|
|||||||||
|
Thread Tools | Search this Thread |
June 14th, 2009, 10:56 PM | #16 | |
Major Player
Join Date: Dec 2008
Location: Laguna Niguel, CA
Posts: 277
|
Quote:
|
|
June 15th, 2009, 04:48 PM | #17 | |
Major Player
Join Date: Dec 2008
Location: Laguna Niguel, CA
Posts: 277
|
compresor vs. motionperfect results
Quote:
Comments: 1) Motionperfect appeared to crash at the end, even though it did all 15 seconds. This trial is limited to 15 secs so that may have something to do with the crash. 2) Motionperfect required an avi file so I used neoscene. This doesn't affect the actual input frames. It does slightly affect the job motionperfect had to do because it converted 29.97 to 24fps instead of 30fps to 24 fps. 2) Motionperfect put out a GIANT file. This might have something to do with the crash: 86 mb Your original 106 mb Your 24fps file. 319 mb After original -> neoscene conversion 2.3 gb After motionperfect conversion to 24fps 3) The compressor output looked like simple frame blending with no interpolation. I think you had something wrong with your settings. I could have gotten the same results by telling premiere to export at 24 fps. Here are two consecutive frames. They are typical of all the frames in the compressor output. http://elleh.com/public/comp-1.jpg http://elleh.com/public/comp-2.jpg 4) The motionperfect output blew me away. I had trouble finding artifacts. The two frames I chose to use (same two as above) had the worst defects I could find. Look at the bottom of the right leg which is moving pretty fast: http://elleh.com/public/motp-1.jpg http://elleh.com/public/motp-2.jpg I will try to find out how to get the giant motionperfect output file into a reasonable size so people can download it and see what I saw. I'll start by contacting support. All in all I am very happy that good tweener frames can be produced. Now it is just a matter of getting the software to work right. Can you please take another stab at the compressor conversion? P.S. My output had the blacks crushed. I know neoscene doesn't do it so motionperfect must have. So that is another thing to bitch to them about. |
|
June 15th, 2009, 09:32 PM | #18 | |
Major Player
Join Date: Jan 2005
Location: Wellington, New Zealand
Posts: 513
|
Quote:
Each time I had "High quality motion compensated" checked, as per the pic. Hmmm... |
|
June 15th, 2009, 09:42 PM | #19 | |
Major Player
Join Date: Dec 2008
Location: Laguna Niguel, CA
Posts: 277
|
Quote:
It is very easy to tell the difference between frame blending and interpolation. Frame blending has the constant blur problem and interpolation has less frequent, but worse artifacts. Right now I'm not sure motionperfect is good enough for my needs. But it is definitely a step in the right direction, especially if I fix bad frames in photoshop. |
|
June 15th, 2009, 10:01 PM | #20 |
Inner Circle
Join Date: Nov 2005
Location: Elk Grove CA
Posts: 6,838
|
Here is a motion perfect render. I liked what it looked like in general, but if you watch "qi" on the teeshirt at the very beginning, you will see an aritfact from the motion perfect render.
__________________
Chris J. Barcellos |
June 15th, 2009, 10:19 PM | #21 | |
Major Player
Join Date: Dec 2008
Location: Laguna Niguel, CA
Posts: 277
|
Quote:
I just want to know what the state of the art really is. I've heard over and over on different forums about compressor and I want to see how good it is. MotionPerfect has already exceeded my expectations. You're "qi" artifact could easily be touched up in PP. |
|
June 16th, 2009, 02:05 AM | #22 | |
Major Player
Join Date: Jan 2005
Location: Wellington, New Zealand
Posts: 513
|
Quote:
"High quality motion compensated" is the critical tab to have checked. Here's a couple of frame grabs to demonstrate that with identical settings (only 25p this time) interpolation is taking place, but... it has the kind of blur one would associate with frame blending, just not so severe. On the close up (100%) you can see on the HQ motion compensated 25p version the pole distorts in front of my son's head as Motion generates a new frame. On the 50% shot, four frames later, you can see how Motion struggles with the arm, just as it did in the stream shots I sent you. For reference, I have attached grabs from the 30p original and the 25p frame blending version. You can clearly see the differences. |
|
June 16th, 2009, 02:56 AM | #23 | |
Major Player
Join Date: Dec 2008
Location: Laguna Niguel, CA
Posts: 277
|
Quote:
Since my last post I have found that while MP will usually make objects (people) in the foreground look perfect with no blending (which is amazing to me), the background will appear to "flow" around the foreground object. It is hard to describe because you cannot see it in a still. Speaking of stills, we have been concentrating on them and they mean nothing. The final appearance in motion is what matters. I'm going back and playing those files a few times to try to get a subjective feel for their qualities. P.S. In my MP experimenting, I came across a still or two where the foreground leg moving rapidly would have a gaping whole you could see right through to the background. Very gross. Again it happened in single frames which could be patched. |
|
June 16th, 2009, 03:25 AM | #24 | |
Major Player
Join Date: Dec 2008
Location: Laguna Niguel, CA
Posts: 277
|
MotionPrefect loses the war
Quote:
MotionPerfect was unusable and is not fixable. There is a "breathing" of the background following the moving objects that jumps out at you and looks unnatural. Compressor looked the best with a blur that some might consider filmy. However, I tried a straight blended output from Premiere and it looked identical to my eye. I was only looking at the high speed artifacts. I need to study natural talking head speed also. Maybe compressor blends at high speed and interpolates at low. I also need to go back and look for MP alternatives for the PC. I compared originally by looking at tweened frames. MP frames look awesome. I'm going to look for something with compressor type logic. Maybe we now know why MP is so cheap. |
|
June 17th, 2009, 01:05 AM | #25 |
Inner Circle
Join Date: Nov 2005
Location: Elk Grove CA
Posts: 6,838
|
Yes. Motion perfect is usless with high motion shots. The attached clip of clearly shows how bad Motion Perfect can get. Clearly, this is not fixable.. Oh well, back to the drawing board.
__________________
Chris J. Barcellos |
June 17th, 2009, 06:07 AM | #26 |
Regular Crew
Join Date: Jul 2005
Location: Spain
Posts: 67
|
It does what is expected to. Nothing less, nothing more.
Having a 30p converted to a 24p with no flaws, is like expect converting an 720x480 footage to a 1920x1080... it's NEVER going to be great. Besides, I wold never want a single of that artifacts on my videos. I rather prefer to stick to 30p, wait for a firmware update or buy another camera. |
June 17th, 2009, 02:47 PM | #27 | |
Major Player
Join Date: Dec 2008
Location: Laguna Niguel, CA
Posts: 277
|
Quote:
I think maybe blending looks better at high speed because natural motion blur is going to be big anyway. So a little more doesn't hurt. |
|
June 17th, 2009, 02:50 PM | #28 | |
Major Player
Join Date: Dec 2008
Location: Laguna Niguel, CA
Posts: 277
|
Quote:
|
|
June 18th, 2009, 05:06 AM | #29 |
Inner Circle
Join Date: Feb 2009
Location: Norwich, Norfolk, UK
Posts: 3,531
|
I used Josh's test clip & converted to 24p using the method described by Denver Riddle & demonstrated by Philip Bloom in this video tutorial How to convert Canon 5dmk2 footage from 30p to 24p - Canon 5D Mark II on Vimeo In brief the workflow converts the original H.264 .MOV to ProRes as the intermediate CODEC & then uses Cinema Tools to conform the footage to 24p i.e. turns the 30p into 24p by simply flipping the bits in the header which keeps the same number of frames plays them back more slowly. Effectively an overcrank. Compressor is then used on this clip to bring the duration back to the original. The unaltered audio is dropped back onto the clip after the video conversion to 24p.
Philip describes it as how to convert footage shot in 30p to perfect 24p Well perfect it isn't. The video looks good (which is I suppose the ultimate test) but stepping through frame by frame it is easy to find artefacts anywhere there is a lot of motion. Philip's sample video has very little motion so doesn't show these artefacts. Perhaps it is unfair to compare individual frames after conversion as it doesn't matter how many artefacts there are present as long as the viewer doesn't notice them & the video looks good. If/when Canon gives us 24/25p we will be able shoot footage side by side & do a proper comparison between native 24/25p & 30p converted to 24/25p. |
June 18th, 2009, 11:05 PM | #30 |
Inner Circle
Join Date: Nov 2005
Location: Elk Grove CA
Posts: 6,838
|
Tonight I had a bit of time to try a couple 24p renders using different methods out of Vegas, and using my the Cineform codec that comes with NeoScene.
The first one, I took a cue from Phil Bloom recommended process, and slowed a section footage down to play as 80 % frame rate. I put the resulting file on the same time line, sped it back up to 1.250%, and then rendered it to 24p. I rendered it to a window media file for posting here. The other, I just did a direct render to 24p using the same Cineform codec. They are both posted below. I am thinking this is not so bad...
__________________
Chris J. Barcellos |
| ||||||
|
|