Bitrates. Where do you draw the line?

[IronFilm]

3 hours ago, M_Williams said:

Nikon's response to RED's lawsuit is a legitimately valid argument and stands a good chance of invalidating the patent. Why no one else ever used that argument (except JinniMag) is beyond me. And notice how RED quietly dropped their lawsuit against JinniMag.... no deals were made there.

RED's patent never should have been granted in the first place. But Nikon has a strong argument legally as for why it's not valid.

But JinniMag shut down two years ago? And their videos have been pulled down. (did anybody save copies of them??)

[Alexis Fontana]

My observations:

(Always shooting log)

h264/265 has too much sharpening and macro blocking - the latter especially when trying to push a vibrant photochemical look with deep film like tones. Blue and red gives up before I’d like em to. Don’t like the idea that a tiny insufficient computer does heavy compression in camera. Especially noise reduction -  in camera noise reduction needs to fuck off and die  

On Red cameras I rarely go above 8:1

BRAW rarely above 5:1

On Arri cameras I mostly shoot Prores 444 - sometimes 444 XQ. Never ArriRAW as the file sizes are too large for the kind of jobs I do (also very close to diminishing returns)

Love the Sony Venice 16-bit X-OCN(compressed??), it’s just insane how much latitude it holds.

 

422 LT - I only think about it as a deliverable format, have no idea how it looks from a Fuji camera. I think the idea of shooting 8k on a s35/LF size sensor is ridiculous, the file sizes becomes too big and hence you need to accept more compressed codecs which defeats the purpose of the so called “high resolution”. Don’t really know much about the 12k ursa but it’s another kind of sensor so not comparable right?

 

I own a pocket 6k pro, love the camera but would prefer larger photosites and 4K.

[TomTheDP]

1 hour ago, Alexis Fontana said:

My observations:

(Always shooting log)

h264/265 has too much sharpening and macro blocking - the latter especially when trying to push a vibrant photochemical look with deep film like tones. Blue and red gives up before I’d like em to. Don’t like the idea that a tiny insufficient computer does heavy compression in camera. Especially noise reduction -  in camera noise reduction needs to fuck off and die  

On Red cameras I rarely go above 8:1

BRAW rarely above 5:1

On Arri cameras I mostly shoot Prores 444 - sometimes 444 XQ. Never ArriRAW as the file sizes are too large for the kind of jobs I do (also very close to diminishing returns)

Love the Sony Venice 16-bit X-OCN(compressed??), it’s just insane how much latitude it holds.

 

422 LT - I only think about it as a deliverable format, have no idea how it looks from a Fuji camera. I think the idea of shooting 8k on a s35/LF size sensor is ridiculous, the file sizes becomes too big and hence you need to accept more compressed codecs which defeats the purpose of the so called “high resolution”. Don’t really know much about the 12k ursa but it’s another kind of sensor so not comparable right?

 

I own a pocket 6k pro, love the camera but would prefer larger photosites and 4K.

The Venice 2 has 3 different RAW compression options. 

X-OCN XT which is 965mbps 

X-OCN ST which is 660mbps 

X-OCN LT which is 389 mbps 

The RAW LT is even smaller than 4k prores 422. 

I am not sure if the older Venice has the same RAW options or not. 

I have been shooting 2k 422 a lot on my Alexa Classic. I still get the amazing highlight retention and I am not usually pushing the footage too far. It is nice to have 444 as an option though. Arri Raw in 2.8k wouldn't be terrible but it edits terribly. By far the most choppy footage I have ever encountered. 
 

[Django]

On 10/19/2022 at 11:12 PM, BenEricson said:

The motion is kind of all garbage on the C70. It might be the DGO. I don't know. It feels a little funky.

Is there any tangible proof concerning  this? I've only really noticed the footage is a little choppy when the camera is moving a lot handheld but that's kind of to be expected with no mechanical stab.

Have you tried Canon Raw on the C70? From the test footage I came across the motion felt really nice, well on the normal side of things.

Getting back to codecs, I still think the C300 mk2's 12-bit 444 2K option is the nicest option. Real shame no other Canon cam seems to have that particular one.

[M_Williams]

On 10/22/2022 at 2:43 AM, IronFilm said:

But JinniMag shut down two years ago? And their videos have been pulled down. (did anybody save copies of them??)

Huh? I assume you mean the Mags he was making. The videos were never pulled. YouTube channel is still up, vidoes have been on there for 2-3 years it says. Maybe they were reposted, I can't remember when he originally did them but it seems longer than 2-3 years ago...

But yeah, here's the REDCODE video, the rest are on his channel (Jinni.Tech). Nikon used his arguments in their response to RED because they appear by all accounts to be valid - I've read/watched several patent lawyers say Nikon is correct about the filing date of the patent. You're not entitled to a patent if the invention was described in a printed publication or in public use or on sale more than a year prior to the date of the application. RED filed the two patents on on April 11, 2007 and Dec 28, 2007, but Jared Land posted specs in 2005 and demonstrated a prototype at NAB in April 2006 and took hundreds pre-orders. There's also a video of an early prototype in action in early 2006. The patent filed in April 2007 doesn't describe key parts of the REDCODE patent, so the earliest date that can be applicable is December 28, 2007 - more than a year after they took pre-orders and demonstrated the camera.

 

[IronFilm]

10 hours ago, M_Williams said:

Huh? I assume you mean the Mags he was making. The videos were never pulled. YouTube channel is still up, vidoes have been on there for 2-3 years it says. Maybe they were reposted, I can't remember when he originally did them but it seems longer than 2-3 years ago...

ah my bad, I was just looking up some of his videos, such as: https://youtu.be/qv7phUqoIE4 , saw "Video unavailable. This video was removed due to a legal complaint" and didn't think to check if his other videos are still up. 

Although if you look at this latest video:

Years later and "coming soon" still hasn't happened yet. 

But anyway, his store is definitely closed, you can't order anything: http://jinnimag.com/store.html#!/JINNIMAG/c/33340804/offset=0&sort=nameAsc 

 

 

[The Dancing Babamef]

I think it's great that they brute force the quality with "excess" bitrate rather than have you suffer from macro blocking at the highest quality 422 HQ. IF you shoot LT then you probably know the risks.

[kye]

54 minutes ago, The Dancing Babamef said:

I think it's great that they brute force the quality with "excess" bitrate rather than have you suffer from macro blocking at the highest quality 422 HQ. IF you shoot LT then you probably know the risks.

This raises an interesting question for me about what role resolution plays in the quality of Prores.  

Prores bitrates scale proportionally with resolution, so 4K is ~4x the bitrates of 1080p because there are ~4x the pixels.

I would assert that 2K Prores HQ is "sufficient" (not perfect, but sufficient) as countless feature films were shot in 2K/1080p Prores HQ (at ~180Mbps) and were projected in theatres worldwide on the largest screens available (short of IMAX), so the macro-blocking can't have been too bad.  
So then, if we're talking 4K, Prores HQ is ~700Mbps and LT is ~330, which is almost double the 2K HQ bitrate.  If you film 4K you don't project it onto a larger screen just because you have more bitrate (plus there aren't really many screens larger than real cinemas anyway), so even if the macro-blocking is larger from LT in terms of how many pixels wide the artefacts are, the fact that there's more bitrate for the whole image, surely the artefacts would be less visible than on 2K HQ?

Does anyone know how this comparison actually fares in the real world?  I've never tested this particular aspect.

[Django]

16 minutes ago, kye said:

I would assert that 2K Prores HQ is "sufficient" (not perfect, but sufficient) as countless feature films were shot in 2K/1080p Prores HQ (at ~180Mbps) and were projected in theatres worldwide on the largest screens available (short of IMAX), so the macro-blocking can't have been too bad.  

If you're referring to the Alexa Classic, 2K ProRes HQ was closer to 300Mbps and it could shoot up to 430Mbps in 4444.

Second gen Alexa XT really changed the game though with ARRIRAW /Prores XQ and resolutions up to 3.4K in OpenGate.

So I'd say "sufficient" by 2010 theatrical standards but ARRI IQ took a major bump as soon as the XT came out around 2013. 

Besides I heard there were software upgrades to give the Classic 3.2K/3.4K XQ/Open Gate so those bitrates and resolutions may have been used much earlier than 2013.

Long story short I don't know if that many features were actually shot in 2K ProRes HQ..

[Django]

On 10/22/2022 at 5:20 PM, Alexis Fontana said:

My observations:

(Always shooting log)

h264/265 has too much sharpening and macro blocking - the latter especially when trying to push a vibrant photochemical look with deep film like tones. Blue and red gives up before I’d like em to. Don’t like the idea that a tiny insufficient computer does heavy compression in camera. Especially noise reduction -  in camera noise reduction needs to fuck off and die  

On Red cameras I rarely go above 8:1

BRAW rarely above 5:1

On Arri cameras I mostly shoot Prores 444 - sometimes 444 XQ. Never ArriRAW as the file sizes are too large for the kind of jobs I do (also very close to diminishing returns)

Love the Sony Venice 16-bit X-OCN(compressed??), it’s just insane how much latitude it holds.

 

422 LT - I only think about it as a deliverable format, have no idea how it looks from a Fuji camera. I think the idea of shooting 8k on a s35/LF size sensor is ridiculous, the file sizes becomes too big and hence you need to accept more compressed codecs which defeats the purpose of the so called “high resolution”. Don’t really know much about the 12k ursa but it’s another kind of sensor so not comparable right?

 

I own a pocket 6k pro, love the camera but would prefer larger photosites and 4K.

Good observations. I agree that h264/h265 can really kill footage with sharpening & NR done in-cam.

But it really depends on camera brands and models. Even within Sony, the FX3/A7S3 has tons of NR you can't adjust.

The same sensor on FX6 is a lot more adjustable and on the new FX30 the same XAVC codecs display much less NR applied in-cam so it seems Sony listened to feedback.

Canon allow you to fully turn off high ISO NR. And higher end models have internal compressed RAW.

On Fuji you can reduce NR & sharpening but not fully turn off. Now with ProRes I assume there is a less processed image but its still a baked in format so not sure of the specifics. 

All-in-all prosumer hybrid cameras are getting better with XH2/XH2S/GH6/Z9 taking the lead with internal Prores. Followed by Canon with internal RAW. Sony is really still in last position with its heavy NR induced h26x codecs. 

[The Dancing Babamef]

Yes. but I also think that having high bitrate for 10bit 422 will account for the missing information in color, comparing to 444. But all of this circles back to the use case: are you filming a talking head or in a forest with lots of trees and moving micro elements that take up the bitrate so even thou the files are large it's justified as the utilization is high. 

But just like Andrew said in the original post: "- ProRes 422 HQ in GH6 and Nikon Z9 is too big, not practical in my opinion - may as well shoot raw?" 

 

Go watch DpReview's episodes and ProAV TV. Their shows might as well be audio only except the actual image comparisons. 

[Phil A]

I'm shooting H.265 200 Mbit on the X-T4 and used to shoot 150 Mbit H.265 on the GH5.

I fully accept the weaknesses of those codecs, but I create only content for social media or personal usage so there is absolutely no point in recording in higher bit rates or codecs. Most people are obsessed with technical image quality when in the end, anyway no one cares about their results (cat video on YT with 7 views).
The only moment where I might use a higher bit rate option is if I record something with water, because that's always a disaster in LongGOP.

I always used to transcode to ProRes and wanted a camera that can record ProRes because of my editing system, but since I switched to Apple M1 based hardware, everything is golden with H.265.

[kye]

4 hours ago, Django said:

If you're referring to the Alexa Classic, 2K ProRes HQ was closer to 300Mbps and it could shoot up to 430Mbps in 4444.

Second gen Alexa XT really changed the game though with ARRIRAW /Prores XQ and resolutions up to 3.4K in OpenGate.

So I'd say "sufficient" by 2010 theatrical standards but ARRI IQ took a major bump as soon as the XT came out around 2013. 

Besides I heard there were software upgrades to give the Classic 3.2K/3.4K XQ/Open Gate so those bitrates and resolutions may have been used much earlier than 2013.

Long story short I don't know if that many features were actually shot in 2K ProRes HQ..

What were they distributed in?  and what about the DI?  I ask because once something has had things like macro-blocking then there's no fixing it downstream.  I thought that a 2K Prores DI was common for colourists, and they'd deliver in 2K as well.  Not sure if that would have been 4444 or better though.

[Django]

1 hour ago, kye said:

What were they distributed in?  and what about the DI?  I ask because once something has had things like macro-blocking then there's no fixing it downstream.  I thought that a 2K Prores DI was common for colourists, and they'd deliver in 2K as well.  Not sure if that would have been 4444 or better though.

I'm talking acquisition from a DPs perspective.

As the following video from 2014 states, 80% of films back then where shot above 3K yet indeed 99% were finished in 2K.

That doesn't mean it was ideal as far as IQ as the guy explains nobody printed Super35 to Super16 which is sort of what was being done back then. Not to mention they were already anticipating 4K streaming (or "broadband" as they called it then lol). He goes on to explain/push the benefits of shooting in ARRIRAW at 3.4K with the then new XT and details a workflow where you'd transcode to ProRes 444 2.8K for a 50% gain in file size and almost 3K resolution. That would have definitely been the forward-thinking workflow:

 

[John Matthews]

10 hours ago, kye said:

I would assert that 2K Prores HQ is "sufficient" (not perfect, but sufficient) as countless feature films were shot in 2K/1080p Prores HQ (at ~180Mbps) and were projected in theatres worldwide on the largest screens available (short of IMAX), so the macro-blocking can't have been too bad.  
So then, if we're talking 4K, Prores HQ is ~700Mbps and LT is ~330, which is almost double the 2K HQ bitrate.  If you film 4K you don't project it onto a larger screen just because you have more bitrate (plus there aren't really many screens larger than real cinemas anyway), so even if the macro-blocking is larger from LT in terms of how many pixels wide the artefacts are, the fact that there's more bitrate for the whole image, surely the artefacts would be less visible than on 2K HQ?

But Kye, the human eye has evolved, you see. What was more than adequate in 2015 is no longer going to cut the mustard in 2022. Funny how things have changed.

[kye]

4 hours ago, John Matthews said:

But Kye, the human eye has evolved, you see. What was more than adequate in 2015 is no longer going to cut the mustard in 2022. Funny how things have changed.

lol...  I guess it's just me and my logic again, getting too big for my britches 😛 

[kye]

7 hours ago, Django said:

I'm talking acquisition from a DPs perspective.

As the following video from 2014 states, 80% of films back then where shot above 3K yet indeed 99% were finished in 2K.

That doesn't mean it was ideal as far as IQ as the guy explains nobody printed Super35 to Super16 which is sort of what was being done back then. Not to mention they were already anticipating 4K streaming (or "broadband" as they called it then lol). He goes on to explain/push the benefits of shooting in ARRIRAW at 3.4K with the then new XT and details a workflow where you'd transcode to ProRes 444 2.8K for a 50% gain in file size and almost 3K resolution. That would have definitely been the forward-thinking workflow:

 

Yes, acquisition tried to keep pace with the tech I'm sure, and this was also in the days when cameras didn't downscale in-camera which made oversampling at capture even more important.

My point was simply that if a 2K Prores HQ didn't have intolerable macro-blocking when used in a workflow that was projected in multiplex cinemas, that 4K Prores LT with a similar/superior bitrate wouldn't be worse than that.  Also, the people in this discussion are pretty unlikely to be shooting for projection on a 590" cinema screen, so a codec with 400Mbps being streamed at 15-25Mbps, LT is probably good enough....   even with our post-millennial biologically superior eyesight 😉 

[TomTheDP]

20 hours ago, Django said:

If you're referring to the Alexa Classic, 2K ProRes HQ was closer to 300Mbps and it could shoot up to 430Mbps in 4444.

Second gen Alexa XT really changed the game though with ARRIRAW /Prores XQ and resolutions up to 3.4K in OpenGate.

So I'd say "sufficient" by 2010 theatrical standards but ARRI IQ took a major bump as soon as the XT came out around 2013. 

Besides I heard there were software upgrades to give the Classic 3.2K/3.4K XQ/Open Gate so those bitrates and resolutions may have been used much earlier than 2013.

Long story short I don't know if that many features were actually shot in 2K ProRes HQ..

The difference between 3.2k the resolution being used if shooting spherical and 2.8k which is what the Alexa classic sensor uses is pretty small imo. Yes the Alexa Classic is 2k, but its downsampled from 2.8k. If you look at the difference between downsampled 2k from 2.8k and 2.8k native it is pretty much impossible to tell the difference. The benefit of shooting RAW is definitely nice, ARRI raw has more texture than prores. Of course it comes at the cost of a little more grain. I am sure the little extra info in Prores 444XQ is helpful sometimes too. To me ARRI didn't get a big boost until the LF tho. You do notice the extra detail comparing the LF to any of the previous S35 Alexas. Comparing different flavors of 3k is pretty negligible though at least in terms of resolution. 

I do think 12bit 444 makes a noticeable difference over 10 bit if you push the color at all, mainly with skin tones.


just my 2 cents 

[Django]

7 hours ago, kye said:

Yes, acquisition tried to keep pace with the tech I'm sure, and this was also in the days when cameras didn't downscale in-camera which made oversampling at capture even more important.

My point was simply that if a 2K Prores HQ didn't have intolerable macro-blocking when used in a workflow that was projected in multiplex cinemas, that 4K Prores LT with a similar/superior bitrate wouldn't be worse than that.  Also, the people in this discussion are pretty unlikely to be shooting for projection on a 590" cinema screen, so a codec with 400Mbps being streamed at 15-25Mbps, LT is probably good enough....   even with our post-millennial biologically superior eyesight 😉 

Cinema screen is actually not a great example as 85% of theatres in the world are still using 2K projectors (that number was probably even bigger a decade ago). Cable TV in the US aren't even all 1080i with some networks still broadcasting at 720p. 

That is the main reason why ARRI have been so slow to develop true S35 4K cameras and why DI/editing/finishing have been done on 2K timelines.

We often like to put cinema on a pedestal for various valid reasons but truth is that concerning resolution, that whole industry has been lagging behind home entertainment and streaming productions for a minute. The 2014 video I linked already assesses that back then. We simply can't ignore what's been happening on streaming platforms and social media like it or not that's most peoples daily reality and benchmark. 

So yeah depending on what industry you're in, what you shoot, what your end viewer is displaying on, your storage capacity and bandwidth, what your post tech skills & min/max requirements are: different workflows, different resolutions, different codecs may be used in the pipeline. That is the reason why there are so many options.

Heck for some, 8-bit rec709 may still be better than 10-bit log or even 12-bit RAW.

So within this diversity of context I don't think we can really blanket one codec or one resolution format as being ok across all scenarios.

That said, sure 4K ProRes LT is a pro standard codec, fairly chunky and superior to low bitrate h26x. Its no XQ 444 though.

ProRes in consumer hybrids is a great leap forward. Much more practical than RAW in editing sense.

On a similar tangent, I also think open-gate has never been so useful. Traditionally for anamorphic shoot but today also serves the purpose of multi-aspect ratio deliveries which is becoming pretty common.

57 minutes ago, TomTheDP said:

The difference between 3.2k the resolution being used if shooting spherical and 2.8k which is what the Alexa classic sensor uses is pretty small imo. Yes the Alexa Classic is 2k, but its downsampled from 2.8k. If you look at the difference between downsampled 2k from 2.8k and 2.8k native it is pretty much impossible to tell the difference. The benefit of shooting RAW is definitely nice, ARRI raw has more texture than prores. Of course it comes at the cost of a little more grain. I am sure the little extra info in Prores 444XQ is helpful sometimes too. To me ARRI didn't get a big boost until the LF tho. You do notice the extra detail comparing the LF to any of the previous S35 Alexas. Comparing different flavors of 3k is pretty negligible though at least in terms of resolution. 

I do think 12bit 444 makes a noticeable difference over 10 bit if you push the color at all, mainly with skin tones.


just my 2 cents 

Oh I'll believe you that the IQ difference is marginal in between 3.2K & 2.8K and 2.8K & 2K downsampled.

However for anamorphic shooting, difference in between the XT's Open Gate 3.4K in ARRIRAW  vs 4:3 2K ProRes of the Classic was substantial as the video details. Even though technically they're both the same ALEV sensor with identical characteristics.

The LF has a larger than FF sensor shooting up to 4448 x 3096 so yeah that's a major boost in IQ as far as detail but also noise levels which was the biggest con of the OG Alexa S35 sensors imo.

[kye]

1 hour ago, Django said:

Cinema screen is actually not a great example as 85% of theatres in the world are still using 2K projectors (that number was probably even bigger a decade ago). Cable TV in the US aren't even all 1080i with some networks still broadcasting at 720p. 

That is the main reason why ARRI have been so slow to develop true S35 4K cameras and why DI/editing/finishing have been done on 2K timelines.

We often like to put cinema on a pedestal for various valid reasons but truth is that concerning resolution, that whole industry has been lagging behind home entertainment and streaming productions for a minute. The 2014 video I linked already assesses that back then. We simply can't ignore what's been happening on streaming platforms and social media like it or not that's most peoples daily reality and benchmark. 

So yeah depending on what industry you're in, what you shoot, what your end viewer is displaying on, your storage capacity and bandwidth, what your post tech skills & min/max requirements are: different workflows, different resolutions, different codecs may be used in the pipeline. That is the reason why there are so many options.

Heck for some, 8-bit rec709 may still be better than 10-bit log or even 12-bit RAW.

So within this diversity of context I don't think we can really blanket one codec or one resolution format as being ok across all scenarios.

That said, sure 4K ProRes LT is a pro standard codec, fairly chunky and superior to low bitrate h26x. Its no XQ 444 though.

ProRes in consumer hybrids is a great leap forward. Much more practical than RAW in editing sense.

I was really just comparing 2K Prores HQ to 4K Prores LT.

If we're talking about anything more than that, then we have to start talking about what is visible, and that means discussing a certain resolution test that makes sensible discussion basically impossible.