Aston martin 3d tracking and insertion

[mdharrington]

Great execution, it will definetely fool almost everyone. The guy's performance though... : D

Ha ya, he's actually talking about a Bugatti...so I had to mute him

[ten]

The tracking problem was .....I was walking then stopped(almost) and did a basically nodal pan, and syntheyes never really could solve that motion properly. So I think the last part of the solve is pretty accurate, while the first part may be a little off.

Have you ever tried Mocha pro? It sometimes excels where usual trackers fail.

ten

[mdharrington]

syntheyes was needed for the texture extraction...
and none of my trackers were slipping....that was not the issue...the issue was the solve..
it was animated between nodal pan and regular, but the nodal pan was not stationary....a bit of a Frankenstein track.

I can trace the entire problem to panning to my left while still very slowly walking right. Syntheyes did not appreciate that move very much.

Mocha can pull of some smooth 3d tracks, and would have made a decent track, but there would be no way to recover the textures....syntheyes can pull a 6-8k still from moving video combining pixels from different frames...in the shot breakdown you see the result of the texture extraction and it is quite high quality

here is a still of the scene no car...the little bit of blur you will notice, is due to incorrect solve
with a good solve this would be razor sharp

[motion]

Good job.. The reflections really sell it. Did you use an hdri or daylight to light your scene?

[mdharrington]

Just daylight....
But I was able to match the direction of the sun perfectly using shadows extracted in the texture, and the daylight sky was pretty close to the actual sky.

[RealityFox]

Really nicely done. I think that the materials need to be tweaked on the car a bit more, the wheels looked pretty nice though. I find that the grating in the front of the car is a bit distracting, like it doesn't feel very natural? The car paint itself is fairly good but needs some tiny tiny adjustments to it.

Needless to say though, well done.

[gordonrobb]

Have you sought help over at Ssontech forum?

Fairly sure you can solve the shot for the first bit, then apply that solve to the second bit (iirc).

[mdharrington]

Fairly sure you can solve the shot for the first bit, then apply that solve to the second bit (iirc).

yes you are right...and that's what I did. Also had to animate the solver from normal to nodal and back during the shot

I have been using syntheyes since beta, and was actually going back and forth with russ adding some features specifically for this shot, namely using an embedded alpha instead of a garbage spline to block out the actor from the texture extraction. He added a few features within hours of communicating and would send me the beta's

Not saying someone else couldn't have solved this better, it was getting to the point where it would require manual adjustment of the seed path....and I kinda get lost there

it's really a low parallax portion that is so bad....its kinda embarrassing demo of my tracking abilities....this was more about the texture extraction an integration workflow

[mdharrington]

Really nicely done. I think that the materials need to be tweaked on the car a bit more, the wheels looked pretty nice though. I find that the grating in the front of the car is a bit distracting, like it doesn't feel very natural? The car paint itself is fairly good but needs some tiny tiny adjustments to it.

Needless to say though, well done.

Thanks

Ya the front grill grating isn't real geometry....just a texture.

The paint i spent a hour or 2 on....using falloff node (if I recall correctly) was really twitchy. There is a bit of paint flecks from turbulence underneath but it is very subtle...

Not sure I have the skills to improve the paint that much more other then hand painting some texture maps for orange peel and swirl directions.