JOURNAL: Melanchthon (S B)

next page
  • Worst workflow ever. 2012-10-09 13:31:55 1. Rip DVDs.
    2. Make d2vs of each whole disc, because the VOBs aren't split where the episodes are.
    3. Make shitty mjpegs from d2vs.
    4. Watch Magix choke and die on said mjpegs.

    Now at this point I really should have gone back to the original d2v-reading Avisynth script and added some #Trim() lines, decommenting each one in turn to export one avi for each episode. But no!

    5. Use Virtualdub to export sections of video. I cut to a pattern, but wouldn't trust myself to replicate it, so let's just say the exported sections started and ended at random frames.
    6. Start editing.
    7. Start work on a particular 4-second section. Find it impossible to time because the processor can't keep up with this particular set of video objects for some reason.
    8. Realise that I compressed the video way, way too much and that the computer can't unpick all that compression. I need to make some mjpegs that can be crunched on and spat out faster than 23.97fps.
    9. Realise that the only way of frame-matching with 100% certainty is to recompress the shitty mjpegs. Fuck!
    10. Recompress the shitty mjpegs.
    11. Post a journal entry about it instead of getting back to work. 
  • Editing haaaeeeet 2012-09-27 23:39:48 This time I really am making a Trigun AMV, and I'm not spending four months trying to bash the footage into shape either (plus, it's been six years and filters have improved since then, so I need to start over tech-wise). Somehow I managed to mess up even the simple process of changing the original framerate to the one I'll be using -- the avis are in 23.976, but my video editor can't handle that and is apparently working in 23.98 instead. Not a big deal, although sometimes on playback cuts made at a scene change show a frame from the scene they're next to.

    I'm going to use a weird hybrid method to get the clips. First I'll use the low-quality files of the whole series to make the video itself (sans any complex or picture-dependent effects), then I'll go back to the VOBs and work on only the portions that are going to get used. I'm not tidying up 26 episodes of horrible crap for a video that's going to use less than half an episode's worth of video. The clips will have to be totally replaced anyway -- material with blends moves differently to material with clean frames, and I use clip movement for synching purposes.

    So anyway, as the title suggests, onto the editing haaaeeeet. I've reached the point where I've put down all the neat clips with the cool visuals, and they're just sitting there waiting to be expanded on. The clips that expand on them are generally missing that certain visual something because they need to serve both the video and those initial clips, but without more of a video to compare them with I can't tell whether this is fine or not. It's very annoying.

    For instance, there's this one part near the beginning where the song starts kicking off, and that's matched with a nice blue sky to contrast with the less colourful footage at the start of the song. Problem is, I can't think of anything to go with it because that section of the timeline is otherwise blank, and it's blank because I'm still working on the high points of the song that get used as a reference for everything else. How nice it would be to be able to simply go through a song start to finish.

    The second annoying thing is that the lyric synch is... incidental? The lyrics work for Trigun, but on the whole they're pretty vague. The most that can be grabbed onto is "this chance to drop", which appears twice and actually already has been grabbed onto. One of those is in the long middle section that's planned to alternate between Episodes 17 and 24 instead of grabbing whatever random clip goes with its predecessor.

    I guess I could place more sweeping cuts too. 13 cuts in 8.5 seconds is great to watch, but not so much fun to edit, and especially not so much fun to delete if it turns out to not look as good later on as it did when you finished work on it for the first time. But I think I will be keeping that part.

    The third annoying thing is that even though I don't get much done sitting with the MEP window open, I get nothing at all chatting to people from the org, and I like doing that. :/ (It occurs to me that typing this isn't exactly achieving much in the way of editing, but it could be used for reference later.)

    If you want to see an example of clip movement used for synching, check out http://www.animemusicvideos.org/members/members_videoinfo.php?v=181187, most specifically the rising 'its' at 0.25 and the part from 1.19 to ~1.44. Trigun doesn't really have the shinies to do that, so I guess enjoy your six minutes of scenery and child abuse. At least it's not an MMV? 
  • Video pimpage, GO! 2006-11-09 19:42:34 http://www.animemusicvideos.org/members/members_videoinfo.php?v=132643

    I made up half of that as I want along, heh. But I do think it's funny that Episode 10 set up some kinky games between Ranka and the boys, and it turns out that they're all going to follow Haruhi and then Kyouya's all annoyed. It's lucky for the world that he's so self-contained and controlled, because that guy gives off major serial killer vibes. 
  • Bollocks. 2006-11-09 00:20:39 I compressed my video, checked it, uploaded it, checked it for the last time... and noticed that I'd screwed up on a bit of masking. >.<

    So now the vid info is sitting there all alone. :/ 
  • Fuck Trigun. Fuck it right in the the ear. 2006-11-07 22:17:57 I found something else to do.

    I can't believe I went from watching my first episode of a series to complete video in six weeks. Still, it's done and I can add 'working with subs and raws' (hey, the series only finished airing in September) and 'lipsynch' to the fictional list of things I've done with AMVs.

    Boring crap that would send people to sleep if I put it on the forums:

    Video breakdown -- the vast majority of the footage is from Solar / Lunar's subs, which are nice and bright and clean but suffer from a touch of mosquito noise and occasional blocking. Of that footage, most of it didn't need desubbing and a good portion of the rest of it could be done in MSPaint by cloning clean pixels. One episode couldn't be done manually and it couldn't be cropped either, and so began my descent into raw-downloading madness. Probably the most annoying part of that was altering the brightness and contrast and such so that the clips all matched. Also, the raw clips are a little on the soft side and don't like being filtered. Well, tough, you're outnumbered.

    Lip-synch -- It's both easier and harder than I expected. 8fps and three positions was good enough for the animators, and it's good enough for me. If the mouth movements are suited to the lyrics, then you can roughly approximate the rhythm of the speech / singing and it passes without notice. If they aren't, you can arse about as much as you like and it still looks like you slapped the clip on the timeline and left it at that. By far the easiest method of lipsynch is to let the original clip do the work, stepping in only for long notes or to alter that one odd mouth position that throws the whole phrase off.

    Odd little things I've found:

    1. Sibilant sounds are almost impossible to synch accurately. Whether it's because the mouth shapes I had weren't right or because they fell where I needed some closed-mouth frames to keep up the illusion of movement is something I still don't know.

    2. A well-placed reaction shot is your friend for life. The prevailing forum wisdom appears to be that the camera should be on the speaker as much as possible, but I think there needs to be more natural interaction between characters in lipsynch vids; more of a sense that the other characters are present even though they're off-screen, or even off-episode.

    3. Using the movement in the clip instead of slaving over each syllable leaves more of a margin for error when it comes to motion lip-synch. If the lipflap is good enough and no better, the slight inaccuracies in the movement you can't control won't stand out... hopefully.

    4. blur.sharpen = quick antialiasing. I hate being able to see the joins in composited shots. It's a pain, but it's worth it when I see the elements of a frame locked firmly together. :/

    5. Sparkly pink WAFFy shounen-ai shoujo provides ample opportunities for lyric synch involving descriptions of graphic violence, even the sort that involves penises.

    6. Temporal smoothing increases the size of the filtered, compression-ready file. wtf? 
next page
Current server time: Jul 26, 2025 20:02:35