Instrumental Anime Project

Discussion & organization of Multi-Editor Projects
Locked
User avatar
rose4emily
Joined: Fri Jan 23, 2004 1:36 am
Location: Rochester, NY
Contact:
Org Profile

Post by rose4emily » Sun Sep 05, 2004 2:21 pm

The plan is to offer the widescreen and fullscreen versions as separate files, largely because encoding widescreen videos as fullscreen videos wastes a lot of bits and tends to make the upper and lower edged of the video look like crud.

I did a complete merge of all of the fullscreen section videos, in the same encoding, and it came out to 256MB (nice round number - at least if you think in hex). Add the narratives and credits, and I expect it to be about 300-350 MB.

As I said before, the lower bitrate encodings weren't terribly pretty - big, ugly macroblocks and jittery motion, especially on panning sections. At just over 10 MB a minute (it's just over 22 minutes long without credits and narratives), I don't think this bitrate is too bad. The widescreen version is going to be longer (at about 37 minutes - damn me and my freakishly long video), but will also be spatially smaller (512x288 vs 512x384), so it should still probably encode somewhere in the 400-450 MB range with credits and narratives. Small enough for a CD.

As to not making both "high" and "low" quality encodings - I don't want to misrepresent the work throuh the "low" quality encoding, which would probably proliferate more quickly than the larger file.

It looks as though I will be able to use my server as a bittorrent tracker, using CBTT, and can augment my limited uploading power by letting others who've downloaded Animasia help with the distribution. If ever there was a legitimate use for p2p file sharing technology, I think I'm staring it in the face.

If I can find a print shop in the area that can do low-volume printings of quality CD and CD case labels, I might offer CDs through the mail. I'm not sure about how to handle the expenses, though. Maybe PayPal with an option of cash-and-good-faith for people who (like me) don't want to mess with buying stuff through PayPal? Then there's the time factor. Lots of variables to think about on that one.

Okay, I still have to encode all of the titles and make a second attempt at the narrative speaches, so I'll be going now.

And classes start tomorrow...
may seeds of dreams fall from my hands -
and by yours be pressed into the ground.

User avatar
jasper-isis
P. Y. T.
Joined: Tue Aug 13, 2002 11:02 am
Status: catching all the lights
Org Profile

Post by jasper-isis » Sun Sep 05, 2004 8:30 pm

You have classes on Labor Day? Bummer...

The encoding looks all right, though it's a little soft for my tastes. (I usually assiciate soft video quality with downloaded footage; the fact that Forbidden Memories uses slightly under-saturated color doesn't really help either.) However, if the blurring makes the other segments look nice, then I am willing to sacrifice some fidelity from my source footage. (Learned that usage of "fidelity" from DWP :wink: )

There are two main ways to crunch down on video size for disro: decrease resolution, or decrease bitrate. Personally, I'd rather have the online distribution be of a smaller resolution (say, 320x240) than of lower quality.

The CD distribution can be of whatever resolution we want. I say don't bother crunching the bitrates too hard for the CD's -- we can afford to use more space. But don't encode the bitrate so high that somebody with a 90's computer wouldn't be able to play the project. (There's nothing that I hate more than "padded" video files... ugh.)
Image

User avatar
Otohiko
Joined: Mon May 05, 2003 8:32 pm
Org Profile

Post by Otohiko » Sun Sep 05, 2004 8:40 pm

Yea, speaking of which - have you tried using a lower resolution? Maybe that would solve some issues for online distro; and it might not be as much of a difference since you're using some blurring anyway. Something like 384x288 or 448x336 (I think those would be the sizes), or something along those lines? If it saves, say, 100 megs off the final version, it might well be helpful in getting the project better-distributed.

But, again, I'm not the one presently working on compressing/distributing... but try and see if something like that works, if possible.
The Birds are using humanity in order to throw something terrifying at this green pig. And then what happens to us all later, that’s simply not important to them…

User avatar
Bakadeshi [AuN Studios]
Joined: Wed Mar 24, 2004 7:59 pm
Location: Georgia / S. FL WIP: ROS2, VG3, AR2
Contact:
Org Profile

Post by Bakadeshi [AuN Studios] » Mon Sep 06, 2004 12:55 am

yea I agree to try the smaller size. also did you try working with the filters that reduce noise spacially and temporally as mentioned in the guides? those can significantly reduce file size for mpeg4 compression. I just don't know if AVIsynth or an effective alternate with these filters is available on linux.
[size=0]
Image
Image[/size]
Recommended Underated video (Not Mine): Jasper-Isis - Ever Searching

User avatar
rose4emily
Joined: Fri Jan 23, 2004 1:36 am
Location: Rochester, NY
Contact:
Org Profile

Post by rose4emily » Mon Sep 06, 2004 2:44 am

Denoising filters: Yes, I used those on some of the segments. Not all of them, as they seemed to help in places and hurt in others. I tested each with a bunch of different pre-processing filters and just picked the best looking one - which was usually the smallest one as well, being the one with the least "noise" for the encoder to deal with.

"Soft" Footage: Even at the relatively high bitrate I've used, the unsoftened footage had much more noticable spatial artifacts around sharp lines and edges. "Forbidden Memories" was actually the video I used as the primary point of reference for my encoding decisions, as it was so clean in the first place and therefore easy to detect the degree of image degradation present for various options. The people who make the downloaded footage soften it for a reason - it reduces the "noisy" spacial loss with a more visually acceptable form of loss - namely the blurring. The same thing is also commonly done for images encoded in the JPEG format on the web. The DCT (Discrete Cosine Transform) used by the JPEG image format, and all of the MPEG video codecs (along with almost all other lossy codecs for audio and video) has a hard time representing sharp edges, and creates "ripples" around those edges due to it's inability to accurately reprent a discontinuous function. All digital signals are discontinuous, but some (such as flat-shaded animation) contains much larger discontinuities that than others and thus causes greater errors in the DCT. By reducing the discontinuity (blurring the picture), the DCT can more accurately represent the resulting image, and do so with fewer bits - as it seems more natural.

On reducing the video resolution for the "Local" distribution:

The DCT comes into play again here. Signals containing a greater proportion of high-frequency components, especially non-regular high-frequency components, are harder to represent through the DCT because they contain, on average, more "information" in any given subset of their length.

Think about how you would, if you were recording from a record onto a tape, make the song shorter so it would fit on less tape. You'd speed up the record so it would finish playing sooner. But this changes the sound coming out of the record. The song is not only shorter, it is played at a higher pitch - a higher frequency. Now think about video in the same way, as a one dimensional stream (I know, it seems weird, but that's pretty much how it works on television and most magnetic media - hence "interlacing" and "fields" vs. "frames"). How would you make it fill less space? By speeding it up so it would form a shorter stream. In digital terms, you'd scale the image to a smaller size so it would contain fewer pixels. Problem is, the analog representation of the image is now at a higher frequency, and the DCT essentially acts on that analog representation rather than just performing arithmetic and lookup operations on the digital pixels (as would be the case for most lossless codecs). The higher frequency image, because it contains a greater density (higher frequency) of edges and introduces greater discontinuities to even soft gradients, wil not encode as cleanly or efficiently as the larger, lower frequency image. Even 512x288 is considered smallish for lossy digital encoding from a professional standpoint. Going farther below that will cause further diminishing returns in bitrate reduction.

All that having been said, I'll try it. I'm just explaining why I think a 60% reduction in pixels (about that achieved by going from 512x384 to 320x240) will yield nowhere near a 60% reduction in file size at the same settings, but will either increase the noticability of the edge-ripple effect of DCT-based compression or the softening required to alleviate said effect.

~~~

I did a test on encoding the 16:9 section this evening, at it actually compressed into about the same number of bytes (255 vs. 256) as the 4:3 section. Half of Pen-Pen's video is missing (am I going to see the full one anytime soon?), but everything else is there, so I think they'll actually be about the same size despite the difference in temporal length.

~~~

The test merges are available here:

http://www.thewired.info/instrumentality/

~~~

I still havent put the titles into the merges. I appearently have to create an audio clip (could be of silence, it just has to exist as a file to merge into the encoding) matching the length of the title segments before I can merge them. Am playing with other short, ambient clips to see if I can do better than silence - but I can always do the silence if it comes out that the silence would work best.

No further luck with the narratives yet. I'll try asking around this week to see if I can get access to a studio mike somewhere around here. I live in a community composed primarily of rich white suburbanites who own too much electronic equipment (well, I'm three out of four on that one. at least all my ridiculous collection of electronics was funded through my own employment and a couple of lucky discoveries at our local recycling drop-off - which has so far provided members of my family with a perfecty good evergreen bush, a self-shifting bicycle, a 600-watt home audio stack, and a working floor jack that should prove to be a big help with getting my father's truck onto blocks so he can fix its rusty undercarriage), so I have to imagine someone has some good audio recording toys lying around.

~~~

Visit your recycling centers from time to time. Suburbanites throw away a lot of really useful stuff all the time. The one in Dover even saw a new, working bicycle almost every day. Replace the punctured inner tubes or fix the derailed chain (yes, I've seen people throw away bicycles because the tires were flat or the chains derailes, or they never rode the things anymore it it didn't occur to them to put a "free" sign on the thing and drop it on their front lawn), and you have cheep, healthy, and environmentally friendly transportation. Of course, it helps to live on what used to be a farm and still produces enough organic waste (leaves, sod, weeds, branches - whatever won't fit into the home compost bin after the grass clippings - always keep the grass clippings, lots of nitrogen) to warrent weekly trips to empty the truck. Or a renovation and landscaping project that makes those trips happen about every third day for a few weeks on end. Then you'll have plenty of opportunity to treasure-hunt while dropping off branches and old vinyl flooring.

So long as that rusty 39-year-old Ford holds up, anyhow.

Imagine a full-size truck with a manual transmission, unpowered steering, no heat, no ac, leaky windows, holes rusted through the pickup bed, and an engine with the same horsepower as a 95 Taurus wagon - and you have "Fred" the truck. Yet it still gets complements from old guys whever it goes. I just think it's about time we get the windows and bed fixed.

~~~

That was one hell of a digression. I suppose that's what happens when you write at 3:40 in the morning. You ramble. I do that anyway, the insomniac hours just make it worse.

But, now I stop.

~~~

36 posts to go before we're at a four-digit count. :)
may seeds of dreams fall from my hands -
and by yours be pressed into the ground.

User avatar
rose4emily
Joined: Fri Jan 23, 2004 1:36 am
Location: Rochester, NY
Contact:
Org Profile

Post by rose4emily » Mon Sep 06, 2004 6:06 am

Now there's a poetic signature. :)

Odd... no one, to the best of my knowledge, has mentioned mine in this entire forum. I put those two lines in for my Senior yearbook quote and a dozen or so people felt the need to express their concern for my dismal outlook on life.

It would appear that none of those people were gardeners. 8)
may seeds of dreams fall from my hands -
and by yours be pressed into the ground.

User avatar
Bakadeshi [AuN Studios]
Joined: Wed Mar 24, 2004 7:59 pm
Location: Georgia / S. FL WIP: ROS2, VG3, AR2
Contact:
Org Profile

Post by Bakadeshi [AuN Studios] » Mon Sep 06, 2004 4:00 pm

rose4emily wrote:Now there's a poetic signature. :)

Odd... no one, to the best of my knowledge, has mentioned mine in this entire forum. I put those two lines in for my Senior yearbook quote and a dozen or so people felt the need to express their concern for my dismal outlook on life.

It would appear that none of those people were gardeners. 8)
I understood the meaning behind your sig perfectly 8)
[size=0]
Image
Image[/size]
Recommended Underated video (Not Mine): Jasper-Isis - Ever Searching

User avatar
pen-pen2002
Joined: Sun Sep 02, 2001 3:39 pm
Location: Grinnell, IA Procrastination Meter: Code Lemon-Lime
Org Profile

Post by pen-pen2002 » Mon Sep 06, 2004 4:08 pm

I always liked your signature, Rose4Emily, it doesn't seem dismal at all in my oppinion.

As far as my video goes, have you changed the password on the server? I can connect but can't get in.

I don't know how well mine will do with a smothing filter as it is pretty detailed animation most of the time, I actually ran a sharpening filter on the AWA compresion.

I downloaded the fullscreen test and the peice as a whole flowes very well, the order was perfect. The quality was not as good as I hoped, however. The size of the final product is a tricky thing. I think that dial-up users are not going to be able to get it regardless of what we do, so we should not worry about them (aside from making a snazy DVD) and concentrate on broadband users. I think we should have no trouble getting extra space considering the fact that there are about 5 full videos. Also the doughnut is expanding both in storage and bandwidth. I think we should concentrate on making it look good.
Image

User avatar
rose4emily
Joined: Fri Jan 23, 2004 1:36 am
Location: Rochester, NY
Contact:
Org Profile

Post by rose4emily » Mon Sep 06, 2004 8:46 pm

Unfortunately, I wasn't able to get video quality any better than what I acheived. I set the encoding at the best safe quantizer setting and tried a lot of different options with the filtering, and it looks like this is the best we can acheive. This, again, might be a side effect of the smallish resolution chosen for the project and all of the frequency/complexity issues that come with it, along with the fact that I had to use the exact same encoder settings for all of the videos, despite their great visual differences from one another. The good news is that the filtering actually made a couple of the videos look, in my opinion, better in the new encoding than in their original one. I only wish the same could be said for some of the super-clean submissions like "Requiem for a Nightmare" and "Forbidden Memories".

In my opinion, the only proper codec for traditional 2D animation would be one based on translatable and compositable cells that would be re-rendered onto the screen at playback time (or into a playable file for playback on slower computers). This would produce the same or better file size savings by intelligently removing redundancy without introducing loss. It would also have the added benefit of splitting the video into logical components identical to those that would be used for creating and editing that video. The description format used for editing and the distribution format used for playback would be the same.

In a further stroke of misdirected genius, I came up with the idea of "compiling" a video into its own rendering code in CS this morning. The player would just be a canvas into which the rendered raster would be fed. The file could then be raster, vector, or any mix of raster and vector, and specify its own raster data and drawing code. This would allow for almost infinate extensibility of the file format - the file could do anything the editing software and editor can think of for it to do, so long as the code remains within portable standards (Java would be great for this, but a C-based version might provide better performance at the cost of having to be "compiled" after distribution by the player on the viewer's machine). You could create your own filters and translations, use your own image formats, and generally optimize the file to your own video. To make this all practical, this power should be abstracted by the editing software, so as to provide a "safe" and simple environment to work in for common video tasks, saving any actual "coding" to those who need something really different that the editing software doesn't already provide. Strangely enough, codecs are usually much smaller than most videos, so including the rendering code in the video itself should consume much less space than it is capable of saving.

I think I've found the ingredient for making my media editing suite unique. I'll just have to remember to let it render to normal codecs as well for playback on the machines of people without the player client.

~~~

Pen-Pen: FTP - ?

The process is up, and I can get to it from my other computer. It should work on the outside as well. I haven't changed any passwords - it's still "instrumentality" and "firstchildren".

Are you getting a login screen, or a welcome message, or just something saying the address was found? Most clients I've seen display a little "Welcome to the Wired" message on login to the FTP server.

If you can't get in, and I can't figure out what's going on, is there any chance you can post it on your side and have me download it from you? It'd be the same transfer, just a pull on my end rather than a push from yours.

~~~

I'm currently looking for a development team here in Rochester to help me get the media studio project rolling, as I have a lot of ideas, but not quite enough knowledge to implement all of them at this point. If any of you want in, now's the time to say something. Now, or at whatever point in time I feel the source is ready to be opened and published. I just don't want to make a public release until I have at least the core functionality firmly in place, lest this thing end up looking like yet one more "SourceForget" project or, worse, fork off in a half-dozen incompatable directions.
may seeds of dreams fall from my hands -
and by yours be pressed into the ground.

User avatar
Bakadeshi [AuN Studios]
Joined: Wed Mar 24, 2004 7:59 pm
Location: Georgia / S. FL WIP: ROS2, VG3, AR2
Contact:
Org Profile

Post by Bakadeshi [AuN Studios] » Tue Sep 07, 2004 4:38 pm

heh pretty soon Jasper is gonna run out of space in her sig... if this project doesn't release soon ;p
[size=0]
Image
Image[/size]
Recommended Underated video (Not Mine): Jasper-Isis - Ever Searching

Locked

Return to “Multi-Editor Projects”