Most of today's amateur video is taken on mobile phones and stored as mp4. We happily take this and use it as footage for editing, and the quality is really good.
Why is it that crappy footage that we capture from old analog video becomes even crappier if we use mp4/h.264? Why does it make such a big difference capturing this already crappy video to for instance ProRes? I just made a test and it is obvious that the quality is much better when capturing to a near lossless codec and then using that for editing.
In the below test, "ProRes" the footage was captured in ProRes; while "VideoPro" and "Lars" was captured in MP4 at bitrate 10,000 and 5,000 respectively.
There is a clear difference.
+ Reply to Thread
Results 1 to 26 of 26
Prores bitrate is much higher than what you are using at 5-10Mb/s. If you used h.264 at the same bitrate as prores - it would yield typically ~1.2-1.5x higher quality in I-frame mode, 2-3x higher quality in long GOP mode than prores
There are many different "flavours" of h.264 . It supports everything from low quality to lossless. Multiple configurations like I-frame, 8-14bit depth, 4:2:0 - 4:4:4, even RGB . If you capture using low quality settings - you get low quality (no surprise)
Thanks. I just made a test encoding to max bitrate in handske, and that really improved the quality.
But become that video from my mobile looks very good quality even though it probably uses a pretty low bitrate?
Last edited by jagabo; 16th Aug 2019 at 20:20.
Also, the old crappy footage is probably...crappy . It probably has lots of noise and artifacts which impairs compression efficiency
For what it's worth, ProRes is designed not just for high image quality, but also ease of playback as an intermediate codec. Compared to AVC of similar IQ, it is less demanding of computing resources so is easier to work with during post production. AVC is more end-user friendly, and lossless (Lagarith, HuffYUV, or my preference Ut) is better for archival or if you're OCD (haha) like me as a wedding videographer, capturing ultimate detail.
But isn't it correct that not only mobile phones, but also digital cameras capture in h.264? And there the quality looks great. I googled that max bitrate for iphone is 400MB per minute.
I tried to encode the old video footage to H.264 with highest bitrate in handbrake, and the quality was indeed comparable to ProRes. And filesize was smaller. Is it just as good to archive (to used later as footage) in h.264 with a high bit rate? How high should it be?
Re-read my last post, which quoted the important reasons. That's why.
Another problem with H.264 is that it tends to err on the side of softness.
Handbrake is probably one of the lowest quality conversion tools available. You're definitely losing details, and plasticizing SD video, at the default settings..
HandBrake so would like, if you have time of course, to expand on that statement.
Are you referring to its core/encoder or more to do with its presets/defaults/setup? (I am actually really interested in your answer/points of view).
Compare Handbrake to the options available in MainConcept or selur's Hybrid.
Handbrake is, and always has been, a stripped-down dummy-friendly x264 GUI for the masses. But it's not really quality. It has many quirks and bugs, and even trying to use it for the simple purpose of DVD>MP4 can be frustrating at times.
It is what it is.
A CRF slider doesn't at all address the settings/options that I refer to.
I use MeGUI because it's especially slick for my videography workflow, but HandBrake is no slouch. I think it gets a bad rap because at first glance, it's geared more toward speed and ease of use than quality or filesize (especially for neophytes only using presets), and things like the aforementioned quality slider might seem ambiguous or gimmicky to some of us accustomed to working with concrete values. But it's also highly configurable and can do most if not all an advanced user might want, once you learn to fully use it.
For what it's worth @malling, it's true that h.264 is commonly used in consumer and prosumer grade cameras; my Sony A7III for example records 8-bit AVC up to 100Mbps (which can really be frustrating). But true professional cameras and external recorders utilize ProRes, lossless, raw, etc. Not because it's visually superior straight out of the camera, but because it provides more latitude in post production—you can push and pull the footage around with less quality degradation. In fact, this functionally has even been hacked into cameras like my Canon 5D3 by users, to enable raw video capture.
Also, HandBrake defaults are terrible, e.g. a noob would end up changing the frame rate of all their encodes by using ANY of the presets...
Not only does it force 30, 25 or 60 FPS but VFR by default.
And yes as mentioned above, it softens videos...at least it used to.
Maybe zones or qpfile ?
ffmpeg libx264 can do zones now with -x264opts , but not sure about using a qpfile
I don't really use handbrake, so not sure if zones work. But that's a very useful feature for people that are more than just casual users
I most of the options can be specified in the extra commands box, just like you can with megui . But megui uses x264cli , so it definitely has access to everything, even if it's not a checkbox or slider in the GUI
What handbrake lacks is some of the preprocessing filters available through avisynth and vapoursynth that those other GUI's like megui, handbrake, hybrid have
However, that aside, I do agree it's not something that's easily apparent to a user, the preset system could be improved.
Thanks, everybody, so much for your input. The conversation has indeed made me understand a little more, and I am now experimenting. As this is just for personal use, digitising my old family home movies, handbrake suits my level pretty well, even though there may be more advanced (and hence much more complicated) tools out there.
However, I'm a little confused by the quality settings in Handbrake.
1. Should I use "Constant Quality" or Avg Bitrate to encode uncompressed captured by Svideo from a DV camcorder? In the documentation they recommend the former.
2. In "Constant Quality" the quality should become higher sliding to the right, with smaller RF values. But the documentation recommend higher RF for higher quality:
Recommended settings for x264 and x265 encoders:
RF 18-22 for 480p/576p Standard Definition1
RF 19-23 for 720p High Definition2
RF 20-24 for 1080p Full High Definition3
RF 22-28 for 2160p 4K Ultra High Definition4
1. "Constant Quality" is usually recommended.
2. Lower RF (slider more to the right) = higher quality
The documentation doesn't "recommend higher RF for higher quality". It basically says you can get away with a lower RF value the higher your resolution is. Higher resolution doesn't imply higher quality. (e.g. how a 1080p web streaming video can look much worse than an old 480p DVD).
There's no single setup that's best for all source and all purposes. A general compromise setup for viewing DV sources (as mentioned in one of your other threads) try something like the HQ 480p (NTSC) or HQ 576p (PAL) preset. Apply the Yadif deinterlacer with Preset Bob. Select the h.264 (x264) encoder. Set the frame rate to 2x that of your source (59.94 for NTSC, 50 for PAL). Set the Encoder Preset to Slow, tune for Film, Quality to RF=18. Use lower RF values for higher image quality (bigger files) or higher values for smaller files (lower quality)
Jagabo, thank you do much. Very helpful indeed!