Why does resolution change in handbrake with "keep aspect ratio" checked?

[ikarius]

Simple question; I'm converting most of my DVDs to digital files, and I don't understand why when I see source showing "720 x 480", and I have "Keep Aspect Ratio" checked, the size on every movie changes; primarily the height changes from movie to movie. Is this cropping detecting that the movie height isn't really 480?

I'm encoding for Plex with a Roku client, and I want good quality, but trying to keep size sort-of reasonable.

[Baldrick]

The video is resized to the DVDs display aspect ratio and with cropping of black borders. So the Pixel Aspect Ratio will be around 4:3, 16:9 or 2:35:1.

You can also keep it as 720x480 for all movies with just 16:9 or 4:3 Display Aspect Ratio flags. But not all video players supports it.

Read https://trac.handbrake.fr/wiki/AnamorphicGuide

[DB83]

Read 'What is a dvd' on the left of this screen.

All US dvds have a SAR of 720*480. But the Display Aspect Ratio, which is what you are keeping when you check that box is different. A pure 4:3 video could be 720*540 whereas a 16:9 video will have even less height pixels.

[hello_hello]

The automatic cropping will remove any black bars etc which will in turn change the height and/or width of the video. Often the remaining dimensions will differ according to the amount of cropping.

DVDs don't use square pixels. A 16:9 NTSC DVD has a resolution of 720x480 but displays using the equivalent square pixel resolution of 853x480. (853/480 = 16/9)
If you select "anamorphic none" and "keep aspect ratio" is checked, HandBrake will resize the DVD to square pixel dimensions for you, then encode it. However instead of taking the 720x480 DVD and resizing to 853x480, it keeps the original width and reduces the height instead, giving you roughly 720x404. If you decide to resize the video further by adjusting the width, HandBrake will automatically adjust the height. The lower the value you specify in "Modulus" the more accurately Handbrake can resize the DVD. Because the "anamorphic none" option resizes the DVD to square pixels by reducing the height you can lose a little of the resolution. Many people resize to 854x480 instead, but HandBrake won't let you do that.

The anamorphic "strict" option doesn't resize the DVD at all. It encodes it at 720x480 (or whatever is left after cropping) and the player resizes the video on playback just as it would the original DVD. This should give you better quality, but not all devices will resize the video correctly on playback. Anamorphic "loose" is much like anamorphic "scrict" except it may resize the video a little. The explanation for why it exists would take a bit of typing..... personally I'd use anamorphic strict if your player displays the encoded video correctly, and anamorphic none with the modulus set to 2 if it doesn't.

Or Try VidCoder, It's an alternative HandBrake GUI. It'll let you resize "up" to 853x480 for better quality (although the file sizes will be larger. Ideally you'd take the original height (480), then subtract the top and bottom cropping. For example if you cropped 40 pixels from the top and 40 from the bottom the new height would be 400. Then you'd adjust the width. Vidcoder will adjust the output dimensions it displays as you do. When the output height equals 400, that'd be the ideal resizing.

[ikarius]

Thanks for the clarifications. As handbrake doesn't show any details on the cropping when you set it to automatic, I was just left wondering why the pixel height was changing when I checked the "keep aspect ratio" box. I'm not dissatisfied at all with handbrake, only looking to understand what it's doing as much as possible.

I've been following the "Rokoding" guide http://roku.yt1300.com/ to encode for Plex+Roku. At this point my only question is why it seems like DVDs end up hardly any smaller than blu-rays. I do scale my blu-rays down to 720p, with constant quality and RF 23, whereas I'm using constant quality/RF19 on DVDs, but I'd think that with more pixes the blu-rays would end up a lot larger, but they only seem to be ~30% larger. I'm going to try encoding the same movie from a DVD and a blu-ray and see what a direct comparison looks like.

[DB83]

No of pixels have nothing to do with video size. Size = run length * bitrate. So if your Blu-rays have more bit-rate (to allow for more pixels) then the size will be bigger.

[ikarius]

Unless I'm completely crazy, pixels would directly correlate to the bitrate. More pixels = more information, and of course it's all being crunched through a very sophisticated compression algorithm, which means that the amount of change, and the compressibility has a lot to do with the ending bitrate, but if you have more information to start with, you're likely to end up with more information after compression. Also note- I'm using constant quality, rather than 2-pass and a target bitrate. I assume that the bitrate varies during the stream; I would assume that more action and more change from frame to frame would create a spike in bitrate compared to a relatively static shot. Now, this is based on understanding of general computing principals rather than intimate understanding of video codec internals, so if I've got anything wrong, please feel free to correct my misunderstandings.

[smrpix]

It may seem counterintuitive initially, but as DB83 says size=run length*bitrate. It's the only truth. It's kind of a mantra around here. Repeat it and be enlightened.

And yes, assuming variable bitrate encoding, your internal logic here is correct:

I assume that the bitrate varies during the stream; I would assume that more action and more change from frame to frame would create a spike in bitrate compared to a relatively static shot.

[ikarius]

Hmm. Is bitrate always fixed in video files? Also, what are the determining factors for bitrate? As I said before, I am using constant quality rather than 2-pass. 2-pass allows me to specify a target bitrate, does constant quality create a variable bitrate, or auto-figure a bitrate?

I saw an encode of Iron Sky where the quality was generally quite good, but early on when the videowas showing a moon lander cruising across the face of the moon, and the shot was also panning across the face of the moon, the video looked like it was dropping frames. My guess is that someone specified a target bitrate, and the rate of change during that pan shot with lots of little details moving was too much to encode at the bitrate, so it ended up dropping a lot of the intermediate infor. Is that a fair guess about what was going on?

I'm curious about how I go about avoiding that- I was hoping that I wouldn't have to worry about that using constant quality. Any tips are appreciated.

[manono]

Is bitrate always fixed in video files?

Not in quality-based modes. You can only figure the average bitrate after the encoding is finished.

Also, what are the determining factors for bitrate?

In CQ modes? The quality you set and the complexity of the video.

...does constant quality create a variable bitrate...

Yes.

[ikarius]

Okay, cool. So, I'm not completely off base, most of my educated guesses were in the right ballpark. The lack of difference probably comes down to using constant quality of 19 for DVD and 23 for BD, which means it's working much harder to compress the blu-rays. I'll do some more experimentation to see how far I can push up the CQ scale for DVDs without seriously degrading them.

[hello_hello]

TAt this point my only question is why it seems like DVDs end up hardly any smaller than blu-rays. I do scale my blu-rays down to 720p, with constant quality and RF 23, whereas I'm using constant quality/RF19 on DVDs, but I'd think that with more pixes the blu-rays would end up a lot larger, but they only seem to be ~30% larger. I'm going to try encoding the same movie from a DVD and a blu-ray and see what a direct comparison looks like.

The difference in resulting bitrate/file size between CRF19 and CRF23 is likely to be reasonably large for any given video. If you were to encode the two using the same CRF value there'd probably much more of a size difference.

[ikarius]

Yep, the Rokoding site suggested those CQ settings, with a lower compression setting for DVDs, without explaining why they recommend it. I imagine it comes down to not wanting to further degrade the DVD picture, as it's only 480p to begin with. I'll experiment more, now that y'all have helped me understand what's going on a bit better. Cheers!