aluminumstudios (William Milberry)
I should be editing right now, but it's all Vic, Jeff, and dji's fault.
Remakes ... they just don't work like the originals
I've been trying to redo some of my old videos with DVD footage and update them a bit.
First I started with my Armitage III - Hey Man Nice Shot video. I recreated the original down to the frame (using better footage). Then I went in and modified all of the parts I didn't like. When I sat down and watched it I hated it.
The best way I can describe it was a bad video with good editing ... It just didn't jive. Also, when I watch the original I remember how much fun it was to make it. My first video in 1998. Everything was new and an adventure. That sense of nostalgia was lost on the remake ... so I deleted it!
Next remake - Perfect Blue / Every Breath. I personally think this is one of my best videos in terms of theme, story, editing, and emotion. It's not my personal favorite of mine ... but I believe it to be the best in terms of composition. Anyway, the captures I made it from weren't very good. So I redid it with DVD footage, and behold I didn't like the remake at all! But this time it was something different. The remake was too clean. I discovered that the washed out color, the fuzziness and noise really lent to the mood of the original. It created atmosphere which the remake lacked. So I've been trying tons of filters (gaussian blur, radial blur, diffuse glow, film grain, etc.) on the new one to try and get it to feel like the old one.
As it stands right now there is a 50/50 chance I am going to delete the remake and be happy with the old one!
The only video I ever remade that I liked was Torn. I recreated the original video using better footage and it worked for that one. BTW - Torn was originally made from the caps I did for Every Breath.
Slow filters ... even on fast computers
I have two distinct sides - a creative one and a hard-core tech one.
This is my hard-core tech side talking.
I had a PIII 500Mhz (overclocked a bit to 585MHz) and certain filters such as radial blur, gaussian blur, TMPEG's noise filters, various smart smoothers, etc., were slow as heck. Using them often gave me a rendering rate of like 1 frame per second.
Now I have an Athlon XP 2000 (overclocked to a 2100 - 1733 MHz) and I get a rendering rate of 1.8 fps on similar operations! The same case is true with P4's (so AMD or Intel zeallots need not argue the strengths/weaknesses of their beloved CPUs.)
Why is this? Well, I know it's not my system because many other functions run at 4x what my PIII did. My Athlon rips through MPEG encoding when no fiters are used.
Why can't it's raw power give me more speed on these intensive filters? I think they use algorithms with many branches that have poor predictability and hence just don't run well on your traditional scalar processors.
When a program is running and it comes to a branch (an "if X then do Y" type statement, or a "case" statement), the CPU tries to predict which way the program will branch and starts executing that branch of code ahead of time. If it mispredicted, the instructions that were executed have to be thrown away and the CPU's pipeline is emptied of this incorrect code (because it has to go off and run the other branch now.)
I haven't examined any of these algorithms, but I suspect that radial blur and a number of noise filters may have many branches that are hard to predict. So the scalar CPUs in our systems make guesses which are often wrong which result is lost CPU cycles executing instructions that get thrown out.
I read that the Itanium executes both branches and just discards the wrong one. That is a pretty phat chip to do that. If my theory on this is correct, I suspect that the Itanium would run radial blur filters and the link much faster than current x86 CPUs.
I haven't dont any research to substantiate this ... I'm just theorizing based on what I know of cpu architecture and what I suspect is the case with some of these algorithms.
Wow, I made a journal entry.
I've been so out of this week since AWA. The real world is such a let down. So I've been sitting here at work doing basically nothing ... I'm runing out of nothing to do so I decided to make a journal entry.
I'm currently trying to figure out what I want to do for my next project! I'm 1/3 done with a nifty action video dedicated to my favorite type of anime characters ... but now I'm considering adding a lot of compositing and effects to it. It's the type of video that would support it well. I kind of want to make a massive multi-series video (conventionally edited) and a video with tons of stills and compositing ... just to be artistic. Unfortunatly, I don't have as much time as I used to since I'm taking Japanese this semester.
Japanese is a lot of fun, not to hard, but a lot of work. I have to study every night or get behind. So far I'm getting an A. I'm finding it's easier being a student now that I'm older. I'm in class because I WANT to be.
Back to fighting bioterrorists
Well, so many people talk about animemusicvideos.org I figured I would do a journal entry for the heck of it.
I'm back from AnimeUSA. It was a small con but I had a good time hanging out with the other video editors (who I also saw at AWA and Otakon) and seeing the Cowboy Bebop movie Knocking on Heaven's Door especially made it worth it. I used to go to one con a year, now I've gone to 3 in 3 months and been invovled with panels and such at all three of them.
I have a ton of shit on my hard drive and I come back from every con more confused about what I want to devote my time to. There are some damned good new videos showing up. I especially like Hsein's FLCL video and this one Parasite Eve video (I don't remember who did it off hand.) I have been noticing a new generation of fast, extremely cleanly edited videos and I would kind of like to do one in that vein. At the same time I have ideas for a really good drama and a off-beat comedy. ARGGHHH, there aren't enough hours in the day! All of these ideas are pretty ambitious and would take a long time to do.
I've always tried to do my own thing, but maybe I'll put a poll on my site to see which video I should work on next since I can't seem to decide!
Love and Peace!
Current server time: Sep 17, 2014 07:40:27