X264/Handbrake settings for old cartoons?

[mikehunt69]

I have a mixture of old cartoons (pre-1965 sourced from DVD and Blu-Ray, currently stored on my hard drive in their original video + audio.
These are ~5mbit VBR MPEG2/192kbit CBR AC3 mono for the DVDs and 17.5mbit VBR AVC/192 kbit CBR AC3 mono for the blu-rays.


I would like to re-encode in handbrake for smaller sizes.


Any suggestions for settings I should use?


Is noise reduction a good idea to make the files more compressible?

[coltonhelp]

If you mean DVNR, then no.

[mikehunt69]

I was talking about the 'Denoise' filter on Handbrake.

[hello_hello]

Denoising will probably make the files more compressible but any denoising is some sort of compromise between removing noise and blurring.

I don't know anything about Handbrake's denoising but maybe run a few test encodes (you can pick the start and end points for encoding so there's no need to encode a whole file each time). Start with the lightest denoising and work your way up. It's all personal taste. Chances are, if nothing else the lightest denoising will only remove very light noise without causing any noticeable blurring, and still make the video a little more compressible, but I don't really know.

[Prot]

I ripped my Duckman DVDs to my PC using AnyDVD. My intention is to convert the episodes to mp4 for portability. When I use Handbrake to do this, the quality is terrible, and file size close to 400 meg per episode. Using comb filter and changing resolution to 640x480, all I manage to do was improve the quality, but only reduce file size by a small amount.

Any ideas how to reduce the file size significantly?

I'm trying a resolution of 480x360 next to see if that helps.

TIA

[mikehunt69]

I have been doing some testing, source is DVD 720x480, Handbrake 0.9.9 (use 64-bit if you can for speed).

My source:


French Rarebit, from NTSC DVD, source for 7:22 minute episode 285MB file size, of which 10MB is AC3 1.0 audio at 192kbit, and the rest (275MB) is the MPEG2 video.


First:
Preset - High Profile
Then, Picture tab, Anamorphic Loose - try Strict if your device doesn't do aspect ratios properly


Video tab:
x264 tune - animation - this is designed for film cartoons (i.e. not for modern CGI)


Quality: Constant Quality, change this
x264 preset - very slow is best


Audio: Auto-passthru for me, keeps the audio stream the same, around 85MB/hour. For mono, 64kbit AAC is probably fine though, reduces this to 28MB/hour, check compatibility on the AAC though.

So then I tried different Constant Quality RF settings, for the video only:


original: 275MB (5.1 mbit)
18/very slow: 164MB: 43% saving
18/slower: 176MB 36%
18/slow: 180MB 35%
18/medium: 206MB 25%
19/medium: 177MB 36%
24/medium: 59MB 79% (1.1mbit)

As I understand it, you can't necessarily compare 18/very slow with 18/slower - they aren't necessarily the same quality (19 slow is better than 20 slow, but 19.5 slow isn't necessarily better than 20 medium or whatever), I would appreciate more detail on this point though.

Visual results, close-up:


original - the original source with film grain well-evident
18 - basically identical in motion, grain still there but reduced, archival grade
24 - obvious differences in motion, picture still sharp, all grain gone, some macroblocking, still very watchable


For archival on a computer store the original Mpeg-2 video + AC3 audio without any transcoding.
Watching on portable devices, you can afford to throw away a lot of detail by upping the rate factor. At RF 24 the video stream is down to around 480MB/hr from the original 2200MB/hr, which is good, and you can certainly go lower than that, it's just balancing file size vs. quality. For better quality maybe a higher RF.

[mikehunt69]

And one more, 20/very slow 107MB. RF 18 very slow is 20% smaller than RF 20 medium. Though I suspect you are looking for file sizes like 80% smaller, so it might not be worth going for very slow if it takes too long.

Here's some screen shots from Media Player Classic at frame size: double (because nobody watches in a tiny window) for RF 18, 20, 24 and the original file

[jagabo]

At the same CRF, veryfast usually gives smaller files than faster, fast, medium, and sometimes even slow. Not as good quality though.

Watch for posterization artifacts in dark grainy areas. And rough edges on moving objects. That's where defects show up first.

[hello_hello]

First:
Preset - High Profile
Then, Picture tab, Anamorphic Loose - try Strict if your device doesn't do aspect ratios properly.

They're both anamorphic so both will cause the image to display incorrectly if a device doesn't support aspect ratios properly.
The difference is, "strict" doesn't allow resizing. It uses the same pixel aspect ratio as the source. "Loose" allows resizing (how much I can't remember as I don't use Handbrake myself) and the pixel aspect ratio will be adjusted accordingly if need be.
I'm pretty sure anamorphic strict and anamorphic loose with a modulus of 2 and no resizing should produce the same result.

Anamorphic "none" would be the one to use if a device doesn't do aspect ratios. It resizes to square pixels.

[mikehunt69]

So I have a few questions:

4.1 vs 3.1 - I find they are giving basically the same file size - should 4.1 give better PQ given that I am encoding at RF21 in both cases?


High vs Main - Main seems to give a smaller size - is High going to give better PQ?


All this given a NTSC DVD source.


Also if I convert my VFR source to CFR, is that bad for PQ, or does it just mean bigger file sizes (about 4% I am seeing). What disadvatages does VFR have?


The grain preset seems to kill compression, e.g., source: 283MB, RF18-animation 167MB, RF18-grain 285MB (bigger than the source!), RF20-animation 86MB, RF20-grain 192MB.


From an mpeg-2 source what settings should work best to preserve grain while still achieving some compression (h264 hopefully more efficient than mpeg-2)?

[jagabo]

4.1 vs 3.1 - I find they are giving basically the same file size - should 4.1 give better PQ given that I am encoding at RF21 in both cases?

Not necessarily.

High vs Main - Main seems to give a smaller size - is High going to give better PQ?

Not necessarily.

Profiles and levels are limitation placed on the encoder regarding what features, frame sizes, bitrate, etc. the encoder can use.

http://en.wikipedia.org/wiki/H.264#Profiles
http://en.wikipedia.org/wiki/H.264#Levels

They don't directly determine the quality of the encoding. They are used more to guaranty compatibility with playback devices.

Also if I convert my VFR source to CFR, is that bad for PQ, or does it just mean bigger file sizes (about 4% I am seeing). What disadvatages does VFR have?

Variable frame rate allows for greater compression. Ie, higher quality at the same bitrate, or smaller files at the same CRF. But the difference between a VFR and CFR encode, with all else equal, is usually just a few percent. Animation tends to benefit a bit more (because of many repeated frames) than live footage. VFR is harder for an editor to work with, and players play, so I usually avoid it.

The grain preset seems to kill compression, e.g., source: 283MB, RF18-animation 167MB, RF18-grain 285MB (bigger than the source!), RF20-animation 86MB, RF20-grain 192MB.

It takes a lot more bitirate to encode all that grain. Grain/noise is a big enemy of compression. Because all high compression codecs get a lot of their compression largely by not re-encoding parts of the picture that don't change from frame to frame. The more grain you have, the more the picture changes from frame to frame. Grain also makes individual frames harder to compress because the encoder has to keep track of more "detail" within the frame.

[mikehunt69]

I guess for DVD source material >3.1 doesn't matter, but I'm not really clear why Main is bigger than High

Anyway, I am looking at anamorphic settings now:


None - 640x480 112MB
Loose/Strict (both the same) 720x480 143MB


That's an insane increase in file-size - 28%, any idea why?


Assuming that it will be viewed on an effective 1440x1080 screen, how much better is 720x480 anamorphic over 640x480 non-anamoprhic, if at all?


Finally, I have Looney Tunes Golden Collection Volume 2, and the cartoons are encoded by Handbrake using FPS 'Same As Source' at 23.976fps, but on Volume 6 they encode at 29.97fps. Is this just because they were differently mastered by Warner Bros? I would have thought the original framerate in 1952 or whatever would have been 23.976fps. Does it matter? I guess I shouldn't force the 29.97fps to 23.976???

[coltonhelp]

I guess for DVD source material >3.1 doesn't matter, but I'm not really clear why Main is bigger than High

Anyway, I am looking at anamorphic settings now:


None - 640x480 112MB
Loose/Strict (both the same) 720x480 143MB


That's an insane increase in file-size - 28%, any idea why?


Assuming that it will be viewed on an effective 1440x1080 screen, how much better is 720x480 anamorphic over 640x480 non-anamoprhic, if at all?


Finally, I have Looney Tunes Golden Collection Volume 2, and the cartoons are encoded by Handbrake using FPS 'Same As Source' at 23.976fps, but on Volume 6 they encode at 29.97fps. Is this just because they were differently mastered by Warner Bros? I would have thought the original framerate in 1952 or whatever would have been 23.976fps. Does it matter? I guess I shouldn't force the 29.97fps to 23.976???

IVTC'ing the DVD source is highly recommended for film sources, and also, the true framerate for Looney Tunes is 24fps. It's evident that you're trying to encode an episode of Looney Tunes.

[mikehunt69]

So are these good Handbrake settings for my 29.97fps source:

Detelecine: default
Framerate: 24fps
Constant framerate


?


How about for the 23.976 fps source?


Should I do:


Detelecine: default
Framerate: 24fps
Constant framerate



also?


How can I tell that the detelecine has done something to the video? It seems that the detelecine doesn't force the video into 23.976 or 24fps (for one source it happens automatically), I have to force it.

[jagabo]

Use 23.976 fps. 24 fps will get a duplicate frame every 1001 frames. Or you'll have to speed up the audio to match the faster video frame rate.

[PuzZLeR]

I guess for DVD source material >3.1 doesn't matter, but I'm not really clear why Main is bigger than High

This will mean very little except as a flag, and for compliance with devices/decoders. Much more depends on your settings.

Just use Level 4.1 and High Profile, which is what the higher quality playback devices expect, such as media players, blu-ray players, etc, and it also allows you to include more higher quality features of the spec (but may not be applicable in your case).

In short, Level 4.1 and High Profile super-set the features allowable with Level 3.1 and Main Profile respectively. If any device does not play Level 4.1 and/or High Profile today, then it is NOT a good decoder of H.264.

[mikehunt69]

Use 23.976 fps. 24 fps will get a duplicate frame every 1001 frames. Or you'll have to speed up the audio to match the faster video frame rate.

Is that the same with PAL too? I guess NTSC 29.97 is very easy to change to 23.997 fps because it's 4 frames in 5, but how does that work with 25fps PAL sources? Is 24fps not better with PAL?

[jagabo]

Leave PAL at 25 fps.

[mikehunt69]

High vs Main - Main seems to give a smaller size - is High going to give better PQ?

I looked at this, and High is actually quite considerably better in terms of detail on things like pastel shades (non-solid blocks of color), which is why I guess the filesize is bigger. So Main Quality 20 != High Quality 20.

Anyway, quick question: what's the best way (any encoder/filter/software at all) to de-interlace an interlaced source?
Screenshot from the DVD:
https://forum.videohelp.com/images/imgfiles/tkVgkYzh.jpg

[manono]

Anyway, quick question: what's the best way (any encoder/filter/software at all) to de-interlace an interlaced source?

"A video is worth a thousand pictures". 10 seconds of steady motion will be plenty.

Anyway, for film-based videos (which yours probably is) you don't deinterlace at all - you IVTC with an AviSynth IVTC. For true interlaced video the very best deinterlacer (and you asked for the best) is AviSynth's QTGMC. Several programs (XviD4PSP, for example) can do either.

[mikehunt69]

<delete>

[mikehunt69]

Video link:

http://www.putlocker.com/file/55E1F0D4EA535F3E

This is 10s of the original MPEG-2 video, not changed from the DVD.

[jagabo]

It's film based. It should be inverse telecined, not deinterlaced. Try TFM().TDecimate(). Here it is after IVTC and a little cleanup. It's lost a little detail from the downsize/upsize to get rid of the dot crawl artifacts.