[Isaac Someah-kwaw]

Yeah it’s disabled. Ive also converted most of the project layers to image sequences, deleted unused effects and converted as many expressions to keyframes as I could. The original render time was cut down quite a bit, but still not yet to a point where it doesnt largely impede on my overall workflow

[Isaac Someah-kwaw]

I just wish I could get a complete result overnight, instead of waking up to see that it’s only halfway done.

Seems like the most effective thing I’ve figured out so far is using affinity to isolate an instance of BG render to each of my 8 cores, so they don’t wrestle with each other over processing power. Was able to squeeze out 7-9 frames per minute this way, as opposed to 4-5.

[Isaac Someah-kwaw]

After more observation, it looks I’ve fallen for a placebo effect of sorts. Turns out that the more instances of BG renderer I open, the longer it takes for each instance to render frames. Using 6 instances of BG renderer operates with not much better efficiency than using two. About 4-7 frames per minute…

[Isaac Someah-kwaw]

-So I’ve been experimenting with Michael and Chris’s suggestions.

-I tried multiprocessor rendering in CC 2014, but it was giving me even bigger initial estimates than CC 2015 and AME were.

– So then I went and re-rendered all of the PNG sequence layers in the project as uncompressed TIFFs. I’ll say off top that the huge file sizes make this an unrealistic option in the long run. I tested rendering with these settings, and I’m not sure if there was a noticeable improvement in rendering times.

– I also looked into disabling hyperthreading, but just my luck, it would appear that my laptop doesn’t allow for it to be disabled.

-I did stumble over a promising option in the BG renderer script. I purchased it and have been testing out rendering with multiple instances of it running. It certainly made the PNG to TIFF conversion much faster. I’m currently trying this option with the final Render to TIFF. I launched around 30 instances of BG rendererer earlier. So far, it IS rendering quicker, But….. It’s been 7 hours and it’s just now getting to the 1 minute mark in what is a 5 minute project. Half of the renderers I launched quit after about 2 hours because of errors, but I still have 14 running.

-The real problem is the time it takes to render one frame. It’s taking about 3 minutes per frame on average, and 4-7 frames per minute between the different instances. Is there another format I could use that would still make quality results, but at a much faster pace?

[Isaac Someah-kwaw]

From the looks of it, this would be a job better suited for Photoshop than After Effects, unless you intend to animate the face somehow. It shouldn’t be to terribly difficult using some masking.

I imagine you could just use the magic wand selection tool to select the black area, and then create an inverted mask that turns it into transparency. Then place another metal image similar in color and texture to the other metal face parts in a layer below the hole layer. Then, either mask or delete any extra metal area that would hang out from under the hole layer’s outer edges. That would get you part of the way there. And then you can do some color correction and clone stamp/healing brush work to make the hole layer better mesh with the skin tone and texture.

After you finish that, then you could import the final product into after effects and do anything else you had in mind.

[Isaac Someah-kwaw]

To be honest I’d rather not, at least not immediately. The music playing is actually self-made content that I will also be hosting on my soon-to-be-released website. Maybe it’s an unnecessary precaution, or just force of habit, but I’ve been keeping everything I’ve been working on pretty much confidential for the past…3 years lol. I won’t mind uploading it though, if any other recommendations don’t end up working for me

[Isaac Someah-kwaw]

Thanks, I’ll have to give that a try

[Isaac Someah-kwaw]

Yeah I’m sure that plays a colossal role in why the render is taking so long. There are 6 primary elements in the main composition that utilize a variation of one of Dan’s loop expressions that makes the elements hover in circles. Additionally, these elements all use Video Copilot’s Saber plugin. Each instance of Saber is controlled by an expression that causes their Glow intensity to interact with the audio in alternating increments. It operates in a similar fashion to how the position expression does. On top of that, each element is accompanied by a point light, that utilizes the same position and light intensity expressions, so that the environment around each glowing element appear to be affected by it’s glow.

I tried to prerender these elements to 30 second PNG sequence loops, so that I could at least eliminate the the position expressions. But the problem with that is, in order to use Sabre with the PNG sequence files, I’m forced to auto trace the entire duration of the work area for each Sequence file, as opposed to a single frame as I was doing prior. And from what I can tell, running Sabre in this way actually causes After effects to run even slower, though I never tested a final render using that method. On top of that, that still leaves the glow/lighting expressions running, as well as the position expressions for the point lights.

And to answer your last question, it’s not really volumetric lighting that is my aim, but rather what I was describing in the first paragraph. I just want the light being cast from each of the glowing elements to interact with their surrounding environment in a realistic fashion. And being that the position of each floating element has a z-position ranging anywhere from 300 to 5000, idk how I could mimic the lighting realistically using a prerendered video that doesn’t have any z-Depth

[Isaac Someah-kwaw]

– Yeah ray-traced 3D is disabled.

– I’m using AE CC 2015.2

I have an Acer Aspire V3 laptop:

-Intel(R) Core(TM) i7-4702MQ CPU @ 2.20GHz (8 CPUs), ~2.2GHz
-24gb ram
-Intel(R) HD Graphics 4600
-Windows 10
-214gb SSD for the OS and main programs
-2TB HDD for storage

I did just try rendering as a JPEG sequence, and the estimate did stay around the 8-9 hour range, albeit I only let it render for a minute or two. So if all else fails, I think I can make that work, despite the quality loss

[Isaac Someah-kwaw]

Wooo! After enough poking and prodding, I’ve figured out a solution that suits my needs.

The key was in creating an incremental loop that gets triggered by passing the threshold, and resets itself once its value is greater than 3. Each increment sets a different x value that gets multiplied by the audio output level. And the linear expression keeps the glow from completely disappearing when the output falls below the threshold.

Thanks Dan!

threshold = 38;
x = 0;
n = 0;
audioLev = thisComp.layer("Sound Keys").effect("Sound Keys")("Output 1");
frame = Math.round(time / thisComp.frameDuration);

above = false;
while (frame >= 0){
t = frame * thisComp.frameDuration;
if (above){
if (audioLev.valueAtTime(t) < threshold){
above = false;
};
}else if (audioLev.valueAtTime(t) >= threshold){
above = true;
n++;
if (n==1) {x = 1;
};
else if (n==2) {x = .75;
};
else if (n==3) {x = .5;
};
else if (n > 3) {n = 1; x = 1;
};
};
frame--
};

glow = audioLev*x
linear(glow,0,500,60,500);