The good bitrate question

[kb1985]

Quite long time ago I asked you "btw" what bitrate you suggest for compressing DV video to XVID 720x548 ? I haven't received any replies to my "btw", but you helped me solve the main problem. Now, this is the big question: "is 2000 kbps enough for 720x 'shaky' xvid video?" For tv captures it is perfect, but what about "shaky" home video? I've experimented with some settings and the results are various. I have some thoughts but would like to hear your opinion first. 1500 kbps is definitely not enough, 12 000 kbps is too much. What are your suggestions?

BTW and I'm also not sure if I get better quality on creating the home video movie on DVD 4,4 GB or making it into 1500 MB XVID? 4400 > 1500, so probably DVD should give me better results but maybe it's also matter of compression algorythms?

[junkmalle]

If you don't need files of a specific size use quantization (constant quality) mode and stop worrying about bitrate. Run a few tests to see what quality level you're happy with then use that for all your videos.

I usually set the Q value to something between 2 and 3. At 2 it's hard to see differences between even with still frames. At 3 you can see differences in still frames but they're not noticable at normal playback speed. Depending on the source the final average bitrate may turn out anywyere between 600 and 6000 kbps.

[kb1985]

I don't need a file of specific size, but I would prefer not making files bigger than 2 GB. I thought about creating files of 1400 MB or 1500 MB size and putting three of them on one DVD (but i'm still not sure if three 1500 MB files will fit on one DVD). I've just compressed the 2 pass xvid test file. It had 1500 MB for 60 minutes and over 2000 kbps. The result seemed ok but AFAIK it depends on kind of video footage. Isn't the "quantizer" method something like waste of space, because it uses much HDD space for encoding even very simple parts of the video? I'm really confused what to do now. Maybe XVID is not the best choice at all? Maybe creating DVD is the best solution? However it's not very handy for me, because I enjoy cutting the video and xvid seems easily editable than DVD (no need to rip it everytime when I want to edit something).

[junkmalle]

Isn't the "quantizer" method something like waste of space, because it uses much HDD space for encoding even very simple parts of the video?

No, you're describing Contant Bitrate. Quantizer mode is Constant Quality -- the codec uses whatever bitrate is needed to get the desired quality at each and every frame. If few bits are needed few bits are used, if a lot of bits are needed a lot of bits are used. It only requires one pass but you don't know beforehand how big the resulting file will be.

Maybe XVID is not the best choice at all? ...I enjoy cutting the video and xvid seems easily editable than DVD (no need to rip it everytime when I want to edit something).

XVID is not a good choice if you want editability (unless you use all key frames with much higher bitrates). MPEG has better editability because it has more keyframes than your usual XVID file. For archiving you don't need to convert to VOB files and master, you can leave your files as raw MPEG. You may even find that your DVD player will play MPEG files on ISO data disks -- many do.

The best editability comes with codecs that use only intra-frame encoding (compressing each frame as a standalone image, rather than encoding the differences between frames). But you won't get anywhere as good compression. Here are some examples:

Uncompressed RGB: 112 GB/hr
HuffYUV: ~20 GB/hr
DV: 13 GB/hr
Motion JPEG: ~10 GB/hr

Note that HuffYUV and MJPEG have highly variable compression rates. The numbers given are very rough estimates. Especially with MJPEG where you can specify the desired quality (I used PicVideo MJPEG codec at setting "19").

[kb1985]

[quote="junkmalleNo, you're describing Contant Bitrate. Quantizer mode is Constant Quality -- the codec uses whatever bitrate is needed to get the desired quality at each and every frame. If few bits are needed few bits are used, if a lot of bits are needed a lot of bits are used. It only requires one pass but you don't know beforehand how big the resulting file will be.

So when I use quantizer 1 pass I will have even better results than after normal 2 pass? I'm not sure if i shoudl chose quantizer=2 or variable bitrate max 9500/average 8100/minimum 5000.

[junkmalle]

So when I use quantizer 1 pass I will have even better results than after normal 2 pass? I'm not sure if i shoudl chose quantizer=2 or variable bitrate max 9500/average 8100/minimum 5000.

It will depend on the material. Single pass Quantization mode at Q=2 may be better or worse than 2-pass VBR at any particular setting. The point is: you know exactly what the quality will be in Q mode, but you don't know the size; in 2-pass VBR mode you know what the size will be but you don't know the quality.

So use 2-pass VBR when you need a file of a certain size. If you want a file of a certain quality use single pass Quantization mode with Q set to the quality you want.

You should perform some experiments yourself but I find that at Q=2 the result is almost indistinguishable from the original video even if you examine still frames with 4x or 8x enlargement. Q=1 is pretty much overkill but you might use it for archiving. At Q=3 you can see some loss of quality (a little macroblocking, loss of some small details) in enlarged still frames but you don't notice them at normal playback speed. You can use decimal values too (eg 2.5, 3.389822) if you want something in between.

In general, if you pick a Quantization value that results in the same size file as a 2-pass VBR encode, the files will be nearly identical in quality. If you single step through the video you will see some frames that look sligtly better in one, some that look slightly better in the other.