PDA

View Full Version : 2001 Slit Scan scene in Lightwave almost done yet ...



DonJMyers
07-09-2018, 01:14 PM
This morning I decided to fake in 3d the camera rig to create a 2001 a space odyseey slit scan type effect. I even found a guy who wrote a video program to reverse engineer the original artwork of the shots. He fed the movie into the algorhthm and it brough the images back: http://seriss.com/people/erco/2001/



Unfortunately that process made resulting images my blurrier than in the movie when you use this included image in your renders.

SCENE ATTACHED ALMOST DONE
A 100%+ luminous art card and a black matte card with an offset slit in it sit before the camera. The camera moves within the frame on the Z axis then resets to next frame. The motion blur is set to insanity and just about everything else is turned off. The artwork slowly moves left or right through the slot.

TRY
Changing directions of art
Fooling with blur
Different images
Different sized slits

Have fun playing with this simple LW scene which approximates that look of 2001 but something is not flat enough about it. Essentially, the camera uses motion blur and artwork to sort of mimic the look of volumetrics. My scene is still to volumetric. Three textures are included.

Can anybody make it look more like 2001? I wanted to upload more attachments re: this scene but for some reason my account is limited.

stiff paper
07-09-2018, 03:30 PM
Can't offer immediate help. I only have LW installed on an old box now.

However. Many years ago, a peripheral acquaintance asked me if I would do free VFX for a fan film he was going to make. Stupidly, I said yes.

Luckily, he never made the fan film, and the furthest I got was some experimentation on a title sequence.

I don't remember exactly what I did, but I do remember that I started out trying to properly replicate slitscan. Then at some point, I decided I was being stupid because -duh- I had a computer and LW. What I did then was analyse what was happening physically with real slitscan and then I faked the results of that. I don't even remember how it works.

I doubt it'll help, but hey, it might. Even if only for thinking about another angle on it.

(I've re-saved the images as jpegs and pngs to save space. You'll need to re-point them.)

Edit: Wait. I think I do remember something. Slitscan is -effectively- like projecting your moving image at a slight horizontal diagonal through a geometry tube. I think I used a null as a texture reference to point the image on a slight diagonal and then just made a variety of tubes shaped like the slit would have been. I remember hurting my brain thinking about it at the time, but the image slides at the bottom, and the new bit of image that's revealed is projected further down an illusory "tube" that's made by the camera moving in the physical version.

I might be wrong. It was waaay back.

MonroePoteet
07-09-2018, 03:56 PM
Here's a video of how the slit scanning was done in 2001: A Space Odyssey:

https://youtu.be/6EKreQ5HD4w?t=595

The camera has the shutter held open on a single frame while the camera moves completely (14' down to 1") toward the slit and the artwork is moved behind the slit. Because the shutter is held open, the artwork scrolled behind the slit ends up being streams of light on the single frame. Repeat many, Many, MANY times!

The specific description of how they worked starts at about 10:20.

The rest of that series (Parts 1 thorugh 7) are superb for any 2001:ASO fans out there.

mTp

ccclarke
07-09-2018, 07:12 PM
142167

Maybe this will help.

CC

jeric_synergy
07-09-2018, 10:51 PM
Thinking about the slitscan process hurtz muh bwain too. I don't know how one would emulate the accumulation of light on the film, since film has a "memory" and in general LW does not.*

When they were doing this, the camera reset after each push: did the artwork also reset??? (ouch ouch ouch....) It seems like that would be necessary to get continuity....

I do remember somebody (Mike Green?) came up with a way of using every point in space as a camera-- he was using it to project Lego-character faces onto, well, Lego people. But it seems like it should be able to do many many weird things.


*we used to do funky things like use a negative offset into the rendered image sequence.... another example of using funky numbers in LW's rather liberal numeric fields to advantage.

stiff paper
07-10-2018, 02:26 AM
I don't know how one would emulate the accumulation of light on the film, since film has a "memory" and in general LW does not.
Yeah, I think this was what caused me to stop and think about what the end result really is. The camera movement causes the shape of the slit at the "bottom" to, in effect, draw out an accurate (but illusory) 3d tunnel. If you have a circular slit, you get a normal tubular tunnel; a rectangle produces a rectangular tunnel; a single straight line produces a plane.

The camera movement also causes whatever image is on the other side of the slit to be drawn onto the tunnel's walls. After making a whole frame, the image underneath the slit returns to its original position but then increments the origin slightly each time. That's why the tunnel "moves".

In 3d, you'd be crazy not to make tunnel walls out of geometry, but that's only if you're interested in the end result and not the process.

If you want the same look and you want it to be a real simulation of the physical process, I think somebody would probably have to write a bit of code to replicate the whole "How exposure works with film" aspect of it.

Maybe not though. LW is (or at least was) very flexible and maybe some mad person somewhere could think of a way.

vncnt
07-10-2018, 02:36 AM
Thinking about the slitscan process hurtz muh bwain too. I don't know how one would emulate the accumulation of light on the film, since film has a "memory" and in general LW does not.
You could try Motion Blur to get that "memory" effect.

stiff paper
07-10-2018, 02:52 AM
Gah.

I'm now actually looking at the scene I uploaded and everything I've said so far has been wrong. At some point I must have decided that projecting textures onto the walls at angles was a foolish idea. The tunnel walls have simple UVs and the textures are applied absolutely straightforwardly.

Sorry. I do remember playing around with the projected textures, but it was a long time ago.

So there you go: if you just want the look then all you have to do is make some UV'd tunnel walls. Easy.

stiff paper
07-10-2018, 03:21 AM
I haz thinked.

Either...

Make your slit be a model of just the slit itself and nothing else. Luminous and with the texture mapped to move through it.

or...

Fix up all the "unseen by" settings for your black card somehow so all that shows up in the render is the slit and nothing else - definitely nothing black and in the alpha, for instance (I don't know if this is even possible).

Then...

Make your camera motion be across, say, 100 frames from top to bottom (100 frames for a completed single image). Each single frame should be an image of just the slit in a particular position along the "tunnel". Also make your projected image's motion be across the same number of frames. Turn off motion blur for now.

Set up your frame saving based on Mr Rid's suggestion near the end of this thread:
https://forums.newtek.com/showthread.php?120777-footprints-in-snow&highlight=footprints

Hopefully, what you'll get is a single frame at the very end of your 100 frames of render that has the slit "exposed" on it 100 times, thus painting the tunnel walls.

This will give you a single frame. Renumber this frame to frame 0000 and put it in a safe place. Delete all of the other frames; they're chaff.

Increment the start (and end) position of your projected image very slighty, render another 100 frames and do the whole thing again to make your second frame.

You'll have to experiment with the distance your camera travels and the number of frames needed to make a single complete image so that you get a continuous tunnel wall rather than occasional "slices".

This is a time consuming, laborious and mentally oppressive process. Thus, it's the perfect analogue for the real thing.

No idea if it'll work. But if it's the physical process you're trying to replicate, I think this might be as close as you can get.

jeric_synergy
07-10-2018, 09:17 AM
I haz thinked.

.......
This is a time consuming, laborious and mentally oppressive process. Thus, it's the perfect analogue for the real thing.


LOL. How the heck did John Whitney Sr. come up with this idea in the first place? It's a hell of a long jump from "taillights on the freeway" to "2001's stargate sequence".

For me, the deal breaker always was..... "and now we unload the camera and wait 3 days to see if we got anything."

DonJMyers
07-10-2018, 09:25 AM
Your idea uses a triangular tube which is technically not slit scan. But thanks for trying!

DonJMyers
07-10-2018, 09:27 AM
I haz thinked.


No idea if it'll work. But if it's the physical process you're trying to replicate, I think this might be as close as you can get.

Did you see my enclosed scene? Most of you are tellling me to do what I already did which is a bit too volumetric-like. Perhaps my slit is too wide.

The main thing is "How do you leave the shutter open over an entire frame. I am playing with fractional frames and different motion blurs but the effect is too soft.

DonJMyers
07-10-2018, 09:32 AM
142168

I forgot that I made this scene in LW 2018 and some folks can't load it, but the attached photo above shows the very simple setup. The trick seems to be the movement of the camera within ONE frame.

jeric_synergy
07-10-2018, 10:24 AM
::duped::

MonroePoteet
07-10-2018, 10:38 AM
As Cardboard said in an earlier post, it depends on whether the objective is the effect or the process.

The slit scan process was intended to *simulate* an endless plane (or two parallel planes or a corridor) with amazing geometric patterns on it with the camera zooming across (or between) them. If the objective is the effect, then in a 3D package the only tricky / challenging part (IMO) is finding the high-resolution geometric patterns to move across the plane(s). Rather than moving the Camera across the plane, I'd move the textures on the surface (i.e. animate in the Position tab of the Image to move negative in Z) so the edge of the plane never approaches the camera.

If the objective is to actually simulate the slit scan *process*, then I think you'll have to fall back to LW2015 or prior, but the technique you posted appears to work fine. Sadly, the non-realistic motion blur options appear to have been removed in LW2018, so your ONLY choice is "realistic" motion blur which will produce a lot of "noise" for the motion blur rather than the Classic or Dithered options, which produce individual "instances" of the blurred object / point.

To make a good quality fly through, you'll need to use much higher resolution and higher contrast images. As well, you'll need to make the slit as thin as possible. LW2015.3 sample scene attached. In the sample scene, the slit is 2mm wide and the Camera moves between -10m and -.1m between each pair of frames, using the same repeating Envelope as in the original post. The TCB spline is set to 1.0 Tension to ensure the Camera doesn't overshoot the slit.

As you found in the original post, the Motion Blur Passes needs to be set very high to get a "continuous" line. The sample scene has it set to 100 MB passes (i.e. very low) just to render the sample MOV file. 1000 MB passes is starting to look more like continous lines at a rendering cost.

142171 Frame 3 with 100 MB Passes, about 21 seconds
142172 Frame 3 with 1000 MB Passes, about 3 min 25 sec
142173 Frame 190 with 100 MB Passes

MOV file at 100 MB Passes 142170

The image used for the surface pattern (in the Images folder) was constructed in a few minutes in an Image Editor putting in rectangles, ovals and using a Tile and Swirl effect to convolute it. The Surface Luminosity is set to 100,000% (yes, one-hundred-thousand percent!) and the Color Saturation on the Processing tab (CTRL-F8) is set to 150% to overcome the loss in intensity from the motion blur, both of which could probably even be bumped up more. Glow is applied to the surface and Glow is enabled on the Processing tab at 300% with a 32-pixel radius.

To make the images for the surface, you might be able use the Kaleidoscope technique in this post:

https://forums.newtek.com/showthread.php?154983-filigree-Patterns&p=1521973&viewfull=1#post1521973

Have fun!
mTp

stiff paper
07-10-2018, 11:25 AM
"and now we unload the camera and wait 3 days to see if we got anything."

There's a great story from 2001 where the guy doing it spent forever on a sequence and then on frame 800 or something lost track of whether he'd already given the projected image handle an incremental turn or not on this frame... and had to start again.

Ma3rk
07-10-2018, 12:21 PM
LOL. How the heck did John Whitney Sr. come up with this idea in the first place? It's a hell of a long jump from "taillights on the freeway" to "2001's stargate sequence".

For me, the deal breaker always was..... "and now we unload the camera and wait 3 days to see if we got anything."

Well, I could tells ya, but then I'd have ta kills ya!

Heck, just 'cuz I share a last name doesn't mean I'm related. Always wanted to meet him though. Actually had someone track me down e-mail wise, wanting to ask me some questions, thinking I was his son; about as close as I got. I will assume though that pharmacopia was involved on some level.

Reminded me though of the famous story of the final dolly-in shot for Fantasia. Disney had already committed to the premiere date which wa days away. They had this final shot which was going to take them to within hours of competion, processing, etc. & getting it on a plane to NY. Over the weekend during the shot, SoCal had an earthquake. There wa no way to stop, check, etc. and make the deadline so they just went with it. And they got lucky.

DonJMyers
07-10-2018, 01:52 PM
The image used for the surface pattern (in the Images folder) was constructed in a few minutes in an Image Editor putting in rectangles, ovals and using a Tile and Swirl effect to convolute it. The Surface Luminosity is set to 100,000% (yes, one-hundred-thousand percent!) and the Color Saturation on the Processing tab (CTRL-F8) is set to 150% to overcome the loss in intensity from the motion blur, both of which could probably even be bumped up more. Glow is applied to the surface and Glow is enabled on the Processing tab at 300% with a 32-pixel radius.

mTp


Yes, aren't the settings absolutely NUTS for a standard render? Motion blur out the wazoo and luminosity ramped up to compensate. Ordinarily luminosity over 1,000 is absurd but here you need 100 times even that!

DonJMyers
07-10-2018, 01:54 PM
There's a great story from 2001 where the guy doing it spent forever on a sequence and then on frame 800 or something lost track of whether he'd already given the projected image handle an incremental turn or not on this frame... and had to start again.

That happened with the last shot of Fantasia also, an incredibly long dolly through panes of glass. During the shot there was an earthquake and the glass all moved. In the end fantasia was premiered in NYC with the last reel flown in at the very last second and placed in the projector still wet from developer!

DonJMyers
07-10-2018, 02:26 PM
https://youtu.be/Rk1aQx9hTaE

Another great use of this was the Superman The Movie credits whose analogue FX by R/Greenberg had to have cost a fortune! I also duped this effect in lightwave using flat letters and motion blur. The effect in LW is a little too clean though. And, if you look closely, there is a lot going on besides slit scan. Note how the letters bend and twist as they go through their motions. Unfortunately the final letters don't line up with the slit scan effect. Ooops!

This looks like early CGI but it is not. It is analogue, which is why my LW version is not EXACTLY as soft edged and nice as this. LW is too exact. I can post that scene if folks are interested in playing with it.

The NYT review of the movie said the opening credits had all the amazing power of a "commercial for a new brand of mouthwash." Too funny, and they are right it does go on and on with spectacle.

But so what, I was 12 and permanently imprinted!

DonJMyers
07-10-2018, 02:39 PM
https://youtu.be/Rk1aQx9hTaE

Here's a great example of a similar technique I previously recreated in LW. But my version ended up slightly too crisp and clean compared to theirs. The opening credits of Superman The Movie look like CGI but they are analogue. And, unfortunately, the swooshed letters don't link with the final letters. Ooops!

Nevertheless I was 12 and permanently imprinted!

I managed to fake the FX well and can upload this scene if you are interested.

But I was too lazy to test a text morph they do in the movie. You see, another problem is that, by 1977 when this was recorded, slit scan FX were more advanced. Note how the text does sophisticated warps as it moves. Not just straight on the Z.

Very sophisticated! I don't have time to mimic the look THAT closely.

jeric_synergy
07-11-2018, 12:00 AM
Looking into Whitney, that guy was a freekin' tech Wizard. Analog computers!? That way lies madness.

Trumble's rig, moving ~14 ft in ~1sec? Yikes. Stand clear. I think w/modern ballistic motor ramping they could have made their clutch and brakes last longer.

+++++++++

One thing I haven't grokked: was the background moving DURING a single frame exposure, or just the camera? It seems like the background was only "incremented" after each frame exposure, ie, not moving during exposure.

Why not just use hi-con b/w patterns, and color them in the printing process?

Fun stuff. I rather like what mPt has managed to do even with the breaks in the lines.

Ma3rk
07-11-2018, 12:42 AM
Looking into Whitney, that guy was a freekin' tech Wizard. Analog computers!? That way lies madness.

Trumble's rig, moving ~14 ft in ~1sec? Yikes. Stand clear. I think w/modern ballistic motor ramping they could have made their clutch and brakes last longer.

+++++++++
Here's a custom rig for flying stunt people a co-ordinator friend brought in one day on Lucifer:

142180

I think he said 2+ G's. We fed it a couple hundred amps of 240V, so ya it better move something rapidly.



One thing I haven't grokked: was the background moving DURING a single frame exposure, or just the camera? It seems like the background was only "incremented" after each frame exposure, ie, not moving during exposure.

Why not just use hi-con b/w patterns, and color them in the printing process?

Fun stuff. I rather like what mPt has managed to do even with the breaks in the lines.


Well, that's how they got the stretched blur for one thing. And they probably tried that too but liked the look of this better is all.

I'll have to go back & take a closer look at the progress a bit later.

stiff paper
07-11-2018, 02:49 AM
...was the background moving DURING a single frame exposure, or just the camera? It seems like the background was only "incremented" after each frame exposure

Both things had to happen. If the background didn't travel during the move, then the slot would just be a thin static image, and all that would be written onto the tube walls would be long stripes of whatever was in the slit at that point.

jwiede
07-11-2018, 11:28 PM
Analog computers!? That way lies madness.

Then I'd strongly recommend against digging too deep into how machine learning algorithms really work. :devil:

stiff paper
07-12-2018, 05:59 AM
... [Moderated content]

As you didn't specify why you were doing it, I gave you one way to achieve a perfect end result and another way to replicate multiple exposures along the "tunnel" wall. Why? Because your idea obviously can't work in LW2018.

And with that, I think I'm done even checking the forum to see if I can offer advice.

----

Mr Bowie! Sorry to have done a bryphi, but when you delete this, could you also inactivate or delete my account?

[Moderator's note: Adieu ...]

jeric_synergy
07-12-2018, 09:20 AM
Please don't abandon us, Cardboard: I always enjoy your posts.

The new 2018 Motion Blur, while it may be "better" in some respects (I can't understand) is disappointing for this particular application. It seems like users would be forced at minimum to do some what-I-consider fancy compositing using motion vector blurs and whatnot.

shrox
07-12-2018, 08:10 PM
Most excellent.

jbrookes
07-16-2018, 01:19 PM
If I recall correctly, the warp trails in Star Trek The Motion Picture were also generated using Trumbull's slit scan system (those too were describes as very time-consuming).

So if anyone figures out warp trails in LightWave, please let me know!

As a bit of trivia, during the V-Ger cloud scene early in the film, the background freezes for a few seconds since someone forgot to move it for a number of frames. You can see the pause in the background at 0:45 here:
https://www.youtube.com/watch?v=kbG3N51MEjM

I salute you for tacking the slit scan process. Very exciting! Sadly I don't have any useful LW techniques to contribute (except an image in my head of a LightWave camera's motion looped using an envelope to have it repeatedly approach a plane... whatever good that would do).

Interesting hearing about the limitations of LW2018's motion blur. Again, half the fun in LightWave over the years has been using trickery to create crazy effects. It would be a shame to lose that flexibility.

jeric_synergy
07-16-2018, 10:28 PM
(except an image in my head of a LightWave camera's motion looped using an envelope to have it repeatedly approach a plane... whatever good that would do).
It's simpler than that: 2 linear keyframes on the Z, one frame apart, set on repeat.

Quite disappointing that the 2018 MBlur won't accommodate this effect.

MonroePoteet
07-17-2018, 07:59 AM
Yes, as jeric_synergy says, a repeated "dolly" move on frames 0-1 to rapidly approach the slit then reset simulates slit-scan as was physically implemented in 2001:ASO. If you check out my previous post:

https://forums.newtek.com/showthread.php?157552-2001-Slit-Scan-scene-in-Lightwave-almost-done-yet&p=1550446&viewfull=1#post1550446

it's a few fairly minor modifications of the original scene file posted by DonJMyers, but using Classic motion blur in LW2015. In that scene, the Camera is at 10m away from the slit in frame 0 and 0.2m away at frame 1, with a TCB 1.0 Tension spline and Post Behavior of Repeat. As indicated, you can approach the smooth, "open shutter" slit-scan effect using more Motion Blur Passes in the Camera properties screen at the cost of render time. At some point, the individual "instances" of the slit created by Classic motion blur "should" be basically continuous.

For me, I'm still confused as to what effects wouldn't be much easier in LW using surfaces, volumetrics, lighting, etc. Slit-scan was intended to *simulate* what we can do easily in LW - i.e. an endless flat plan with high-resolution geometric light patterns on it over which the Camera flies. As I said in that post, the "tricky" part to me is finding the very-high resolution images that don't pixelate (or pixelate *much*) as they approach the camera.

For example, this is just two planes with high-luminosity patterns applied (generated in the kaleidoscope scene I referenced and modifed with an image editor using edge detection, color balance, contrast, etc.), and with the Corona Image Filter applied:

142277

MOV file: 142279

Add a little Ligeti etheric music on top, beef up the image resolution even more, and it'd look pretty good, IMO. For example here's the same MOV file with a fragment of Ligeti "Atmospheres" (fair use): 142280

For the V-ger scene, a lathed "vase" type structure with appropriate surfaces / lighting / hypervoxels applied may be MUCH easier to implement in LW that the slit-screen technique.

Finally, as I said in that post, if you're interested in simulating the *process*, then using the prior scene with high-res geometry images and very high MB passes ought to do the trick (I think!). :)

mTp

jeric_synergy
07-17-2018, 09:09 AM
For me, I'm still confused as to what effects wouldn't be much easier in LW using surfaces, volumetrics, lighting, etc. Slit-scan was intended to *simulate* what we can do easily in LW - i.e. an endless flat plan with high-resolution geometric light patterns on it over which the Camera flies. As I said in that post, the "tricky" part to me is finding the very-high resolution images that don't pixelate (or pixelate *much*) as they approach the camera.

mTp
Wondering: how would simply suspending a physical (RW) camera very close over backlit transparencies on a conveyor belt have differed from the slit-scan effect??? DOF? (It appears they adjusted the focus continuously throughout the camera's travel.)

MonroePoteet
07-17-2018, 09:46 AM
Yes, Depth of Field would kill the close-up method. I do model railroading as a(nother) hobby, and photographing models closeup without the DOF fuzziness somewhere in the image is problematic. Pinhole cameras work OK, but produce the standard diffraction aberration around the edges unless the depth of the pinhole is basically zero, and even then the focus tends to get fuzzy around the edges, and you can ONLY get fish-eye type views.

A method for making a "zero depth" pinhole is to take a thin metal sheet, make a slight dimple in it with a very sharp punch (not all the way through!), and then carefully and evenly sand down the dimple until the pinhole appears - the sanded down edges next to the hole will be very thin.

One technique for in-focus model photography is to take numerous photographs from the same position with the same camera settings, but modify the focus for each successive "depth" and then composite them in post. I suppose that same technique could have been used instead of slit-scan, but unless you take a LOT of variable-focus images and composite them all I doubt the multi-image technique would produce the sensational clarity of slit-scan. For slit-scan the focus on the camera is modified as it slides toward the slit, a predictable, continuous focus change so the slit is continously in focus.

mTp

jeric_synergy
07-17-2018, 10:01 AM
"Zero depth"!!!! (0)_(0)

:bowdown:

DonJMyers
07-17-2018, 11:14 AM
For me, I'm still confused as to what effects wouldn't be much easier in LW using surfaces, volumetrics, lighting, etc. Slit-scan was intended to *simulate* what we can do easily in LW - i.e. an endless flat plan with high-resolution geometric light patterns on it over which the Camera flies.


Yes, for me as I worked on it I realized LW can do the effect much better some other way but I really wanted to see what went through Trumbull's mind as he was creating these images by making an accurate slit scan machine. Unfortunately I assumed that LW 2018 was the way to go because I forgot it has fewer motion blur options. Ooops! I'll never uninstall LW 2015!

MonroePoteet
07-17-2018, 03:32 PM
Forgot to post the scene file, so here's the scene file for the double-plane slit-scan simulation but with smaller resolution images. The high-resolution image on the bottom plane was over 14MB by itself (rendered 4096x4096 out of LW), which exceeds the forum upload limit.

Note that the image on the bottom plane was taken directly from the LW render of a frame in the Kaleidoscope scene:

https://forums.newtek.com/showthread.php?154983-filigree-Patterns&p=1521973&viewfull=1#post1521973

and modified to have extreme contrast in the Image Editor=>Editing panel. The image on the upper plane was modified / distored in an image editor (I use Corel PhotoPaint).

At the 640x480 render resolution in the sample scene, the low-resolution images aren't too apparent. Rendering at higher resolutions (e.g. for film) would definitely have clearer / crisper geometrical patterns on the planes with the high- or ultra-high resolution images.

142289

mTp

jbrookes
07-17-2018, 09:51 PM
That .mov file looks awesome!!

As for the V-GER energy field, I'm pretty sure that was airbrush on glass. So the idea of mapping a vase shape is definitely the right idea.

jeric_synergy
07-18-2018, 12:18 AM
How does the 2018 motion blur differ from the 2015 in words of one syllable or less? The methodology, that is. IOW, how do they get MB in 2018 and why does it not do the same thing it used to?

MonroePoteet
07-18-2018, 08:18 AM
Well, I can't even say "Motion Blur" with single syllable words! :) How about: "It don't work for slit scan". :)

In LW2015, there were three options for Motion Blur: Classic, Dithered and Photoreal. In LW2018, there's simply a checkbox which enables the LW2018 version of Photoreal motion blur.

142296 LW2015 Motion Blur Type options

142297 LW2018 Just a checkbox to Enable / Disable MB

For a moving object Classic MB in LW2015 worked by retrieving the Object's position in "subframes" after the current frame and producing semi-transparent copies of the object, with the MB Passes determining the number of copies. For a moving Camera, it uses the combination Camera and Object motions. Each copy was a clean, recognizable copy of the object, but semi-transparent based upon the speed.

Here's a blue ball moving quickly left to right with Classic motion blur, 50% blur length and 6 passes:

142292

Dithered did a similar sampling, but doubled the number of samples and then produced the semi-transparency by dithering the object copies, i.e. only showing a sampling of pixels on the object, making the copies of the object a little less distinct. Here's the same scene and frame, just changing the MB type to Dithered:

142293

Finally, Photoreal did the same time sampling, but made each copy of the object even less distinct through some sort of "noise" algorithm. Same scene and frame, just changing the MB type to Photoreal:

142294

As mentioned, LW2018 simply has a checkbox which implements Photoreal motion blur, but the object copies are even less distinct than with Photoreal in LW2015. Same settings but with the default & only available MB type in LW2018:

142295

So, the issue is that to fake slit-scan you move the camera rapidly toward the slit and MUST enable motion blur to simulate leaving the shutter open during the move. In LW2018, this causes the "copies" of the object (i.e. details of the pattern showing through the slit) to lose most of their distinction, making the resulting pattern fuzzy and out-of-focus.

In LW2015, you could change MB type to Classic and get semi-transparent but clear copies of the original object, so they retain "focus" even close to the camera, giving something akin to the clarity of the slit-scan process. By bumping up the Luminosity of the pattern to overcome the semi-transparency of the copies (I used 100,000% in my sample scene) and having the MB passes VERY large, you could come close to simulating the open-shutter effect of slit-scan.

mTp

MonroePoteet
07-18-2018, 08:26 AM
Yes, for me as I worked on it I realized LW can do the effect much better some other way but I really wanted to see what went through Trumbull's mind as he was creating these images by making an accurate slit scan machine. Unfortunately I assumed that LW 2018 was the way to go because I forgot it has fewer motion blur options. Ooops! I'll never uninstall LW 2015!

Yes, I'll be keeping LW2015 indefinitely for NON-photorealistic endeavors, and indeed may use it primarily for my feeble, hobbyist animation efforts. The "fakery" available in pre-LW2018 was very valuable, IMO, and I'm not that interested in "faking reality" or "physics based rendering" or "photorealism".

mTp

jeric_synergy
07-18-2018, 08:27 AM
Thank you Monroe, for that summation.

I sorta knew how old MB (emm bee, single syllables! ;) ) worked: so it seems 2018 MB does pretty much the same procedure, but with more blurring?

BTW, I upped your example file to 1000 passes, 1000% Blur Length..... it was interesting....

++++++++
2000 passes / 100% BL /Shtr eff=100%:
142298

+++++++++
The "slit" geometry in my version was a bit janky, so I recreated it w/a slightly wider (?) slot, and it gives a smoother effect.

The proper thing for me to do is to step-morph the slit to see what varying widths do. --But first, recreate the effect w/CGI techniques to see if there's any difference (prolly not). Maybe some of the bloom effect might be a bit different. RW film of course is additive, as it appears is the 2015 MB.

Fun, yet academic, times!

DonJMyers
07-18-2018, 12:59 PM
forgot to post the scene file, so here's the scene file for the double-plane slit-scan simulation but with smaller resolution images.

Mtp

thanks so much i will play with it today !

MonroePoteet
07-18-2018, 10:17 PM
You bet, happy to help! But, I cannot impress enough the importance of HIGH definition images to be mapped to the planes (moving in the Z axis to simulate camera motion). Using the Kaleidoscope scene I referenced (with the image of your choice as the Backdrop image) and rendering the output at 4096x4096 will provide some nice detail. The higher the resolution of the mapped images, the better the final double-plane render will be as the images approach the Camera.

Have fun!
mTp

P.S. as a caveat, the Corona Image Filter can take a while to process. If the rendering seems to hang up, just walk away and let it finish - well worth the wait, IMO! mTp

jeric_synergy
07-19-2018, 09:40 AM
Ah, Corona: I wondered why the blooming.

(This rather points out a difficulty w/LW in general: it's rather inconvenient/difficult to 'autopsy' a scene, because you have to look EVERYWHERE in the UI. Maybe it's the nature of the beast, but other s/w seems to have more centralized locations where one can see all the pieces simultaneously.)

BTW, over on FB, someone seems to have managed to get the effect in 2018. But maybe the EFFECT, and not the PROCESS.

MonroePoteet
07-19-2018, 01:15 PM
The double-plane scene file I posted above is a LW2018, although it was imported from a LW2015 scene / objects file. Yes, it's the *effect* but not the slit-scan *process*. If you have a pointer to the effect on FB it'd be nice to take a look (assuming I don't have to log in!!).

RE: dissection of a LW scene, might be worth a Feature Request on the Schematic View to show *all* panes that have active data / changes to be written to the scene file. They clearly have to keep track of this internally to NOT write out extraneous stuff to the scene file, so it may be (I repeat, MAY be) fairly simple to walk the list they already have and draw a schematic, or at least put boxes on the Schematic View for each one. It might be (repeat, MIGHT be) as easy as setting a flag to not actually write-out the scene file, but add a box to Schematic View while executing the same code.

mTp

jeric_synergy
07-19-2018, 04:02 PM
Waiting for confirmation as to whether it's a "effect" vs "process" image.

Centralized info: it's a bit of a "infographics" challenge, to usefully display all the stuff that's happening. In C4D, the only other app I'm anywhere near up to speed on, there's a system of "Tags", but there's also Layers, and Takes, so..... perhaps it's just a global problem.

With Lightwave specifically, some kind of speadsheet-like panel, possibly w/different "views" depending on what you'd like to examine. EG, all motion modifiers, even if buried in the GE. Certainly, and obviously, all the information is implicit in the Scene file, but displaying it quickly would be quite a design challenge. Not much of a code challenge though.

EG, one big hierarchy, kind of the reverse of how a render might work: start with any processing the final image gets, eg Corona, and then examine the first "child of the scene", however that might be defined: a child of an object, for instance, would be a Surface. (Note that in other apps, "Surfaces" or more usually "Materials" are a separate data structure that can be applied and altered to multiple items simultaneously-- in LW a Surface truly seems to be a 'child' of the associated mesh.) Another category of children attached to items would be Motions, etc etc.

Depicting all the possible relationships would be.... great fun!