Instrumental Anime Project

[rose4emily]

I will be offering a preview of each section before the public release, through a web folder I haven't set up yet - so, if there are any glaring errors, I'll be able to fix them. Just be forewarned that such fixes would probably take a while with classes coming up fast, given the amount of time it takes to do this compilation work. I am keeping lossless archives, though, so it will be possible to make changes if necessary.

I'll be doing multiple encodes for distribution when I finish setting everything up for encoding. There'll be an MPEG1, MPEG2, and MPEG4 encoding if all goes as planned, each of them in a high-bitrate range that I haven't fully decided on. I believe I'll also try a lower bitrate MPEG4 encoding at about the quality used for most DivX DVD rips, which won't be as pretty, but should be easier to download. Time allowing, I'll also do a rescaled MPEG2 encoding for Bakadeshi. Time not allowing, I'll save that for my next vacation in November. I'll also try using my PNG-based lossless codec when I get a chance to finish that, using the lossless archives as a source.

I'll check my e-mail for Songbird's pictures.

I'll be mirroring the project through an "indirect link" from the .org to my site, and should be able to provide it at speeds comperable to the .org (still might take a while for people to download, though, where it's such a long project).

What I'd really like to know is how I could go about setting up a bittorrent tracker for this thing, letting it be a controlled P2P distribution for people desiring fast downloads.

I'd also like to know if there's some way to submit the project under all of our accounts, or if we'll just have to pick one (or make one to represent the group). We could, if need be, create a user called "instrumentality" to cover this project and any that may or may not follow, and link that account to all of ours for the project credits.

[Kalium]

I'd also like to know if there's some way to submit the project under all of our accounts, or if we'll just have to pick one (or make one to represent the group). We could, if need be, create a user called "instrumentality" to cover this project and any that may or may not follow, and link that account to all of ours for the project credits.

There's a collaborators thing. It's one of the steps in entering a video. Step six, and you have to add each name individually.
(Yes, I have been making a habit of keeping an eye on this thread)

[rose4emily]

I did some image processing to turn my drawings into frames that I think would look good in the "intermission" / fullscreen section credits.

Basically, I used a semi-random mathematical pattern generator and some color manipulation to create a background resembling old-fashioned paper or an aged goatskin drumhead (take your pick), then ran the original drawings through another color transform to give them a sort of sepia look, and finally set the alpha channel of the drawings to the inverse of their luma channel to composite them onto the "aged paper" background.

To see them all, go here (and please tell me whether or not the images took forever to load):

http://www.thewired.info/instrumentalit ... ssion.html

I still have to add the credits text, but this should give you an idea of what I'm doing. I've also decided against using the Monty Python intermission theme for the audio, as it would just seem too silly after something like "Forbidden Memories". Instead, I have chosen "Barefoot in the Park" from the Evangelion OST. If you are curious as to which song "Barfoot in the Park" is, get it here to hear for yourself.

http://www.thewired.info/instrumentalit ... ission.mp3

It might seem a bit long at first, but I think it's just right at about 2 min, 30 sec for the 20 images and the credits for six videos and the related project details.

I'm also putting compressed versions of the narrative audio here:

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

I dunno. The static is pretty bad (think of the Miasato voiceover in ReDeath). That, and I sound something like a speak-and-spell trying to be a hypnotist when I try to read and speak casually at the same time. It just doesn't work that way. Hence my posting the narratives - so you can decide whether it'd be a good idea for someone else to try making them.

That, and I didn't have the final script for Warrior's Dance, so I did a heavily modified version of the draft I had at the time.

For right now, I'll do a test run of one of the videos to find any technical issues I may run into adding the narrative and bumper, but I'm not sure if my narritives are recorded at high enough a quality to belong on this project, so I'm holding back a bit on the full compilation.

[downwithpants]

your old-paper looks really nice. very nice job creating it (not to mention the well-drawn sketches). 2:30 does seem kind of long for the credits though (~20 seconds per video?) are you going to do scrolling credits or fading? but i guess that's long enough as an intermission for a pee break. maybe a countdown timer, if you can make one, would be considerate for people who don't need a break/sports fans.
if you have avisynth, this will make a countdown from 2:30

Code: Select all

blackness(length=150*24, width=512, height=384, fps=24, color=$000000)
showsmpte(fps=24)
reverse()

the intermission song is a good choice, more appropriate and less annoying then the monty python intermission song.

the static is quite noticeable in the narratives. i tried using the noise removal in audacity. it removed static well, but your voice came out sounding too bass-heavy and slightly metallic. which noise remover did you use? try audacity's noise removal on the un-equalized audio or upload the un-equalized audio wavs and let me try. http://audacity.sourceforge.net/

your recitation is pretty good, better than I could do. although you could show a bit more enthusiasm i guess- the narrarator in fantasia was pretty enthusiastic. i guess my best tip would be to think of how excited you are this project is getting produced! i think we could use some narratives from songbird or jasper-isis as well, and alternate between your narratives and theirs for some variety.

[rose4emily]

Chris:

Length of "intermission":

It's not just credits for the videos, it's also a summary of many of the project credits, as I'd like to have them placed in both files in case some people end up having one and not the other. I figured on showing the credits for each video one per background image, at about 6 seconds apiece, with a the rest of the information on the remaining background images, with a few plain images used as spacers if there is room left over.

Noise removal:

I recorded and edited in Audacity, and the noise removal I tried was Audacity's build-in filter. Trust me when I say it sounded even worse when I tried it on the un-equalized verson. As to the "bassiness", that's actually what you'd get were you in a room with me (not something I'm terribly fond of, makes me difficult to understand if I'm not also speaking rather slowly loudly) - I only applied a heavy roll-off above 4KHz, well beyond the range of the vowel sounds associated with the pitch and resonance of a human voice. This did remove the high-end noise, and helped compensate for the fact that an electret lavalier mic (which is all I've got to work with right now) is generally biased toward the high end in the first place. Maybe I should try rolling of some of the band under 200-300Hz to remove some of the "boomyness".

Enthusiasm:

So, then, I should go with the "presentation voicing"? At least that'd sound like natural public oration rather than an unusually flat attempt at speaking. It might also help with the noise thing a bit by boosting my acoustic SPL at the time of recording. Either that, or try to memorize the speaches so I don't have to read them at the same time.

Conclusion:

Maybe I should try to find someone here with a better mic (the good ones tend to be pretty expensive, so I can't just go out and buy one right now) and record these things again in more of an oration styling. I'd also like to hear one of Helen's or Wendy's versions. They might be able to make a much cleaner and clearer version of the narratives than the one I've produced.

[NeoQuixotic]

<b>rose4emily:</b> My name is Stein as in "Einstein" and berg as in "iceberg". Steinberg, I think it means stone mountain in Dannish or German, I don't really remember which one. For my monologue you can use this, unless you feel like changing it more

End of All Eva blends two outstanding works of art to tell a story that's beautiful both visually and musically. The visual portion is excerpts from the critically acclaimed Neon Genesis Evangelion anime series. Studio Gainax, along with Hideaki Anno, created one of the best anime series ever conceived. Evangelion showcases biblical prophecies, heart stopping drama, and collosal mech battles. This video is a short depiction of the grand story of Evangelion. Originally, the music considered for End of All Eva was a piece composed by John Williams. However, John Williams' music was too recognizable and the full effect of the video would be lost. The search for musical accompaniment was finally settled with "The End of All Things", a selection composed by Howard Shore, from the Lord of the Rings: Return of the King soundtrack.

Not only did End of All Eva combine these two already excellent artworks to create a new piece, but it also showcased and paid tribute to their magnificent achievements. End of All Eva was created out of respect and admiration for all forms of art, and all who push themselves to present their ideas and emotions any way they can. As Orson Welles once said, "Create your own visual style... let it be unique for yourself and yet identifiable for others." Please watch and enjoy this mixture of Evangelion and Lord of the Rings.

[rose4emily]

I've just finished my technical testing of codec settings and the AVIMerge application I mentioned earlier, and here is an example merge of "Ararat" and "Forbidden Memories":

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

I have not found a way to create an MPEG1 file (I can do MPEG1 in an AVI - but it doesn't make much sense to do so), and the MPEG2 encodings weren't looking as good as the MPEG4 (might be the MPEG2 implementation I have, I've heard there's a lot of variation), so I think I'll stick with just the MPEG4. If Bakadeshi needs lossless, I'll make a set of matching Huffmans for him.

I decided to do "Ararat" and "Forbidden Memories" for the test to find the settings that would look best with both super-clean flat-shaded footage (as was the case for "Forbidden Memories") and already-compressed, highly detailed footage (as was the case for "Ararat"). In the end, I found that applying a wide-radius selective gaussian blur, followed by a narrow-radius normal gaussian blur (these reduced existing compression noice and made the flat-shaded areas "flatter" (even DVDs are somewhat lossy)), and setting the encoding to a constant "q" of 2, with a four-second keyframe interval and the mpeg (as opposed to the h.263) quantizer produced results of a quality very near the originals.

The blur filters are also applied before encoding, so I can elect not to use them on a video that looks too fuzzy with them - if I run into that problem. Everything else has to remain the same for all segments for the AVI join to work.

At 90MB for 7 minutes and 22 seconds of video, the file's somewhat on the largish side, but the overall project should still, at those settings fit comfortably onto two CDs - one for each section - using this encoding. Also keep in mind the fact that the mostly static credits and narratives will compress better than the videos.

I tried lower quality/bitrate settings, but wasn't satisfied with them, and wouldn't advise them for the "official" distribution. CDs can be offered, even if the DVDs don't come around any time soon, for those without bandwidth. I'm also seriously considering the possibility of running a bittorrent tracker for this project off of my server (at least if I can limit the tracker to just the content I chose to track - I don't want to become a hub for cheep people trading "warez").

Of course, the "warez" traders are idiots in the first place - 90% of what they are ripping off has a free and legal equivalent in the open-source world. Once that's out of the way, it's no hardship to pay for the remaining 10%. Examine a copy of MS Office. Whose company is named on the box? Give to Gates what belongs to Gates and to everyone else, well, W.W.J.D? .

Anyhow, check the test merge, and give your opinions and suggestions concerning the ecoding quality. Otohiko and Jasper-Isis can also compare the encoding directly to the files it was encoded from, and should be able to give very well-informed feedback.

And yes, I'm aware that none of the in-between stuff is, well, in-between the two videos. That's just because I haven't figured out what I'm going to do about the narrative audio yet.

[TriGGiT]

how is this doing, i so much want to see it i mean with all those views and replies surely something tasty has got to be cooking here heheh.
that dvdbox cover looked very professional.
*promisess hell be patient*

[Otohiko]

how is this doing, i so much want to see it i mean with all those views and replies surely something tasty has got to be cooking here heheh.
that dvdbox cover looked very professional.
*promisess hell be patient*

Well, as you can see, we're talking about the final touches here - the vids are done. rose4emily is now making intros to videos (to make them sort of like Disney's fantasia, in that sense), intermission stuff, and getting closer to compressing the first final section of the project.

***

rose4emily: good to hear the compression is working out. I've got nothing against the blur filters, etc., so long as you manage to get it to look good. I had to do some tweaking to get 'Ararat' resized properly (without creating some nasty-looking deinterlacing problems - looks like IVTC didn't completely work as I hoped), but that's probably just me and my poor technical literacy...

Anyway, a CD per section is more than reasonable. BT tracker? It could work, I guess it depends on how many people try to get it. I hear AniMix has tried this sort of distribution, maybe it would be a good idea to ask them how it worked out.

[Bakadeshi [AuN Studios]]

The sample looks good. but... I think we'll have to get the compression size down considerably if we expect people to even bother downloading it, for the online distribution version. I'd say close to the size of a fansub episode should be good. 200-300 megs for the HQ online version, and a smaller 100meg version for the dounut. (if it'll even look anywhere descent at that size...) Perhaps we can even split up the widescreen and fullscreen segments to 2 diferent distributions for the dounut, so the entire project would equal 200 megs. Unless we can get the mods to grant us extra space than the preset 100 for most videos.