Is prores or h264 more efficient compression at prores bitrates

[ezcapper]

I know prores is better for editing, but is it more efficient compression than h264 at prores bitrates (105 - 200 Mbps hd video)? Which codec would typically be better for video quality for the original encoded file, or are they usually similar? I understand each case and implementation would be slightly different, but just seeing if there is a general conclusion. Thanks for the help.

[KarMa]

Probably x264.

[poisondeathray]

Quality/compression wise, definitely h264 , even if you mean intra encoded . Significant differnce

But prores usually performs better on the timeline (ie. navigation performance in an editor), especially on macs, even if you use "fast decode, low latency" settings for AVC. The same goes for "professional" AVC-I variants by Panasonic (e.g native AVC-I and ultra variants), they are very slightly more "sluggish" . It's perceptible

Prores and HQ variants are considered "noisy" when you pixel peep, until you get into prores 4444, and 4444 xq variants (ie. very high bitrates, higher than what you referred to). But to the "normal" person not involved in post production, or if you're just watching, you won't notice the difference - still a high quality codec

[SameSelf]

Some NLE/finishing sw are so inefficient at decoding h.264 footage (no matter how powerful the 'puter) that transcoding to ProRes (Mac) or DNxHD/HR (Win) is unavoidable. h.264 is considered a delivery format unsuitable for post. For example, can it handle timecode? I don't even know. Furthermore, as an acquisition format, h.264 is only encountered on consumer cameras. So is h.264 better? A moot question for professional workflows.

I archive all my ProRes/DNxHD footage (acquisition format) as shot. I have no desire to transcode it—if only for the simple reason that I want to avoid generational losses. If there are some other compelling reasons to transcode my ProRes/DNxHD footage to something else (e.g. h.264), I am all ears. Similarly, do I transcode my h.264 footage to ProRes/DNxHD for archival purposes? Of course not and for the same reasons. Typically the two are mutually exclusive. IOW, at least up until recently, I am not aware of too many devices that offer both h.264 and ProRes/DNxHD.

[poisondeathray]

h.264 is considered a delivery format unsuitable for post. For example, can it handle timecode? I don't even know. Furthermore, as an acquisition format, h.264 is only encountered on consumer cameras. So is h.264 better? A moot question for professional workflows.

Not entirely true ; the higher end professional Panasonic cameras use AVC-Intra / AVC-Ultra (up to 4K 12bit 444) . Its a very well established workflow on P2 cards . All pro features, timecode, metadata etc...

Generic garden variety h264 is 8bit and typically use lowish bitrates, but the AVC spec is very wide. It even includes lossless varieties (something even prores 4444 xq cannot claim), up to 14bit , higher than the max 12 supported by prores 4444 xq (but good luck finding support for 14bit AVC)

You can disable some AVC features to make it decode faster, but then you start to lose the compression advantage

[SameSelf]

pdr, do you come from the broadcasting world (I have always been curious about where you acquired your expertise)? I forgot all about the P2 format which is nearly dead, right?

[poisondeathray]

P2 used to be very popular for broadcast, documentaries. Yes it's less popular these days

But don't forget Sony's XAVC , which is basically an AVC-Intra/Ultra clone . Same deal , up to 12bit 4K 444. The higher end Sony's use them . Very popular especially in Europe for all types of shoots. There are lower end 8bit XAVC variants in consumer camcorders as well. But there well established professional workflows as well for the pro XAVC variants including clip management, metadata, timecode etc...

The point is not to think all AVC is the same, there are many varieties, very high (lossless) to very low qualities; very high low latency performance to very high compression , high latency performance (heavily compressed, very long GOP's even infinite)

[SameSelf]

Good point. I am well aware of XAVC MXFs (the lower end cameras use XAVC-S MP4) which is a better example versus Panny's aging P2 format. I am not sure I understand the underlying intent of the OP's question, but these are all in-camera acquisition formats, right? And I know that at least for Resolve, even XAVC/S footage needs to be transcoded for efficient decoding. I know being a DR user heavily colors my perception (pun intended), so I am wondering under what circumstances would one choose to transcode footage to h.264 (even if it is 12-bit 444 Intra-only) unless for some workflow conformance reasons? Otherwise, this discussion sounds more like, "which camera should I choose?"

[poisondeathray]

Good point. I am well aware of XAVC MXFs (the lower end cameras use XAVC-S MP4) which is a better example versus Panny's aging P2 format. I am not sure I understand the underlying intent of the OP's question, but these are all in-camera acquisition formats, right? And I know that at least for Resolve, even XAVC/S footage needs to be transcoded for efficient decoding. I know being a DR user heavily colors my perception (pun intended), so I am wondering under what circumstances would one choose to transcode footage to h.264 (even if it is 12-bit 444 Intra-only) unless for some workflow conformance reasons? Otherwise, this discussion sounds more like, "which camera should I choose?"

If you re-read the OP, he's only asking about 2nd generation compression efficiency. The answer already given is AVC. In long GOP, it wins by an astronomically large margin. But in typically used intra settings such as for editing, it still offers substantially better compression (Higher quality at a given bitrate. This is a measurable and observable difference.) . The main tradeoff for higher compression is editing responsiveness . I would also argue in some programs AVC can have compatibility issues, although prores is far from problem free or issues.

[SameSelf]

I see and agreed. Prores is definitely not without its problems. But it looks like I have been maligning h.264 as an exclusively delivery codec for far too long. Time to repent.

[poisondeathray]

For Windows users considering prores or dnxhd as in intermediate, I would also look at cineform . At equivalent bitrates, the quality is similar but the decoding performance is much better. Prores works better on a Mac. Even custom 3rd party PC Prores decoders aren't as fast as native Mac PR decoders. For picky programs like resolve , cineform is also supported, and the AVI variant is supported on all windows programs

[SameSelf]

You should head over to the DR forum. All those users ever recommend is PR or dnxhd, gets a little tiresome.

[Cornucopia]

I also prefer the quality of cineform, as it is a wavelet-type codec, so it has less inherent blocking artifacts, and it is inherently efficient through VBR, compared to some of the other editing codec options. But until Cineform gets a new life as implemented through VC-5, it probably isn't the best choice for continued sharing.

Due to Apple's closed-system structure, it is extremely unlikely that they will develop their own codec for VC-5 (and they still have not, and probably will not, allow 3rd party codec dev for AVFoundation apps). That basically means that on a Mac - ProRes. See link: http://www.divergentmedia.com/blog/dealing-with-codecs-on-modern-macs/.

Scott

[_Al_]

Also guys do not forget a volume of users that prepare all sorts of "internet broadcasts", videos that needs to be backed up for later, lots of videos, gathered from all kinds of sources, even restored a bit. It needs to be deinterlaced, prepared for timeline editing and stored. In those kind of productions, so to speak, even Prores is too big. You decide what to store and change all kinds of footage , say SD 25i,30, HD 25i, 30i to say 60p (as a format for future) and convert videos to H264 CRF17-18, very short GOP, low ref frames, not resizing, just making sure aspect ratio is included, settings with small latency and convert videos to those archived H264. NLE cannot handle those transfers (frame rates, interlaced), results are usually pretty bad so it is done outside of NLE. Edit those converted , say 60p and finally export to different resolutions, or half frame rate, or archiving those again as 60p H264 as above.

I want to stress that there is lots of internet stuff today, not really broadcast only, where handling 120Mbit/s archives is too much and those H264 suffice anyway because end product is internet video.

[ezcapper]

Wow, thanks for such in-depth answers, very appreciated!