Hello again, folks!
I've been running some tests on the new Handbrake (v. 0.98), with the intention of upgrading the guide above (if necessary). In the end, I got a lot of data to share with you.
Let's review things step by step:
1. Source file:
4 minutes exactly, 1GB .MOV file, 1080p/24fps, 35,232 kbps bitrate, stereo sound - basically, the highest-possible definition you can obtain with the Powershot camera series, from the SX50 down to the Elph 300.
Note: As a matter of principle, I won't even look at cameras which can record HD video, but only in mono sound.
The footage was from an indoor tennis tournament, so it has quick panning, difficult lighting and some video noise.
Note: ISO 200 was used during filming, and White Balance was left to "Auto" - even though it would have benefited from a "Fluorescent" compensation. All in all, I think the footage should be quite representative for normal camera use.
2. Setup:
I installed Handbrake 0.98 on a machine with the following specifications:
OS: WinXP Service Pack 3
CPU: Intel Core i5-650 @ 3.20GHz (quad-core)
RAM: 3.46 GB (yes, this is a Windows 32-bit installation)
All in all, a relatively middle-of-the road setup, it's even using the integrated Intel graphics.
3. Settings:
Basically, most of what's written above (post #4) has remained unchanged.
Here are the refinements:
- Pick the "High Profile" preset.
- Choose .mp4 or .mkv as your resulting file.
(I personally prefer using .mkv, but to each his own... note that, if you choose .mp4 - which will actually produce .m4v video files! - you should also check the "Web optimized" box, which allows you to stream the resulting files. It adds just a few seconds to the end of the encoding process to write metadata for faster streaming, and makes no change to the quality or file size.)
- Set Anamorphic to "none" in the "Picture" tab.
- Turn off decomb and detelecine in the "Video Filters" tab (the .MOV files are neither interlaced nor telecined); set Denoise to "weak", in that same location.
- Click the "Video" tab. RF 20 is set by default. You can go anywhere between 20 to 30, and a higher number indicates a higher compression rate (and a smaller resulting file). You can try RF 25 for starters (more on that later, when we discuss the test results).
- While you're still in the "Video" tab, change the default setting from "Variable Framerate" to "Constant Framerate". This will ensure better playback compatibility, and reduce the risk of dropped frames on some software/hardware combos.
Further encoding optimizations are as follows:
- Click on the "Audio" tab. Choose AC3 (ffpmpeg) in the second drop-down menu.
(AC3, also known as Dolby Digital, is the most widespread audio format for home theatre, and my personal choice... feel free to use AAC or mp3, if that's your game! Just don't select any of the "passthru" options, since your source file has a different audio format, so some form of transcoding WILL be necessary.)
- Your original audio is stereo only, so choose "Stereo", "Surround" or "Dolby ProLogicII" from the third drop-down menu, depending on what your tastes are.
- By default, the audio bitrate is set to 160kbps. I find that to be a little on the skimpy side, the sweet spot should be around 256 kbps.
- Last, but not least, click on the "Advanced" tab. Change the Motion Estimation Method setting to "Uneven Multi-Hexagon". Leave everything else the same.
4. Test results:
I produced four different files using different RF values (the crucial setting for compression and resulting file size). Consequently, the machine required markedly different encoding times.
- RF 20 (25 minutes encoding time) - 339 MB size
- RF 23 (22 minutes encoding time) - 185 MB size
- RF 27 (18 minutes encoding time) - 92.2 MB size <-- 11 times smaller than source file!
- RF 30 (10 minutes encoding time) (!) - 62 MB size <-- 16 times smaller than source!
I played the resulting files in VLC, set in a looping playlist, with filenames enabled (allowing me to see which particular encoding scheme was used when I jumped from one file to another).
Subjectively speaking, although I knew that some differences should be discernible between the (enormously different in size) files, I was not able to distinguish between RF 20 and RF 23.
RF 27 looked "almost" as good as the lower two values - it *felt* a bit grainier, but that could be a psychological effect.
By the time I got to RF 30, however, some degradation was perceptible - occasional antialiasing and artifacting were present in high-speed movements, and some slight color banding occurred in background surfaces.
Make no mistake, however - it was still OK for a non-discriminating eye!
5. Conclusion:
The sweet spot for maximum size reduction with minimal quality loss seems to be in the RF 25 - 27 range (which produces files between one-sixth to one-tenth of the original...) Anything above that takes a visible quality hit (law of diminishing returns!)
If you're into archiving, and do not wish to edit the files any further (I know that .mkv can be a pain for non-linear video editing software, but I'm not sure if this applies equally to .mp4/.m4v), you can use a RF 25 value, to be on the safe side.
If you absolutely need to reduce the original video file to its smallest possible size (i.e., for easier transfers), you can go to RF 30.
. If you don't want that stuff (I didn't), you should step through the installation carefully. The installation also closed all my Firefox windows and tabs and restarted Firefox with its own window to get the installation going (odd, and pretty inconvenient to say the least, and a complete shock). It also wanted to associate all audio and video files with Splash Lite. I deselected the audio and a couple of the video extensions.
Facebook
Twitter