Instrumental Anime Project

pen-pen2002 · Post by **pen-pen2002** » Tue Dec 21, 2004 10:50 pm

Any word on uploading to the dounut vs other distribution methods?

Also are we going to be able to have it under one name in the database or are we going to have to have two listings, one for each section?

downwithpants · Post by **downwithpants** » Thu Dec 23, 2004 12:01 am

hey rose, can you apply a lightening filter or anything to my section of the video? looking at the project file, the middle sections of my video came out really dark, probably because of the softening you did to the video, which blended dark pixels over dark pixels. sorry, to point this out so late, i didn't notice it/pay attention to it before.

jasper-isis · Post by **jasper-isis** » Thu Dec 23, 2004 12:22 pm

rose4emily wrote:I meant to ask, what's your major? You seem really sharp with design, and I was wondering if you were somehow in the arts.

Nah, I'm merely a high school junior trying to stay afloat in one of the most hellish magnet schools of this country. Hopefully two years from now I'll be doing the same at JHU.

My major will probably end up being biology, though I'll probably double-major in the arts somewhere.

rose4emily · Post by **rose4emily** » Thu Dec 23, 2004 6:23 pm

downwithpants wrote:hey rose, can you apply a lightening filter or anything to my section of the video? looking at the project file, the middle sections of my video came out really dark, probably because of the softening you did to the video, which blended dark pixels over dark pixels. sorry, to point this out so late, i didn't notice it/pay attention to it before.

Actually, I did increase the contrast on your segment by just a little, as the re-encoding made the very dark (but not quite black) regions look like crud. Since the detail that was in those regions of the picture was more or less lost to the encoding artifacts over them, I tried bringing up the contrast a little bit to make those sections a true black. While this did exchange the intended "night-time" look for more of a sihouette effect, it also helped to bring out the foreground and the nicely drawn tie between the music and the movements of Kenshin and the other swordsmen.

Sadly, this isn't even a matter of your encoding - which was a very attractive Huffman. It seems to be more the result of the nature of the footage itself - and the fact that many MPEG4 encoders seem to discriminate against very light and very dark regions when deciding where to spend bits.

I actually did try running it through a double lighting filter - to reduce the "noise" in the background, but also cut down on the severity of the stark black backdrop produced by the first filter. The plain dark grey background this produced was a little "blah", however, so I thought the true-black version looked best.

Another thing I should bring up is the difference between monitors. PC monitors tend to have widely varying gamma curves, color saturation levels, and color balances. This often means that image and video regions near boundry conditions (very light, very dark, very saturated, pure reds, pastels...) look very different on different displays. For instance, LCDs tend to have problems representing all of the brigntness levels toward the darker end of the scale, greatly exaggerating blocking and banding in dark regions of a video. Many CRTs "bleed" color along scanlines, which can be highly visible with large regions of black/white contrast - as is seen in portions of 13'37". Many monitors also have gamma and brightness curves that are way off from where they should be, making dark regions look darker, bright (but still detailed) regions look "washed out", or blacks look more like bland greys. The display I used when deciding on encodings and filtering options is one with a very nice color response (especially for the consumer price range it falls within), and which I keep calibrated for proper color, gamma, and brightness response. Unfortunately, the display I'm working on now has pretty awful color/brightness response, and isn't adjustible. Consequently, any changes I make to the filtering or encoding settings at this point would probably make the picture look worse, rather than better - because I couldn't really see what I'm doing, especially in the case of the specific scene you're talking about.

---

Coincidentally, "Clair De Lune" is playing in the background of some parfume ad playing on the TV right now.

---

Jasper-Isis wrote:
rose4emily wrote: I meant to ask, what's you major? You seem really sharp with design, and I was wondering if you were somehow in the arts.
Nah I'm merely a high school junior trying to stay afloat in one of the most hellish magnet schools of this country. Hopefully two years from now I'll be doing the same at JHU. My major will probably end up being biology, though I'll probably double-major in the arts somewhere.

Y'know, this is probably one of the most intelligent groups of people I've ever been a part of. Bio at John Hopkins - you really must be as sharp as I've imagined. I never would have guessed you were still in high school, though. I don't know where I'm going with this - I guess I'm just very impressed.

---

pen-pen2000 wrote:Any word on uploading to the doughnut vs. other distribution methods?

Also, are we going to be able to have it under one name in the database, or are we going to have to have two listings, one for each section.

No word on uploading to the doughnut yet. I'm still waiting for AD to answer the PM I sent him about two weeks ago. I also check AIM every once in a while to see if he's on, but no luck with that yet (time-zone differnce probably doesn't help).

I am planning to upload the film to the doughnut when I am cleared to do so. I have to wait on the admins, even if I have both sections finished, though - as they will have to make special exceptions and probably take a few manual steps to get these files onto the server due to their large size.

I'm really hoping to get a yes on the "Instrumentality" account. Otherwise I'll be creating an "Instrumentality" studio, listing you all as collaborators, and offering each of you the option of "joining" the studio (however that's done - it's been a while since I've set up my account).

The film will be listed as two separate entries, plus a third entry for the intermission (which viewers may or may not want to see along with the two major sections of the film). The entries will be labelled as follows:

Animasia (Part 1 of 2)
Animasia (Part 2 of 2)
Animasia (Intermission)

Each segments' description will mention the other segments, give an overview of the project, and then discuss specifically the contents of that segment.

---

I'm still having problems with audio dropout in WMP. I've tried using four different tools to rebuild the AVI index, including VDubMod on Windows. I've extracted the entire audio stream to a separate WAV file, re-encoded the entire stream, and remuxed it back into the AVI. I've tried muxing the joined AVI into a n MKV, muxing the joined AVI and the outside, re-encoded (in both LAME MP3 and OGG) audio into an MKV container.

The resulting video file had the exact same audio problem, following each of these fixes. It also obviously still had all of the audio, because WMP was able to play it whenever I moved the shuttle a bit.

The one thing I have left to try is to encode every one of the files to a Huffman stream with WAV audio, join these files, and re-encode the monolithic result to the desired XviD encoding.

Some good news, however, is that installing he MKV pack needed to play MKV files in WMP was an incredibly simple and easily understood wizard installer, from .exe format. Because of this, I think I will try to distribute this in MKV format, making use of the chapter feature (to mark chapters at each of the segments), and OGG audio (to further improve the quality/bitrate ratio).

---

Due to the strange difficulties I've been having with merging the AVI files in a manner acceptable to the WMP, I need to take some time to clean out an entire disk partition before I can encode.

Good thing I brought a spindle of blank CDs with me.

rose4emily · Post by **rose4emily** » Thu Dec 23, 2004 10:52 pm

Very long accidental self-quote removed.

pen-pen2002 · Post by **pen-pen2002** » Fri Dec 24, 2004 1:53 am

That was the longest quote I've ever seen . . . And completly pointless.

badmartialarts · Post by **badmartialarts** » Fri Dec 24, 2004 8:42 am

rose4emily wrote:The one thing I have left to try is to encode every one of the files to a Huffman stream with WAV audio, join these files, and re-encode the monolithic result to the desired XviD encoding.

Not a memeber of yer project, but I wanted to comment and say I have yet to have success with split audio-video streams during encoding and since my first video (which had the problems you describe of cutting out on the audio) I have taken care to export the full video with audio and compress the video and audio at the same time. However, I have been told be....well, I can't remember who (maybe TsunamiJones) that the LAME MP3 encoder is actually to blame with this. I know for my first video I used LAME but for my subsequent videos I've used some hacked Fraunhoffer codec (Radium, I think) that so far hasn't given me trouble. I hate to advocate hack jobs, though...not sure what the cheapest Fraunhoffer codec packs run (not cheap, as I recall) but it might be something to look into.

Bakadeshi [AuN Studios] · Post by **Bakadeshi [AuN Studios]** » Fri Dec 24, 2004 5:11 pm

badmartialarts wrote:
rose4emily wrote:The one thing I have left to try is to encode every one of the files to a Huffman stream with WAV audio, join these files, and re-encode the monolithic result to the desired XviD encoding.
Not a memeber of yer project, but I wanted to comment and say I have yet to have success with split audio-video streams during encoding and since my first video (which had the problems you describe of cutting out on the audio) I have taken care to export the full video with audio and compress the video and audio at the same time. However, I have been told be....well, I can't remember who (maybe TsunamiJones) that the LAME MP3 encoder is actually to blame with this. I know for my first video I used LAME but for my subsequent videos I've used some hacked Fraunhoffer codec (Radium, I think) that so far hasn't given me trouble. I hate to advocate hack jobs, though...not sure what the cheapest Fraunhoffer codec packs run (not cheap, as I recall) but it might be something to look into.

I also had problems with Lame and joining the mp3s together. I think to correct this, I exported all of them seperate to wave, joined the wav files to one big wav, then used AVIsynth to join the video together, and encode it in Virtual dub, but using the one big Wav File as the audio source.

The problem seems, that you can;t join mp3s together with lame. You have to join the files together in Wav format first then encode the one big file with Lame. I think even re-exporting the one big file to wav (After it was already joined in mp3) and then recoding still won't fix it. My guess is maybe Lame screws up a bit in the stream where it joins it, throwing off WMP when it gets to that bit.

rose4emily · Post by **rose4emily** » Fri Dec 24, 2004 9:08 pm

Chris - regarding "Tomoe", I reduced the contrast increase. This will lighten the scenes you were saying looked too dark, but might make the dark scenes look a little less clean. Anyhow, this should take care of the lighting issue.

---

I tried the Huffman encodes, with PCM audio, and found that the files didn't join properly using AVIMerge. I also tried using AVIDemux (more or less a clone of VirtualDub), which encoded everything properly - except that the audio fell about two seconds out of synch after the intro.

I have gotten one thing to work - re-encoding the joined XviD file. This is good in that I have one more fallback, even after my original fallback plan (spending several hours and 30 additional gigabytes of disk space on Huffman encodes) failed. The problem is in adding yet one more generation to the process.

What I am going to try next is encode a constant Q=1 all I-frame version, join those pieces, and re-encode using MPayer. The generational loss wasn't too bad with the high bitrate, high-resolution settings I've been using, but I think it'd be best to get that first encode to be as high-bitrate as possible.

I'll also avoid filtering twice, as doing so would excessively soften the footage.

---

Looking at the post about LAME, however, I'll try one more attempt with joining the audio separately. I really would prefer to avoid any more encoding generations than are needed.

---

Oops, apparently the "Quote" button and the "Edit" button are two different things. And I just wanted to fix a couple of ambiguities created by misspellings.

NeoQuixotic · Post by **NeoQuixotic** » Sun Dec 26, 2004 9:28 pm

Ok, I got my computer up and running again! Thank goodness I had a backup of the huffy file of my video. So I took that and filtered it using convolution3d and mfToon (wish I did that for the org distro) and am waiting to render an mpeg2. I have TMPGEnc 3.0 XPress Trial version and am too lazy to find the render info on the thread (if it was stated). The video is the 720x480 direct from the Death and Rebirth and End of Eva DVDs. It is letterboxed so when it is displayed 4:3 the actual video within the black becomes 16:9. I did a test render set to 4:3 and it looked very nice. Just wondering if there is any specific settings I should know.

Oh yeah, I suppose that means I won't be mailing my submission for the DVD. Additionally, is there webspace to upload the mpeg2s yet and has anyone already done so? I would hate to be the last one