That's right, Peanuts celebrated its 60th birthday last month and we mark the occasion with the 50th episode of Brudders, which we're calling, well, 'Peanuts'. (Good grief!)
Read the artist's notes about this strip here and write your own comments and questions.
- Sgt M.
November 22, 2010
November 9, 2010
More cowbell!
In my previous post I mentioned that there were a few missing cowbell hits at the beginning of Sister's motion capture data. As it turns out this is easily correctable from within iPi DMC. First I enable IK or FK on the rig which allows me to manually make a pose correction. Then I click on Refit Pose, and press Track Forwards and Track Backwards a few frames from the corrected frame. Very simple and elegant, especially with the multiple video references visible in the backdrop to 'rotoscope' the action to. This last point is especially important in this case because the motion has to timed to musical events I can't hear in the iPi DMC playback.
Also, if you run into a bad 'pop' in a joint, try stepping to a frame after the pop, hit Refit Pose a couple of times (and maybe Analyze Actor again followed by Refit Pose one or two times,) and then track backwards right through the pop. Sometimes tracking backwards can be more precise. (Note: you may need to track forward a few frames from the frame after the pop to help smooth the motion forward from that frame.)
For you techies who just can't get enough of this 'geek speak', tracking Toullie's 1922 frame performance shown above took 1 hour and 30 minutes to process. (iPi DMC's initial rough estimate was about 2 hours.) This was with a Nvidia Geforce GTX 260; with a GTX 480, I believe the speed could be two to four times better.
- Greenlaw
Also, if you run into a bad 'pop' in a joint, try stepping to a frame after the pop, hit Refit Pose a couple of times (and maybe Analyze Actor again followed by Refit Pose one or two times,) and then track backwards right through the pop. Sometimes tracking backwards can be more precise. (Note: you may need to track forward a few frames from the frame after the pop to help smooth the motion forward from that frame.)
For you techies who just can't get enough of this 'geek speak', tracking Toullie's 1922 frame performance shown above took 1 hour and 30 minutes to process. (iPi DMC's initial rough estimate was about 2 hours.) This was with a Nvidia Geforce GTX 260; with a GTX 480, I believe the speed could be two to four times better.
- Greenlaw
November 8, 2010
Take 2...no wait...
Shot the new session this morning. It seemed like all went well until I got back to the studio and started tracking the footage. That's when I realized I forgot to take iPi Recorder out of Calibration mode and all the footage was underexposed. Sigh! Went back to the stage for a another retake.
Now I'm back at the studio again and this time the data really does look good. Will start tracking the new performances after I pick my daughter up from pre-school and have some lunch.
Update: Okay, back from lunch and have already tracked Sister's performance. Looks pretty good. Some slight jitters when she's standing still but within the context of the shots I don't think it will be noticeable. What will be a problem is that we're losing some of the smaller cowbell hits in the early part of the data. Apparently, there is a threshold for what motions iPi DMC will pickup or ignore. No big deal; it should be a minor issue to keyframe the missing hits.
Oh, yeah, and rolling up the pants cuffs fixed the heel problem.
- Greenlaw
Now I'm back at the studio again and this time the data really does look good. Will start tracking the new performances after I pick my daughter up from pre-school and have some lunch.
Update: Okay, back from lunch and have already tracked Sister's performance. Looks pretty good. Some slight jitters when she's standing still but within the context of the shots I don't think it will be noticeable. What will be a problem is that we're losing some of the smaller cowbell hits in the early part of the data. Apparently, there is a threshold for what motions iPi DMC will pickup or ignore. No big deal; it should be a minor issue to keyframe the missing hits.
Oh, yeah, and rolling up the pants cuffs fixed the heel problem.
- Greenlaw
November 7, 2010
My pants do vex me!
Ran into a new snag today. It turns out that in my previous mocap session, my pants cuffs were too long, covering the back of my feet and preventing iPi DMC from tracking the heels accurately. The result is that, even though the toes in the rig are pinned to the ground, the heels are swiveling slightly from side to side, and this motion is riding up through nearly every joint in the body. Looks like it's back to the stage tomorrow morning to recapture this session.
Fortunately, I have this performance pretty well rehearsed by now and it should only take a half hour to recapture all of Act I.
Otherwise, the software seems to be tracking the performance with reasonable accuracy. Can't wait to see it applied to the actual character rigs.
- Greenlaw
Fortunately, I have this performance pretty well rehearsed by now and it should only take a half hour to recapture all of Act I.
Otherwise, the software seems to be tracking the performance with reasonable accuracy. Can't wait to see it applied to the actual character rigs.
- Greenlaw
Acting!
Today I'm tracking the motion capture performances from earlier this week. So far the motions are looking very good. Some expected jitters in the raw data, but iPi DMC has very good Jitter removal, which I'll run before saving the final data.
I may have to redo a couple of performances; I acted a bit broadly in two segments because I didn't believe the system would pick up on subtleties. Boy was I wrong. iPi DMC faithfully captured my exaggerations with embarassing accuracy. Now I understand why a lot of motion capture performances can look so bad. Less is definitely more here.
- Greenlaw
I may have to redo a couple of performances; I acted a bit broadly in two segments because I didn't believe the system would pick up on subtleties. Boy was I wrong. iPi DMC faithfully captured my exaggerations with embarassing accuracy. Now I understand why a lot of motion capture performances can look so bad. Less is definitely more here.
- Greenlaw
November 2, 2010
Here we go!
I finally got around to setting up the new motion capture gear this morning and did some preliminary tests. The new computer and software appears to be capable of capturing four streams of synchronized video at 640 x 480 at 60fps! Previously on my older system, the best I could get consistently was four streams of 320 x 240 at 30fps. What I think is making the difference with the new computer is having installed two USB 3.0 cards and an internal 1TB SATA 3 drive dedicated for video capture.
Now I'm wondering how well this system can handle six Playstation 3 Eye cameras. I'm currently plugging two PS3 Eye cameras into one on-board controller, two more into a USB 3.0 card, and I have the internal SATA 3 drive plugged into the second USB 3.0 card (which happens to be a SATA 3 card too.) Since there are two on-board USB 2.0 controllers, in theory I could add two more cameras. The question is, is the SATA 3 drive fast enough to record six streams of video data without dropping frames?
For now, I'm sticking with my four camera setup though; I really do want to get this film done this year.
I just finished writing the motion capture performance script, and actual production shooting (i.e., not another test) will begin in the morning after I drop my daughter off at pre-school.
Stay tuned!
- Greenlaw
Now I'm wondering how well this system can handle six Playstation 3 Eye cameras. I'm currently plugging two PS3 Eye cameras into one on-board controller, two more into a USB 3.0 card, and I have the internal SATA 3 drive plugged into the second USB 3.0 card (which happens to be a SATA 3 card too.) Since there are two on-board USB 2.0 controllers, in theory I could add two more cameras. The question is, is the SATA 3 drive fast enough to record six streams of video data without dropping frames?
For now, I'm sticking with my four camera setup though; I really do want to get this film done this year.
I just finished writing the motion capture performance script, and actual production shooting (i.e., not another test) will begin in the morning after I drop my daughter off at pre-school.
Stay tuned!
- Greenlaw
November 1, 2010
Hello Gothtober 2010!
A day late, but is it ever really too late to celebrate Gothtober? Yesterday our friends at Gothtober premiered Act I of the animatic (a moving storyboard) for our next Brudders movie 'Don't Fear the Pooper'. In a few days we're going to start posting the production log on the website so you can follow the weekly progress of the fully animated movie, which we intend to release in mid-December. (Like I said, it's never too late to celebrate.)
In the meantime, you can visit www.gothtober.com and watch the 'teaser' by clicking on the big button for October 31st. And while you're there, please check out the many wonderful Howl-o-ween projects by other Gothtober artists.
Oh, and buy some of their stuff. Fiddle Cat is awesome!
- Sgt M.
In the meantime, you can visit www.gothtober.com and watch the 'teaser' by clicking on the big button for October 31st. And while you're there, please check out the many wonderful Howl-o-ween projects by other Gothtober artists.
Oh, and buy some of their stuff. Fiddle Cat is awesome!
- Sgt M.
Subscribe to:
Posts (Atom)