Decent settings for archiving

[Frost45]

I'm looking to archive my DVD's and I want to maximize the quality, but not in a 4GB to 7GB size. I've chosen to leave the resolution at default (720x480) always, strict anamorphic, no cropping, with most of the advanced options enabled i.e. Cabac, 8x8DCT, All Analysis and Uneven Multi-Hexagon motion estimation method. Audio is the main AC3 option, at a bitrate of 192. Would this achieve good results and a size of 2GB or so per 100 min film, higher for 3hr rips? I've ripped the VIDEO_TS folder with DVD-Fab.

[mduell]

I'm looking to archive my DVD's and I want to maximize the quality, but not in a 4GB to 7GB size. I've chosen to leave the resolution at default (720x480) always, strict anamorphic, no cropping, with most of the advanced options enabled i.e. Cabac, 8x8DCT, All Analysis and Uneven Multi-Hexagon motion estimation method. Audio is the main AC3 option, at a bitrate of 192. Would this achieve good results and a size of 2GB or so per 100 min film, higher for 3hr rips? I've ripped the VIDEO_TS folder with DVD-Fab.

The High Profile preset is a good place to start, plus the anaysis and ME changes you mentioned and drop the AAC track. Leave autocrop on, no reason to keep the black bars. At RF 19 I suspect you'll be under 2GB most of the time.

[Bantha]

I create my movies files as MKV. A 2 hour movie is usually 1 GB to 1.2 GB in size with very good Picture Quality. Most 1 hour TV show episodes are about 400-500 MB in size and most 30 minute TV shows are about 250-300 MB. What follows is what I use under normal circumstances.

Under the PICTURE tab, I leave my DVDs at default resolution 720x480, anamorphic to strict, auto crop to off (this way my vids are always divisable by 16 compatible (853x480 or 640x480)) and as far as the black bars go, very small amount of data added, not enough to worry about.

As for VIDEO FILTERS, I leave everything off except for Decombing. And I leave it at Default.

For the VIDEO tab, I leave the codec as H.264 (x264). I determine what the framerate of the picture is (most movies are 23.976 fps) and set the Framerate to it (I do not leave it as SAME AS SOURCE. This is because txMuxer doesn't like VFR videos.). A lot of TV shows are also 23.976 fps. But I found a lot of TV shows are a mixture of 23.976 and 29.97 fps in the same episode! How crazy is that. Star Trek - The Next Generation anyone? For those types of shows, I just encode them as a CFR of 29.97. As for Quality, I leave it as Constant Quality RF:20.

For my AUDIO, I choose the 5.1 AC3 audio channel in English ( because that is what I speak ) and I choose the AC3 Passthrough as the codec. If I really like the movie, I will also add the audio commentary track(s) to the file as well as AC3 Passthrough them. I rarely add the DTS track due to filesize reasons and to be honest, I can't tell a whole lot of difference between DTS and AC3.

For SUBTITLES, I select the default Subtitle stream in English. I also select for movies that have Forced Subtitle tracks those as well and BURN them in the picture. Movies like 2012 (actually, only one I know of that was made the way it was made) has a "SICK" way of doing it's Forced Subtitles, but I found a workaround for it to where I can burn in the Forced Subtitles while maintaining the non-Forced Subtitles as a stream. If interested in my method, just ask.

CHAPTERS I keep the default chapters that Handbrake detects from the DVD and leave them as they are.

Now the fun part, the ADVANCED tab! First off, I start off with a base profile of High Profile and alter these settings from there. I put my Reference Frames to 4 and leave on Mixed References. I leave my B-Frames as 3. Turn Adaptive B-Frames to Optimal and Direct Prediction to Automatic. I check the boxes for Weighted B-Frames and Pyramidal B-Frames. I leave Motion Estimation Method as Hexagon and Subpixel Motion Estimation at 7. (I have experimented with these two settings by placing them w/ higher values. In my experience, it only shrunk a 2 hour movie by about 40 - 80 MB while doubling to tripling the amount of time it takes to encode the movie. Not worth it to me for so little of a gain.) I put Analysis to All and check the box for 8x8 DCT. Enable CABAC Entropy Coding and put Trellis to a value of 2. I leave Psychovisual Rate Distortion and Psychovisual Trellis as default. I leave No Fast-P-Skip and Not DCT-Decimate unchecked and both values for Deblocking as Default.

I hope this helps!

This is no longer valid! See later in the post for my updated settings I use.
http://forum.handbrake.fr/viewtopic.php ... 706#p74706

[mduell]

Under the PICTURE tab, I leave my DVDs at default resolution 720x480, anamorphic to strict, auto crop to off (this way my vids are always divisable by 16 compatible (853x480 or 640x480)) and as far as the black bars go, very small amount of data added, not enough to worry about.

For a guy with as many x264 advanced tweaks as you, it's surprising to see you wasting bits so freely. Forget mod16 and take the cropping.

For the VIDEO tab, I leave the codec as H.264 (x264). I determine what the framerate of the picture is (most movies are 23.976 fps) and set the Framerate to it (I do not leave it as SAME AS SOURCE. This is because txMuxer doesn't like VFR videos.). A lot of TV shows are also 23.976 fps. But I found a lot of TV shows are a mixture of 23.976 and 29.97 fps in the same episode! How crazy is that. Star Trek - The Next Generation anyone? For those types of shows, I just encode them as a CFR of 29.97.

While this is a useful workaround to support your degenerate device, it's not recommended as a general practice

[Bantha]

For a guy with as many x264 advanced tweaks as you, it's surprising to see you wasting bits so freely. Forget mod16 and take the cropping.

I have read somewhere (either on this website's forums or guide) or somewhere else on the internet, that MPEG doesn't like it when you create a video that is not divisable by 16 evenly. You can have some weird effects on the edge blocks on the sides of the frame. And Handbrake's autocrop feature was giving me some weird display resolutions, so I decided to just leave the resolution the same as the DVD (720x480).

For the VIDEO tab, I leave the codec as H.264 (x264). I determine what the framerate of the picture is (most movies are 23.976 fps) and set the Framerate to it (I do not leave it as SAME AS SOURCE. This is because txMuxer doesn't like VFR videos.). A lot of TV shows are also 23.976 fps. But I found a lot of TV shows are a mixture of 23.976 and 29.97 fps in the same episode! How crazy is that. Star Trek - The Next Generation anyone? For those types of shows, I just encode them as a CFR of 29.97.

While this is a useful workaround to support your degenerate device, it's not recommended as a general practice

Actually, it's not the degenerate device's fault this time. It is txMuxer. It doesn't like VFR videos. If I play them through Mencoder, they play fine on the degenerate device. And I would rather view the movies w/ their video/audio streams untouched. Also the WDTV Live doesn't like VFR videos also. So, I have two degenerate devices.

[mduell]

For a guy with as many x264 advanced tweaks as you, it's surprising to see you wasting bits so freely. Forget mod16 and take the cropping.

I have read somewhere (either on this website's forums or guide) or somewhere else on the internet, that MPEG doesn't like it when you create a video that is not divisable by 16 evenly. You can have some weird effects on the edge blocks on the sides of the frame. And Handbrake's autocrop feature was giving me some weird display resolutions, so I decided to just leave the resolution the same as the DVD (720x480).

It's an outdated overblown warning that has since been removed from x264. Autocrop can be wrong sometimes but you can increase the number of previews in preferences to help it or manually specify the correct cropping. Encoding black bars is [censored].

For the VIDEO tab, I leave the codec as H.264 (x264). I determine what the framerate of the picture is (most movies are 23.976 fps) and set the Framerate to it (I do not leave it as SAME AS SOURCE. This is because txMuxer doesn't like VFR videos.). A lot of TV shows are also 23.976 fps. But I found a lot of TV shows are a mixture of 23.976 and 29.97 fps in the same episode! How crazy is that. Star Trek - The Next Generation anyone? For those types of shows, I just encode them as a CFR of 29.97.

While this is a useful workaround to support your degenerate device, it's not recommended as a general practice

Actually, it's not the degenerate device's fault this time. It is txMuxer. It doesn't like VFR videos. If I play them through Mencoder, they play fine on the degenerate device. And I would rather view the movies w/ their video/audio streams untouched. Also the WDTV Live doesn't like VFR videos also. So, I have two degenerate devices.

Mea culpa, your degenerate devices and software stack.

[mkelley]

Also the WDTV Live doesn't like VFR videos also. So, I have two degenerate devices.

Um, what doesn't it like about them? My WDLive doesn't have any issues at all when I leave it as "Same as source".

Now -- I haven't had any video that has two different framerates (you suggest ST-Next Gen, which I don't have) so if you are saying that's an issue with them only I'll be (thankfully) aware of this. Is that the only time you've had problems?

[Bantha]

The problem I have noticed with the WDTVLive is with movies encoded Same As Source. Due to the VRF, as the movie nears the end, the audio/video sync become more and more off. While watching 17 Again, by the end of the movie, the audio was off about 1/2 a second. Not too bad, but enough for someone such as myself with the "eye".

Note, I have noticed since I have last tested VFR files w/ the WDTVLive there are 2 major upgrades to the unit that may fix this issue. I will have to re-encode a movie as SAS and let you know.

[mkelley]

I know MP4 sync was off, but MKV sync has always been fine for me (and I have a very critical eye, since I do animation lip sync professionally). This has been true for the entire time I've had my Live (which, to be fair, has only been about five months).

All of my encodes have been using the snapshot after .94 was released, or later -- I'm wondering also if maybe you were using earlier Handbrake versions. In any case, I can tell you that there are absolutely zero sync issues on my end using SAS encoded for the Live with firmware released in the last four months in an MKV container and Handbrake .94+Snapshot or later (whew! How about that for qualifications?)

[nigo]

Hi there

As I'm very new to Handbrake and digital video+compression so I did a bit of experimenting my self. I started completely from scratch with no idea what the different settings were about. I thought I'd share the results even though they might not apply for the rest of the world More experienced people will probably consider this trivia, but it might keep newcomers from asking questions Consider it as guidelines and nothing else

I did my testing on a 16 minute snip of Return of the King (LOTR). A lot of action, foggy scenes, panning etc. I used a AMD Athlon 64 X2 dual, 4200+ CPU running winxp. I used DD 5.1 pass through, the normal profile and changed only the container type and compression settings under the video tab. The rest was left at default. Here's the results:



Size of the original is the data size for the clip in % of the original. Time factor is how long the processing time was compared to the duration of the original clip. 100% means 16 minutes. > 100% means it took longer time. Of course quality is very subjective, but I rated the quality 1-5 with 1 being terrible and 5 hardly/not distinguishable from the original.

As you can see there is a huge difference in the time factor. The H.264 codec usually takes longer time than the original clip, where as MPEG4 codec is much faster.

Apart from pictures being a bit darker MPEG4 cant beat the H.264 codec when it comes to quality. Only at a constant compression factor of 97% it got close, but here only a file size reduction to 83% was obtained.

I know the tips in handbrake state that ~60% is a "sane" value for compression (also at MPEG4), but it isn't unless quality isn't an issue. I'd rather choose a much higher (>90%) for MPEG4 or an average bitrate of min. 1500 kbps. For H.264 the 60% seems nice.

Notice also that the process time seems to depend on the container type.

The GB/movie is just a rough guideline assuming that all movies compress the same (they don't) and the duration is 2 h.

If I got the time I'd probably go for H.264 at a constant comp setting of 60%. That should also give you 500 movies on a 1TB drive. However that will also take more than 1000 h (on a system like mine) to process!

Hope its a bit useful

[jbrjake]

No one told you to use 60% quality for the ffmpeg mpeg-4 part 2 encoder.

[Bantha]

I found something out quite interesting, I think anyways.... I decided to do an overnight experiment *again* to find a good balance between speed and filesize encoding my movies. See which functions actually gave a decent decrease in filesize, and which were just time-hogs and did nothing or very little to filesize. I decided to start w/ High Profile as a base with only 1 change, I disabled Detelecine. Processed Sherlock Holmes through Handbrake (svn3036) as a MKV, w/ H264, SAS fps, CRF 20, AC3 passthrough w/ main audio stream only, subtitles, and chapters. This gave me a filesize of 1.39 GB.

Next I turned on Pyramidal B-Frames and the filesize dropped to 1.36 GB.

Then I increased my B-Frames from 3 to 6 and my filesize dropped to 1.33 GB. At this point, my encoding time dropped about 8 fps to 47.88 fps from my original setting.

I found that changing Direct Prediction to Auto from Spatial does nothing for filesize but does drop my encode speed. Same goes for changing Motion Estimation Method from Hexagon to Uneven Multi-Hexagon (I didn't dare try the other 2 higher settings). Changing Subpixel Motion Estimation from 7 to 8 actually increased my filesize by about 0.01 GB. I changed Analysis to All which did nothing for my filesize either.

Last setting I played with so far is Trellis. I changed it from 1 to 2, and my filesize dropped from 1.33 to 1.29 GB. So far I have with the settings I have changed under Advanced, I have saved approximately 0.1 GB over a feature length movie. 1.39 GB vs 1.29 GB. But I am not done yet. As Trellis has one more option for me to change it to.

So finally, I changed Trellis from 2 to 0. And this is where it gets interesting (to me anyways). My encode speed increased from when I had Trellis set to 2 (37.43 fps) to 48.90 fps. And not only that, but my filesize shrank once more, this time to 1.25 GB! How is this possible? I thought the whole idea of Trellis was to make sure the bits were used most optimally! Not only does Trellis set to 0 give me the smallest filesize yet, but encodes it almost as fast as HP! Does Trellis not work good when encoding in CRF?

To summarize it all up, I went from 1.39 GB down to 1.25 GB for my 2 hour, 8.5 min movie. That is approximately 134 MB per 2 hour movie. That is equivalent to 201 more movies I can put on my server. Without that much sacrifice to speed.

Based on this newfound knowledge, I will have to update my settings from earlier in the post.

[Bantha]

I create my movies files as MKV. A 2 hour movie is usually 1 GB to 1.2 GB in size with very good Picture Quality. Most 1 hour TV show episodes are about 400-500 MB in size and most 30 minute TV shows are about 250-300 MB. What follows is what I use under normal circumstances.

I start off with High Profile as a base, then alter it as follows:

Under the PICTURE tab, I leave my DVDs at default resolution 720x480, anamorphic to strict, auto crop to off (this way my vids are always divisable by 16 compatible (853x480 or 640x480)) and as far as the black bars go, very small amount of data added, not enough to worry about.

As for VIDEO FILTERS, I leave everything off except for Decombing. And I leave it at Default.

For the VIDEO tab, I leave the codec as H.264 (x264). I used to determine what the framerate of the picture was, but now I just leave it set to SAME AS SOURCE. As for Quality, I leave it as Constant Quality RF:20.

For my AUDIO, I choose the 5.1 AC3 audio channel in English (because that is what I speak) and I choose the AC3 Passthrough as the codec. If I really like the movie, I will also add the audio commentary track(s) to the file as well as AC3 Passthrough them. I rarely add the DTS track due to filesize reasons and to be honest, I can't tell a whole lot of difference between DTS and AC3.

For SUBTITLES, I select the default Subtitle stream in English. I also select for movies that have Forced Subtitle tracks those as well and BURN them in the picture. Movies like 2012 (actually, only one I know of that was made the way it was made) has a "SICK" way of doing it's Forced Subtitles, but I found a workaround for it to where I can burn in the Forced Subtitles while maintaining the non-Forced Subtitles as a stream. If interested in my method, just ask.

CHAPTERS I keep the default chapters that Handbrake detects from the DVD and leave them as they are.

Now the fun part, the ADVANCED tab! I leave my Reference Frames set to 3 and leave on Mixed References. I change the B-Frames to 6 (For live action). Leave Adaptive B-Frames as Optimal and Direct Prediction as Spatial. I check the boxes for Weighted B-Frames and Pyramidal B-Frames. I leave Motion Estimation Method as Hexagon and Subpixel Motion Estimation at 7. I leave Analysis as the default setting, and I check the box for 8x8 DCT. Enable CABAC Entropy Coding and put Trellis to a value of 0. I leave Psychovisual Rate Distortion and Psychovisual Trellis as default. I leave No Fast-P-Skip and Not DCT-Decimate unchecked and both values for Deblocking as Default.

I hope this helps!

[Bantha]

The problem I have noticed with the WDTVLive is with movies encoded Same As Source. Due to the VRF, as the movie nears the end, the audio/video sync become more and more off. While watching 17 Again, by the end of the movie, the audio was off about 1/2 a second. Not too bad, but enough for someone such as myself with the "eye".

Note, I have noticed since I have last tested VFR files w/ the WDTVLive there are 2 major upgrades to the unit that may fix this issue. I will have to re-encode a movie as SAS and let you know.

I have updated my WDTVLive to the newest firmware (v1.02.21) and tried a MKV file I made (using svn3036) that is VFR. And plays like a champ! Dunno if the firmware update fixed this issue, or was it perhaps I was using files encoded from a pre-0.94 release of Handbrake. Dunno. Just glad it works now!

[creamyhorror]

I found something out quite interesting, I think anyways.... I decided to do an overnight experiment *again* to find a good balance between speed and filesize encoding my movies. See which functions actually gave a decent decrease in filesize, and which were just time-hogs and did nothing or very little to filesize. I decided to start w/ High Profile as a base with only 1 change, I disabled Detelecine. Processed Sherlock Holmes through Handbrake (svn3036) as a MKV, w/ H264, SAS fps, CRF 20, AC3 passthrough w/ main audio stream only, subtitles, and chapters. This gave me a filesize of 1.39 GB.

Why disable Detelecine?? NTSC movies almost invariably require it, else you'll likely end up deinterlacing the movie instead!

Next I turned on Pyramidal B-Frames and the filesize dropped to 1.36 GB.

Then I increased my B-Frames from 3 to 6 and my filesize dropped to 1.33 GB. At this point, my encoding time dropped about 8 fps to 47.88 fps from my original setting.

I found that changing Direct Prediction to Auto from Spatial does nothing for filesize but does drop my encode speed. Same goes for changing Motion Estimation Method from Hexagon to Uneven Multi-Hexagon (I didn't dare try the other 2 higher settings). Changing Subpixel Motion Estimation from 7 to 8 actually increased my filesize by about 0.01 GB. I changed Analysis to All which did nothing for my filesize either.

...

To summarize it all up, I went from 1.39 GB down to 1.25 GB for my 2 hour, 8.5 min movie. That is approximately 134 MB per 2 hour movie. That is equivalent to 201 more movies I can put on my server. Without that much sacrifice to speed.

What you didn't realize is that in making all these changes you have changed the quality of the encode as well. "Constant quality" is not constant across different combinations of settings. If you drop trellis to 0, for example, your filesize will fall but so will your quality, in all likelihood.

The trouble is that it's not easy to tell the quality of encodes. SSIM values are a rough guide, but certain options work against SSIM or have no impact while aiming to improve the "subjective quality" of the encode.

Thus what you were trying to do was reduce filesize at the expense of quality (though probably only a small amount).


edit:

Last setting I played with so far is Trellis. I changed it from 1 to 2, and my filesize dropped from 1.33 to 1.29 GB. So far I have with the settings I have changed under Advanced, I have saved approximately 0.1 GB over a feature length movie. 1.39 GB vs 1.29 GB. But I am not done yet. As Trellis has one more option for me to change it to.

So finally, I changed Trellis from 2 to 0. And this is where it gets interesting (to me anyways). My encode speed increased from when I had Trellis set to 2 (37.43 fps) to 48.90 fps. And not only that, but my filesize shrank once more, this time to 1.25 GB! How is this possible? I thought the whole idea of Trellis was to make sure the bits were used most optimally! Not only does Trellis set to 0 give me the smallest filesize yet, but encodes it almost as fast as HP! Does Trellis not work good when encoding in CRF?

It works just fine. You generally get more quality per bit with trellis on. You just can't predict how it'll affect the final bitrate/filesize.

[nigo]

No one told you to use 60% quality for the ffmpeg mpeg-4 part 2 encoder.

Ok... I just used this tooltip as I had no idea what settings to start at:



...but perhaps its a bug then?

[Bantha]

Why disable Detelecine?? NTSC movies almost invariably require it, else you'll likely end up deinterlacing the movie instead!

I disabled Detelecine after reading an piece (can't find it now, it's been a while) that stated something to the effect that by leaving it on, could adversly affect movies that were not hard-telecined.

What you didn't realize is that in making all these changes you have changed the quality of the encode as well. "Constant quality" is not constant across different combinations of settings. If you drop trellis to 0, for example, your filesize will fall but so will your quality, in all likelihood.

The trouble is that it's not easy to tell the quality of encodes. SSIM values are a rough guide, but certain options work against SSIM or have no impact while aiming to improve the "subjective quality" of the encode.

Thus what you were trying to do was reduce filesize at the expense of quality (though probably only a small amount).

I put on screen both encodes (Trellis 0 vs Trellis 2) and I have to say that I am hard-pressed to tell a difference between the two. Meaning I don't see one at all! Even if there is technically a visually difference, my naked-eye can't see it.

[s55]

Just a badly worded tooltip nigo. It's out of date in 0.9.4

[creamyhorror]

I disabled Detelecine after reading an piece (can't find it now, it's been a while) that stated something to the effect that by leaving it on, could adversly affect movies that were not hard-telecined.

Since Detelecine is stateless (can be applied to any sequence), it should be pretty safe to apply on anything.

I put on screen both encodes (Trellis 0 vs Trellis 2) and I have to say that I am hard-pressed to tell a difference between the two. Meaning I don't see one at all! Even if there is technically a visually difference, my naked-eye can't see it.

That's why I said "only a small amount". Generally you can't see the differences in quality that small changes in settings produce (especially at high bitrates), not without pausing, zooming in and flipping back and forth between frames. (This is why many people also rely on the SSIM values reported by x264 to see if any improvement was gained, since those are more sensitive than the human eye.) Since the improvement is too small to be worth the speed tradeoff to you, feel free to drop it.

The main point is that you can't change settings and then simply compare bitrates to determine whether one setting is better than another. That assumes that quality is constant across encodes, which isn't true.

[randomreuben]

Hi jbrjake,

I can confirm that screenshot posted by nigo a few posts earlier. It's what happens when you switch from x264 to FFMPEG encoding. The quality slider automatically sets itself to 60% but the quality marker calls itself 94%. I agree with nigo that it's a bug there.

Take care,

randomreuben

[Deleted User 11865]

Hi jbrjake,

I can confirm that screenshot posted by nigo a few posts earlier. It's what happens when you switch from x264 to FFMPEG encoding. The quality slider automatically sets itself to 60% but the quality marker calls itself 94%. I agree with nigo that it's a bug there.

Take care,

randomreuben

Doesn't matter, percentage scale is gone in current code.

[creamyhorror]

Doesn't matter, percentage scale is gone in current code.

yessssss

wait is that for x264 as well or just for ffmpeg

[Deleted User 11865]

Doesn't matter, percentage scale is gone in current code.

yessssss

wait is that for x264 as well or just for ffmpeg

All encoders. You haven't upgraded to the latest snapshot yet? http://forum.handbrake.fr/viewtopic.php?f=6&t=15452

[creamyhorror]

All encoders. You haven't upgraded to the latest snapshot yet? http://forum.handbrake.fr/viewtopic.php?f=6&t=15452

Nope, haven't upgraded since I haven't used HB for anything recently. Great move, though...no more wrangling over quality percentage with people who don't know how it works.

[moviebuff]

I seem to find the best setting I get is Normal Preset with Rf 20 for dvd and rf 22 for blu ray, am I shorting myself anything by going normal than high ? Seems everyone does the the high preset over the normal? My encodes seem alot longer on high and dont notice a big difference in quality.