help you as well. I am glad with your progress.

[wayne_m]

Agreed - we now know there was no issue with the particle simulation
I also learned about the potential issues you can have with particles colliding if the points on the cv curve are too close or they overlap (see photo below)

iceRender-bad07


Another tip for everyone is to be wary if you use a cache folder in your scene. You might find it takes a very long time to re-open a scene file if you use a cache folder. To avoid the delay in opening that scene file be sure to set the "File Mode" to "Do nothing" before saving & closing that scene file (see photo below)

iceRender-bad04


Thank you Pierre - I appreciate your help & thanks to the community.

I love XSI

[sirdavid32]

Hey, I got more things planned for Softimage this year (regarding video tutorials), so I am glad you do your progress on ICE as well.

[wayne_m]

Thank you Pierre, I will update this thread with my progress on the particle material in my project. I'm also looking forward to your tutorials