Earthquake - VFX Challenge - Astrofish


#1

(Edit 10th Dec)
Here’s a link to a page containing my final animation.

Remember to come back and read the thread!
(End of edit)

Ok, after playing around with some tests over the weekend I’ve decided to give this challenge a shot.

I’m interested in VFX work, but I’ve never really done any before so this should be interesting.

This is also the first challenge that I’ve ever entered, hoping to learn a few things in the process.

Good luck to everyone, and I’ll post my storyboard tonight when I get home.

Cheers!


#2

Accidental duplicate posting - deleted.


#3

Ok, Here’s my incredibly badly drawn storyboard. Don’t laugh…

In case it’s not too clear, the sequence is:-

  1. View down onto motorway from bridge, cars driving along the road.

  2. Crack appears from under the bridge and rapidly travels up the road. Cars steering to avoid it.

  3. Second crack appears further across the road just as a car reaches that part of road.

  4. Second crack splits wide open, and the piece of road between the two tips over and drops into the depths, taking the car with it…

I’ve already got the footage for this, it’s something left over from another project. Here’s a frame:

I’m probably going to crop this to focus on the lower right of the frame, there’s too much of the other side of the road in this shot, and it’s too distracting.


#4

Hey there astrofish,

I really don’t know a lot about VFX…but from what i know it seems to be difficult removing all of the miving cars and replacing them with cg.

I dunno…just thinking about it, if i am wrong, well :smiley: go for it cuz it seems to be a great project u got there :smiley:


#5

Yeah I agree. You’re going to have to paint out the road so that there are no cars, then replace them digitally.

The footage also suggests that the road is wet - notice how the cars have a slight reflection on the road? This one is going to be tough to do…

Leo


#6

Thanks for your comments!

I’m hoping that I’m going to be helped a lot by the fact that there aren’t actually many cars in the raw footage, and by using the right 5-sec piece of the footage I can get it so that most of the cars are clear of the immediate vicinity of the quake area anyway.

I’m planning to only do a proper CG replace of one car (the one that goes down the hole).

For the other cars swerving out of the way, I’m going to see what I can do in After Effects. If I can’t manipulate the raw footage enough to give convincing swerving behaviour, then I’ll fall back on just retiming the cars so that they hit the brakes to avoid going down the hole. This won’t look as good, but will be acceptable I think.

Because there aren’t that many cars, I managed to create a clean plate by stacking 3 or 4 frames, and painting transparency onto the cars so that the road from other frames showed through. This was actually quite easy to do, so I’m going to repeat it a few more times using different frames, so that I’ve got a short looping clean plate sequence with the original grain on it.

Regarding the reflections - yes, it was a miserable wet day unfortunately.
If I end up retiming the original cars, then their reflections will be OK anyway. If I get them swerving then yes this could be difficult…

If the reflections do cause trouble then I thought that instead of trying to mimic them for the CG car, (and the swerving cars if needed), I could instead try using the clean plate to remove the reflections from the original footage.

I haven’t got any suitable footage from a dry day and it’s pouring down outside, so there’s no chance of getting anything better in time, so I’m going to work with it…


#7

Go for it! I have complete confidence in you! This is gonna look so cool!


#8

Ok, overdue for an update…

Here’s a summary of what I’ve done so far. I’ll try to post some images later.

  1. Aligned my CG camera with the clean plate and locked it.
  2. Placed a horizontal plane at the road surface level.
  3. In camera view, drew freehand splines to mark the edges of the roadway and the area that I want the effect to occur in.
  4. Also drew splines to mark out roughly the positions where I want the major cracks to appear.
  5. Projected the splines from 3 and 4 onto the plane, from the camera view. (Discovered that the C4D project spline function fails when projecting onto an undivided plane. Wasted hours figuring out the problem. Works OK with a subdivided plane.)
  6. Switched to top view. Checked that the splines that marked the road edges remained equally spaced in plan view - this helped to confirm that the camera alignment is about right.
  7. Still in top view, redrew freehand splines over the top of those drawn in 4, using those as a guide.

Notes on 7:
So why did I do this? Because by drawing the cracks in the top view, I get consistent size wiggles in the splines. The ones drawn in the camera view in step 4 are not consistent because they have constant sized wiggles when viewed from camera perspective. Since they should be the same size in plan view this is wrong.

So why bother with step 4 at all? Because I want to sketch out in the camera view where I want the action to be, in top view I can then see where my cracks need to be, and draw them in properly.

  1. Hid the splines from 4, and switched back to the camera view. Note that the scale of the wiggles from 7 reduces naturally as the cracks recede into the distance and look more believable, whereas the splines from 4 look like they were just drawn onto the picture… :slight_smile:

  2. Recreated the road surface from a large subdivided cube, cut into five major pieces along the spline edges.
    (To do this I had to split the splines into the individual sections joining the various cracks, duplicate them, then rejoin the appropriate sections to form the outline of each piece, then extrude and boolean union with the original cube.)

  3. Now I have five perfectly meshing pieces that represent the subsurface for the undamaged road.

  4. In top view again, drew a lot of fragments over and around the main cracks to represent the pieces of road surface that will be churned up. Adjusted these so that they mesh together properly to form a continuous road surface. Also created pieces that roughly correspond to the five ‘subsurface’ meshes. Extruded these elements slightly to give some thickness to the road surface.

This is where I’m at at the moment.
The stuff following is what I’m planning to do next:-

  1. Apply camera mapping onto road surface geometry.

  2. Group the large surface pieces to their corresponding subsurface pieces, and apply an animated free form deformation to each of the five pieces.

  3. Adjust the deformers to pull apart the elements at the cracks, and to adjust the road surface heights slightly as well.

  4. For the central road section, keyframe the animation of it falling away. This will finish the gross animation of the road breakup.

  5. For the fragments of road surface made in 11, make keyframe animations adjusting the positions and rotations to provide some more detail to the breakup.

I expect this bit to be quite slow and time consuming - so I’m going to see how this goes before I make any more detail plans for the next stages…

I’ll try to post some images tonight.

Lots still to do (good fun though)…


#9

It’s great that you show us the plan for how you do things. Making me discover new ways of doing things. Never thought of projecting splines drawn on the plate onto a plane for easy positioning of objects. This is good! :thumbsup:

Keep it up!


#10

Hi Knut,
Glad you found something useful in the post!

Here are a few pictures related to what I talked about in my previous post:

  1. Picture of camera view with freehand splines drawn over the top in this view.
    Green splines mark road border, blue ones mark the rough area where the cracks are to appear.

  2. View of same splines in top view, after they have been projected onto the plane. Note that the green road edge splines remain constant distance apart, confirming that camera is in about the right place.

  3. Same view as 2, but now showing the new crack splines drawn in top view.

  4. Same view as 1, but showing the splines drawn in 3. Note the improved perspective effect compared to the original splines.

Finally, I’m using deformers to pull the cracks apart, but rather than deforming the shapes directly, I’m making a deformer ‘slide past’ the shapes. This causes the crack to progress down the road rather than opening all at once.

Look at:

This is a top view of the road area in the editor (no textures).
I’ve colour-coded the five different main chunks of road so you can see how they relate to the splines themselves (shown in red).
The blue gridlike thing is a deformer which I have set to affect the two left-hand side blocks.
As it slides up the screen you can see that it pulls the left crack open from the bottom up.
Another deformer (not shown) is doing a similar thing to the two blocks on the right.
Also shown is the results of some simple keyframe animation on the central block, which falls away as the cracks go past on both sides.
I need to add a few more deformers to open up the smaller cracks at the left and right.

Hope this is of interest to people!


#11

Just a quick animation test.

A lot of the road surface is missing - this will be replaced later with a lot of fragments that will react to the crack as it progresses.

The parts of the road surface that are visible here will remain undamaged. I think I’m going to increase this area, to localise the damage more closely to the cracks.

The subsurface is colour coded here to show the underlying surface blocks. At the moment the edges are straight - these need to be broken up, and of course they need texturing.

No traffic in this shot - this is done entirely with the clean plate.

The animation is mainly to show the timing of the main elements in the shot (except the car). I’m thinking of starting the crack earlier, since it seems too rushed at the moment…


(It’s small - 83k).


#12

NOOOOOO!! I can’t see it! Help!!


#13

This IS the right link for the movie!

Woah that is one deep hole in the ground. I think the animation gives the viewer a great sence of depth and size. Maybe not to realistic, but who carees! Great work so far.

But since you deform the surrounding road surface and if you will hav cars from the video there, won’t they be deformed too?

:thumbsup: Knut


#14

pretty cool !!!

must read everything now to know how U did it !!!

evn like this it’s good !!!


#15

Hmm, I’ve just checked and the link should work fine (I’ve just downloaded it myself via the link). It shouldn’t be a codec problem as I’m using the Sorensen codec…

For what it’s worth, here’s a higher res still from the sequence.


#16

Hi Knut & baby,
Glad you like it. It’s just a test so still missing a lot of stuff.

Fasty: Have you managed to get it yet?

Knut: Yes, the road surface deformation is potentially an issue, but there’s a few things that will help:

I’m probably going to crop the shot to focus more on the lower right anyway, so the road should look a bit more like its moving rather than just stretching, and less of the other side of the road will be in the shot anyway - so those cars won’t matter.

The oncoming cars will be in the area where the road is least stretched, so they shouldn’t be too badly affected.

If it still looks to bad, then I’ll just make the cracks narrower - since the bulk of the hole comes from the segment that falls away. (No - it’s not realistic, but hopefully it will look good!)

Camera shake will help hide some of this as well!

I was originally intending to have the cars swerving to avoid the cracks but given how much else I’ve still got to do to the shot, I’m probably just going to retime them so that they slow down and stop - except for the one that falls in of course…

Anyway, time for sleep…


#17

Yes I just got it! Looking GREAT! Can’t wait to see more!


#18

WHO’S DA MAN!
great work!! but in my point of view i would prefer the falling block to just fall a few feet instead of disapearing rom the cmeras…as it were to hit another layer under the road…

but again

U’DA MAN!!

:applause:


#19

Glad you like it so far.

Yes, It would be more realistic if it just dropped a bit, but I quite like the idea of it opening up a gaping chasm… I might try it the other way as well, to see what it looks like.

Anyway, since when were effects films realistic! :slight_smile:

Cheers - Steve


#20

Just thought I’d post a few observations/issues that I’ve made when processing my footage using the Sorensen 2 Codec…

Raw footage started as PAL DV.

I imported this into Icarus to stabilise it, as it had a small amount of camera shake on it, and saved the result out as raw JPG frames.

I then re-encoded the frames using the Sorensen 2 codec, with a keyframe every 10 frames. This produced stabilised footage in a MOV file.

Call this ‘mov_in’.

I then used ‘mov_in’ as an animated background plate in C4D, with some CG elements over the top of it, and got C4D to render the results out again using Sorensen 2, this time with a keyframe every 25 frames, quality=high. Call this ‘mov_c4d’.

Results ‘movc4d’ looked ok, EXCEPT that every tenth frame (i.e. corresponding to the keyframes in ‘mov_in’) was noticeably brighter than the other frames. Noticeable colour banding was also present in all frames.

As a comparison, I then tried loading ‘mov_in’ into After Effects 5.0, and rendering the output again as a Sorensen 2 file with keyframes every 25 frames, quality=hig (so basically just transcoding from Sorensen to Sorensen). Call the output ‘mov_ae’.

In ‘mov_ae’, EVERY frame was overbright in the way that every tenth frame in ‘mov_c4d’ was.

I then repeated the C4D and AE tests but using the raw jpg image sequence as input (thankfully C4D supports this!), and the results were both fine.

Although C4D and AE produced different problems, both produced bad output through the Sorensen codec when the source footage was also Sorensen encoded. This suggests that the problem is with the codec rather than either of the apps.

Conclusion:
Sorensen 2 seems to have problems when re-encoding footage that was decoded from a Sorensen 2 source.

Since the decoder and the encoder are completely seperate processes, I can only assume that the encoder produces artefacts in the images which the encoder is susceptible to.

Actually, the other possibility is that it is the Quicktime player that is at fault, although most of the work there is being done by the codec again.

In a different project I was tearing my hair out trying to figure out why in two identical compositions rendered from after effects (Soresen again as it happens), one of them seemed not to apply the gamma correct that I’d applied, and the other one did.

I spent ages cursing After Effects for not rendering the same as it was previewing. Then I noticed that although one movie was clearly darker than the other when viewed side by side in two quicktime windows, which one was darker depended upon which order I opened them in. In other words it was a playback problem, not a rendering problem.

One final comment, here’s a tip for reducing file sizes whilst keeping decent quality:
It seems that you can push things a long way by setting keyframes to be very infrequent - at least when there’s not much motion. It remains to be seen what adding camera shake does for the compression - depends how smart they are…

Anyone else have any comments on any of this?

Bye for now!