Felt, per my experiments if I remember correctly PJPEG with 99% quality produce about 2x size vs. H.264 and visual quality about the same. PJPEG with 100% quality I don’t remember exactly but kind of 6x (like 2GB per minute). And not much visual difference even comparing frame by frame. Of course it depend on content. If content is partially static (like most footage) then H.264 efficient, if a lot of full screen changes then yes H.264 may have mor visible artifacts. Anyway 2GB per minute is huge for stock footage and not so many PJPEG footage here with such bitrate.
So I agree if to compress with q100 pjpeg better for all content. Otherwise I think optimum quality/size is H.264.
It’s interessting discussion and it’s repeating from time to time. I will try to find time to do few footage encodings and post full-hd screen shots to some kind of blind test.
Like, I say, to the casual observer with naturally shot footage, h.264 can look good for its data-rate and for me, it’s a very impressive codec, with little else in its league. But you’re right - I’m talking in absolutes here. Compared to its lightly compressed brethren, h.264 is visually muddy and swimmy, the chroma blocks are huge and update intermittently. In shots that contain little movement for instance, you’ll often quite clearly see the jump as the pixels update. What’s more, h.264 is terrible at dealing with unsteady lighting conditions - flickering flames for instance. There it becomes an absolute festival of banding. And for any very clean digital gradients, you can forget h.264 completely. Where the real lack of quality in h.264 becomes apparent though is in the post-production process. Putting a strong grade on footage will often make the hidden quality lack visible. And pulling a chroma key on an h.264 shot blue-screen for example is an absolute nightmare. The chroma resolution is so low, that you’ll rarely be able to get a good key.
I do understand your point, but just to be completely clear. While h.264 provides an excellent quality to data-rate ratio (better than most other codecs all told), if you put all issues of data size aside and talk about image quality at an absolute level, the best quality possible with h.264 compression simply doesn’t come anywhere near to the best quality possible with ProRes or PJPEG. The latter codecs are in a completely different league.
In high end broadcast (i.e. commercials) and feature film, they often won’t even use ProRes or PJPEG (although most digital cameras record a compressed image). There they’ll often use 4K 32bit dpx sequences as an intermediate format, which retains all original information at very high bit depth, with no generational loss. The cost? 1GB per second
From the point of view of $5 stock though - I guess you need to weigh up all factors. And I guess in many instances that h.264 will provide the optimal solution.