This post caused by recent events in my immediate surroundings.
- No matter what filters you used, you’ve overfiltered it. Almost all new encoders invariably make this mistake. Don’t be one of them.
- VFR isn’t as complicated as it seems and you would do well to learn how to use it to your advantage.
- Be very careful with frequency-domain smoothing filters, they kill a lot of background detail if you aren’t careful.
- Related to the above, STOP FUCKING BLURRING EVERYTHING. It doesn’t look good, it doesn’t solve any problems, the source probably isn’t that noisy to begin with. Don’t do it. Yes I know some people say “controlled blur isn’t a bad thing” or whatever the fuck, I don’t care, they’re dumb. It might have been true to a certain extent when everyone had to use XviD or Divx3 for everything but it sure as fuck isn’t anymore.
- Don’t forget your chapter points. They’re very convenient and are easy to include, and not including them makes people complain about you on the internet, and you wouldn’t want that to happen, would you?
- Slower doesn’t automatically mean better. (Yes, trellis=2 and me=esa users, this means you.)
- Don’t let people convince you to hardsub shit just because. Hardsubbing is gay, doing it makes you look dumb.
- Related to the above: karaoke, credits translations, fansub credits, “ego screens”, intrusive group logos, watermarking etc etc are all gay and retarded, and you should go to all possible effort to make sure you don’t have to include them in your encode.
- If you have to include karaoke, don’t wait for lazy timers to shift karaoke for you; Real Encoders shift karaoke with Avisynth.
- JESUS FUCK STOP WARPSHARPING EVERYTHING TO SHIT, IT ISN’T FUNNY, DOES NOT LOOK GOOD, GIVES YOU AIDS AND MAKES OTHER PEOPLE GET EYE CANCER.
- A limited amount of grain can be pretty useful to keep on clean sources.
- When picking TV raws, you should usually use the one with the most detail, unless it has extreme flaws. It is to a certain extent possible to filter away noise, blocking, aliasing or halos, but you can’t recover lost detail.
- Don’t be afraid of overcropping slightly and resizing back up to the desired resolution; it can let you use the resizer as a pseudo-sharpener and noone cares about a few pixels anyway. Overcropping is better than undercropping, black borders/border noise is bad for you.
- Related: http://lipas.uwasa.fi/~f76998/video/conversion/
- Don’t fuck around with the color balance unless you really have to.
- Don’t encode 60fps segments as 60fps, because they aren’t 60 frames per second. Decimate to 29.97.
- Sharpening might look attractive at a glance but don’t overdo it, because oversharpening looks really fucking annoying. See the first point in this list.
- Transport streams are serious fucking business and you should treat them with respect. Learn to IVTC, nub!
Comments (19)
Ha. Nice tips.
How to shift kara with AviSynth btw? (that’s about the only thing I’m lacking :P )
Trim out the OP/ED, add a blankclip of suitable length before it (so the song ends up at the same point along the timeline as on the previous episode that synched), apply textsub() or overlaying, trim the blankclip away, splice the now karaoke’d song together with the rest of the ep again.
Example:
ep = avisource(“lol.avi”)
op = ep.trim(500,3000)
blankclip(op, length=2000)+op
textsub(“kara.ass”)
trim(2000,0)
ep.trim(0,499)+last+ep.trim(3001,0)
Regarding the over-sharpening point, would you prefer to sharpen a natively smooth source, e.g. Gundam SEED, only to undo the blurring from denoising or to make it more visually satisfying if you prefer the edges to be bolder?
Also, do you ever filter only to increase compressibility? If so, what are some good filters for this?
“Don’t encode 60fps segments as 60fps, because they aren’t 60 frames per second. Decimate to 29.97.”
BOLD THIS!
I’ve seen too many people complain about not being able to play OP/ED only because the encoder decided to leave it at 60fps.
Are the 60 fps segments bobbed?
Zanejin: I’m not sure what your point about natively blurry sources is. A moderate level of sharpness is good; a blurry source can naturally be sharpened more than an already sharp source without getting oversharpened, but oversharpening it just because it was blurrier to start with doesn’t make much sense.
I used to filter just to increase compressibility occasionally, but these days I really don’t do that anymore. Potentially good candidates for the job would be any decent denoiser, some kind of antialiaser, and of course our old friend dup().
And yes, the 60fps segments are usually bobbed.
Don’t upscale DVDs to 848/856 when you have MKV and MP4 with their wonderful anamorphic flag? Or even worse, to 720p… and mind you, quite many groups have been doing that and still are AFAIK. @_@
Just to clarify my question on sharpening: Let’s say the average anime’s sharpness is at 0 and that of the anime in question is at -1. Do you prefer keeping the sharpness at -1 (e.g. denoising brings it to -1.3, so you sharpen it back to -1) or “improving the source” so that the sharpness is at 0?
Thanks for your answers!
what are some examples of frequency-domain smoothing filters? ¯\(º_o)/¯
WaryWolf: fft3dfilter, fft3dgpu, dfttest are all frequency domain filters.
Zanejin: Keep in mind that viewers can usually sharpen on playback easily. While removing the bad effects from oversharpening is usually difficult to do during playback.
And considering most viewers have tastes for certain levels of sharpness. It may just be safer to leave it all up to the viewer.
Zanejin: I’m not sure you can quantize it like that because it’s very hard to tell what the actual drawings look like. In either case filtering always has to be decided on a case-by-case basis, but in general, unless the source is oversharpened already, I usually prefer to sharpen it so the end result is a bit sharper than the original source was. The key words being “a bit”, though, because as Harukalover says you should be careful with sharpening.
Harukalover: Problem is most viewers are too dumb/lazy to figure out how to sharpen rather than making a horrible fuss out of it… but w/e. However as TheFluff mentioned, better than oversharpening since fixing that in the decoder is much harder to do. *stares at his ffdshow profile for that which barely runs in real-time with 720p*
besides being slower, what’s wrong with trellis 2? I haven’t been able to find any reason besides speed to not use it so I’m curious.
It removes a lot of noise, to the point where some of the detail might be gone. That’s not really a problem if you use AQ/FGO, though.
Hmm, ok, thanks. Now that you mention it I think I might have read that once somewhere along the way, though I forget where.
Thanks!
@ 7. (martino) Even DivX 6.8 [AVI] will allow anamorphic encoding … certainly in the hands of a capable GUI like AutoMKV ( http://automkv.a.wiki-site.com/index.php/Main_Page ).
Not all players are equall. I found that Windows Media Player worked just fine but some players may have to be manually tweaked to 4:3 or 16:9 .
It’s all good!
Pascal
i was wondering thefluff if you could give hints what filters that elitist/newcomer/1sttimer encode must know. there are overwhelming filters available out in avisynth site. could you list some filters available in that site that is mostly used by you or your fellow encoder?
ps: another question, what filters you or mentar used encoding the dvdrip of claymore if you dont mind telling. just the list of filter that was used. i dont want any specific setting of them cos thats i know thats a *secret*.
I was wondering, especially with VFR raws, would it really be possible to shift karaoke? I’ve tried, and it hasn’t turned out well.
Also, I’ve heard that it’s “easy” to fix a missing first keyframe. As far as I know, that’s a Very Bad Thing, and unfortunately the best raw I could find for a show had fsck’d up first frames. Could you possibly tell as to how that could be done? (simplest being to re-encode, but the content is forever lost.)
Great tips thank you very mutch.
/me Copy pastes it into Encoding Bible.txt