The compression argument.

roxics

roxics

Veteran
I've heard this argument used over and over again. Something like
"It's hard to tell the image quality because of the compression of... youtube/vimeo/netflix/etc."

But if this is the new world we live in, where most people are only going to see your work in this way, then does it really matter if the out of camera quality is better?
If they aren't going to see it in a theater or buy a blu-ray or download a much larger less compressed file, then the only image quality that actually matters in the one people are actually seeing and using compression as an excuse is not a very good excuse.

What do you think?
 
roxics said:
But if this is the new world we live in, where most people are only going to see your work in this way, then does it really matter if the out of camera quality is better?
Click to expand...

Better in = better out. That's also true of compression.
 
David Jimerson said:
Better in = better out. That's also true of compression.
Click to expand...
Hmm, yes and no! Sometimes camera records image to clean, so in some parts distracting macro blocking artefacts from compression happen. To fight that people put grain back on the image. But in general definitely yes!
 
David Jimerson said:
Better in = better out. That's also true of compression.
Click to expand...

I think maybe I phrased this wrong as I'm not trying to say that you shouldn't start with the best you can get. I'm all for that.
What I'm trying to say is, at what point is the compression excuse invalid when judging quality because we live in a world where most people will see it compressed and how youtube/vimeo/etc. compress it is beyond our control.

Especially when new cameras come out I often hear "It's hard to tell because of the compression." But if the videos you are making are going to ultimately end up on these sites anyway, then what you're seeing is what the end user will see.
 
Can you really judge the quality of an extra-virgin olive oil after it's been cooked into something? No, you have to taste it straight from the barrel. The end-product may be satisfactory to you, but how do you know that the oil will hold its own when it really needs to?

How does the codec hold up to post-processing? Does it fall apart in color correction? Color grading? Is chroma-key going to have nasty artifacts? Is the noise from YT compression, or did it come from the camera?

Know your source. It cannot get better than what comes from the camera (diminishing returns).

Sure, if all you shoot for is the Web and you aren't personally that concerned about the differences between the source files and the uploaded and compressed end-product, go for what suits you. Most any camera can be made to look good if the images are properly lit and exposed.

Fact is, the 'Net may be the most prolific distribution platform these days (after all, anyone at any experience level can upload anything at any time with no jury and no editorial control), but there are still plenty of us out there who still shoot for broadcast, theatrical screenings, film festivals, etc. There are still plenty of clients out there who are interested in quality. It may look fine at a compressed 1080p on YouTube, but how does it look upstream? Will it suit both our needs and our personal standards (and those of the client)?

As for online distribution, y'know what stands out above all the noise on the Internet? The stuff that maintains the same level of quality as if it were for broadcast or big screen. "Good enough" may well be just that, but "better" still counts for something.

It's kind of like watching gear reviews on YouTube (of which there are far too many, and only a very few that are worth a flip). If I'm looking for user experiences on a new lighting fixture and I start watching a review where the production lighting sucks, I'm going to click away. A review of audio gear with a scratchy mic or, worse, on-cam mic or stereo hanheld recorder? I have no use for that person's review because their production reflects a piss-poor baseline.

Last, and probably most importantly, Internet distribution may be where most people will see the video, but those viewers aren't the ones paying for the production (unless you're one of the few people who have been able to monetize your YouTube videos to the tune of 6 figures a month). The client who pays for the initial production is the one for whom the quality is most important. And if, again, you're shooting only for the 'Net and you're working with essentially no budget, and monetization isn't happening, you go with what suits you and what you can afford. That's hobby-level, which is perfectly fine for what it is.
 
Last edited:
roxics said:
I've heard this argument used over and over again. Something like
"It's hard to tell the image quality because of the compression of... youtube/vimeo/netflix/etc."
Click to expand...

Well, no. In most cases where someone make that statement, the point of interest is 'pixel peeping', and as such, the person would be better to have a bulk archive site and the 'high quality' material downloaded for viewing.

I've been watching some amout of Netflix, and depending on the connectivity quality, most of the material is 'far superior' to the non-pro offerings on Youtube, et al.

Where most people seem to fail is, in terms of image quality, is the original set design, camera operation, quality of light.

Start with 'quality' there, and achieving a good looking Youtube clip is a matter of correctly encoding the media for the viewing platform.

Connection quality also plays into the viewing experience, and so, if one is targeting iPhone displays, and someone 'blows it up' to 1080 on their 50" display and sees 'crap'... well tough.

On the other hand 1080 streaming to the iPhone is wasted 'quality', unless of course the viewer is hooked up to a large display...
 
The thing is that compression algorithms may change (and WILL CHANGE!), thus it's a valid thing to judge source files not compressed version. It does not happen often but at this moment in time we're on transition phase. Today footage from camera A and B on youtube may look almost identical but tomorrow (i.e. when they'll switch it for h.265) it might be night and day. They talk about h.265 and vp9 for a looong time but because of some issues (money of course) switch was halted. But really it can happen any day.
 
Last edited:
For me compression is the enemy in original footage if you want to exercise significant creative control over the look of the footage in post. This is especially true of sub sampled color spaces lik 4:2:0 and any compressed codec under about 200 Mbps.
I recently encoded some D16 uncompressed raw to uncompressed 10 bit 4:2:2 broadcast AVI (800mbps) and after editing exported 10 bit 4:2:2 XAVC (200mbps) as a test. I was surprised at the loss of depth and subtley in the tonal scale, especially skin, compared to the uncompressed original for even that modest bit of compression. What looked like real skin with all of its variances in color, translucency and texture in the uncompressed original had already devolved to a flatter "skin tone" look common to most video cameras these days. Still very good looking image mind you, but it was like listening to an MP3 audio track after listening to the same from an SACD. By the time this gets compressed to 8mbps streaming video, it looks more like a cartoon than the real thing.
 
Obviously we all have different experiences.
My doc is going through film festivals right now and sometimes it screens at HUGE screens. Shot on Canon 7D with some footage in slow mo shot at 720p stretched to 1080p. H264 codec.... Looks fabulous even on mega screens.
What does matter is how well your codec interacts with some problem material. For instance our highly compressed source (i said before: h264 from 7D) looks great when encoded to DNxHD X but not so great when encoded to DNxHD or ProPres444. Some shots with tiny gradations simply fall apart. Now, we all know that Alexa shoots to ProRes and we all know the results is fantastic but it's because the codec simply encodes contents of the sensor. Period. If the source material for your very high end codec is less then perfect = the results might vary.

But the above is just a side note. The cameras/codec battle is over as far as i'm concerned. ALL cameras shoot fabulous image - today, once again, it's about the person behind the lens.
 
And to quote a producer i shoot for often: i don't get all the hoopla over 4k, 13+ stops of sensitivity and raw recording, if i hire a DP and pay him a few grand/day i expect him to know how to frame (no need for 4k to punch in or re-frame), know how to light (no need for infinite latitude) and expose the footage more or less correctly so even highly compressed footage holds up to minor tweaks.

Of course it's a bit simplified view of what those technological advances gain us, but i kind agree with him.
 
To each his own. I loathe 8 bit 4:2:0 H.264 in all its flavors. Been dealing with it in live streaming applications since before it was ever used as an acquisition format for camera originals. It has no depth, character, or subtlety in its color and tonal values. Everything has a flat lifeless airbrushed quality to it. There is no tolerance or room for creative grading since 90% or more of the data the sensor captured has been discarded in the encoding process.
You can take original wide dynamic range footage from a raw camera or from a high end high quality codec like the better grades of Prores and grade final images to be delivered in H.264 that you could never capture with an H.264 low end camera as an original.
Even a cheap camera like the BMC pocket can capture raw images that blow a low end DSLR shooting H.264 away.
It is not about fixing poor shooting choices in post. It is about having incredibly rich image data beyond what most TV displays can even reproduce as a palette for creative aesthetic choices in post. With a raw uncompressed camera, you make the choices about tone, texture and color values to include in the final product with a very refined degree of control.
With compressed video cameras, most of those choices are made by the camera software engineers and you have very little say in the result beyond a handful of menu settings they provide. You pretty much have to live with what comes out of the camera other than "minor tweaks".
 
Last edited:
Compression artefacts are easily separate from camera issues most of the time. If things like, highlights, CA, smear, colours and DR looks bad it's the camera or the way the footage have been treated. So yes the argument that YT compression makes the world of a difference is false.
 
Patryk_Rebisz said:
And to quote a producer i shoot for often: i don't get all the hoopla over 4k, 13+ stops of sensitivity and raw recording, if i hire a DP and pay him a few grand/day i expect him to know how to frame (no need for 4k to punch in or re-frame), know how to light (no need for infinite latitude) and expose the footage more or less correctly so even highly compressed footage holds up to minor tweaks.

Of course it's a bit simplified view of what those technological advances gain us, but i kind agree with him.
Click to expand...

Yes, "preserving options in post" is often fairly translated as "I don't want to think about this stuff until later."

If you can make all that extra latitude and resolution work for you, then it's for you.

If you're using it as a safety net, then you're not really doing your job well.
 
Patryk_Rebisz said:
And to quote a producer i shoot for often: i don't get all the hoopla over 4k, 13+ stops of sensitivity and raw recording, if i hire a DP and pay him a few grand/day i expect him to know how to frame (no need for 4k to punch in or re-frame), know how to light (no need for infinite latitude) and expose the footage more or less correctly...
Click to expand...

I actually like the way your producer thinks.
 
I find uncompressed raw very liberating. It greatly simplifies the actual shooting process to film like fundamentals: exposure, composition, focus, movement, light.
All of the critical decisions about luts, log, cine, REC709, white balance, color spaces that are so critical to getting the best out of modern conventional video cameras are simply irrelevant while shooting. They are decided in a post grading environment either before or after the actual shoot where one has the proper monitoring and tools to make the best creative choices.
The cinematographer is then free to concentrate on what is in front of the camera instead of the camera itself.
 
Part of the conversation is also the "reality factor." In the past 3 years maybe 4 spots i shot were given proper grade. I was in color on 2 of those occasions (one of them because i happen to be in the studio while they were doing color and i simply hanged around). There is simply no time to involve DP anymore and so the color decisions are made arbitrarily in the post. I mean do you really want to let the colorist who wasn't on the set and might not know that you recorded wrong white balance on purpose to "fix it" for you? Moreover how many times have you seen badly colored footage? To get the most out of RAW you need time (to spit out the 1-light files), time to do proper CC in the post, talented colorist (or conscious that some"mistakes" are on purpose). If you are missing one of those elements your RAW "ready for any tweak in the world" footage will suffer. If you go back to my original posts back from the DVX100 days i would insist on the flat took to fuss around in the post. Since then many things have changed - one of them, the reality hit me: unless i get the look i want in camera now, it will never get there with current deadlines. EVEN on your personal projects, there is simply only so much you can do before you hit the deadlines for festivals submissions etc so having the footage as close to what you want from the get go is important.
 
Patrick_Rebisz: busy DP who must entrust his footage to someone else to color
Razz16mm: independent artist who works at his own pace and grades his own footage

--

On a related note, can anyone point me to any articles that explain or videos that show the difference between RAW and 10-bit log, say ProRes? I can't think of why 10-bit log would not be just as flexible to grade. The only thing I can think of, if it is indeed 10 bit and log, is that some sort of crazy noise reduction was done to the footage before it was encoded to ProRes. That might explain why something would be lost. But otherwise with 10-bit log you should be able to push levels all over the place without banding. And of course even with 8-bit linear you can change the colors to anything you want.

--

And finally back to the threadstarter's question: For me, the grade always makes a drastically bigger difference in online footage I've seen. When people talk about Youtube compression, to me it's like noticing a hairline paint scratch on a rusted-out, broken-down Ford Pinto (or on an otherwise pristine Ferrari Testarossa, whichever the case may be).
 
Last edited:
Patryk_Rebisz said:
. The cameras/codec battle is over as far as i'm concerned. ALL cameras shoot fabulous image - today, once again, it's about the person behind the lens.
Click to expand...


This. For the average audience any camera you buy today produces awesome footage, even compressed on youtube or vimeo. If you want your footage to look "cinematic" as opposed to say some sports footage, then that is a function of lighting, framing, wardrobe, set design more so than your choice of camera. If you are pixel peeping, ok sure, compression makes a difference, but your general audience does not pixel peep.
 
With a video camera you white balance and pick a lut from a menu of them available in camera. You can shoot the same way with a raw camera. Pick a predefined lut and white balance. It takes time to develop your own luts, but you can use those developed by others. If you know the characteristics of the lut like you know the settings of your video camera, you can expose to match and these days even load the lut to your on camera monitor so you see what it looks like while shooting. It takes no more time than setting up a video camera, it just happens in post rather than on set.
The biggest difference with raw is dealing with the massive amounts of data. My D16 projects eat 2TB drives for lunch. I use them in a hot swap drive bay like flash drives.
 
You must log in or register to reply here.
Back
Top