|
|||||||||
|
Thread Tools | Search this Thread |
September 12th, 2003, 12:28 PM | #1 |
Space Hipster
Join Date: Oct 2001
Location: Greensboro, NC
Posts: 1,508
|
Magic Bullet 1.5 Speed results
Here are the numbers:
I used a 11 second and 12 frame (344 frames) DV stock footage clip that was shot interlaced. I applied Magic Bullet, Deartifacter for DV, Broadcast Spec (Composite), LetterBox 1:1.85 and Miami LookSuite. I also made a levels adjustment. Project set at 16bpc. Render speed on AE 5.51 PB on 2.0 GHz P4M Toshiba Notebook, 512MB RAM, 4200 RPM Hard disk running XP SP1 with all background processes (anti-virus etc) left on. (1) Magic Bullet 1.1 50 minutes and 15 seconds or 8.8 seconds per frame (2) Magic Bullet 1.5 30 minutes and 19 seconds or 5.3 seconds per frame Just for kicks, running AE 6.0 Pro on my main machine, a custom-built P4 3.0GHz C (HT on) with 800FSB and 2 GB 400Mhz DRR ECC memory, 10,000 RPM RAID 0 SATA drives running XP SP1 with all background processes (anti-virus etc) left on. Secret cache is set to purge every 24 frames which slows down processing. (1) Magic Bullet 1.5 9 minutes and 13 seconds 1.6 seconds per frame My guess is a basic Magic Bullet process at 8pbc would run .5 or seconds per frame or faster with a higher secret cache. I'm very pleased. |
September 12th, 2003, 12:40 PM | #2 |
Warden
Join Date: Mar 2002
Location: Clearwater, FL
Posts: 8,287
|
That's great news Stephen. Thanks for taking the time to do the tests and posting them. Now all I have to do is wait for a Mac version.
__________________
Jeff Donald Carpe Diem Search DVinfo.net for quick answers | Where to Buy? From the best in the business: DVinfo.net sponsors |
September 14th, 2003, 08:09 AM | #3 |
Major Player
Join Date: Jan 2003
Location: Plainfield, New Jersey
Posts: 927
|
Stephen, did you notice Magic Bullet introducing any artifacts to your image?
|
September 17th, 2003, 10:41 AM | #4 |
Regular Crew
Join Date: Mar 2003
Location: Netherlands
Posts: 126
|
Well, I tested magic bullet 1.1 and 1.5 and I must say it's a very good de-interlacer (slow, but good)... I don't know about Stephens discoveries, but I can tell you Magic Bullet 1.1 and 1.5 have a big problem with small lines.
The edges of stones for example.. or the numbers on number plates. This only happens when there's very little to no motion present by the way. Examples van be found on my website: http://home.planet.nl/~snuve011/bram/de-interlaced/de-interlaced.html |
September 17th, 2003, 11:32 AM | #5 |
Space Hipster
Join Date: Oct 2001
Location: Greensboro, NC
Posts: 1,508
|
You need to add the DV Deartifacting (set to DV footage) plugin that comes with Magic Bullet to address the small line issue.
I come from using FilmFX and Reelsmart FieldsKit and Magic Bullet 1.5 beats both hands down, both speed and quality. Unfortunately, still frame comparisons won't tell you that much. You have to look at long clips as each package works on a wide variety of data and then assembles those frames together. Looking at clips will reveal far more than still frames. |
September 17th, 2003, 03:09 PM | #6 |
Regular Crew
Join Date: Mar 2003
Location: Netherlands
Posts: 126
|
Yeah, i've got the moving avi's here too, but there the tilting lines are even more annoying...
I'll try the de-artifacting feature of magic bullet and let you know what happens. (i'll check the clip, not the stillframes ok?) By the way, I thought de-artifacting was ment to get the chroma keys back (?) |
September 17th, 2003, 03:20 PM | #7 |
Space Hipster
Join Date: Oct 2001
Location: Greensboro, NC
Posts: 1,508
|
It's to remove block artifacts from DV footage. Also, I've done all my Magic Bullet at 16pbc which helps improve quality as well.
|
September 17th, 2003, 03:43 PM | #8 |
Regular Crew
Join Date: Mar 2003
Location: Netherlands
Posts: 126
|
Yes, you were right: De-artifacting helped although the jumping lines on certain patterns/details are still not completely gone.
|
September 17th, 2003, 03:48 PM | #9 |
Space Hipster
Join Date: Oct 2001
Location: Greensboro, NC
Posts: 1,508
|
I agree it's not perfect, but with deartifacting at 16pbc, it appears to my eye to give much better results than the other options available.
|
September 17th, 2003, 03:51 PM | #10 |
Regular Crew
Join Date: Mar 2003
Location: Netherlands
Posts: 126
|
I'm now doing the same thing, but in 16 bit and detail pattern set to 7 instead of 4
Lets see (uhh wait a looong time :] ) wait happens. |
September 19th, 2003, 01:05 PM | #11 |
Regular Crew
Join Date: Mar 2003
Location: Netherlands
Posts: 126
|
Ok final result:
The higher the detail pattern, the more likely magic bullet will choke on small details. A detail pattern of 1 gave me the best results. The de-artifactor ensures that no remaining artifacts show up. You can check my website for details |
September 19th, 2003, 01:11 PM | #12 |
Space Hipster
Join Date: Oct 2001
Location: Greensboro, NC
Posts: 1,508
|
Bram:
Thanks for the results and pics on your site. Very useful info. |
| ||||||
|
|