Welcome to our community

Be a part of something great, join today!

Register Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

skintone

James_Mills said:
I think it's pretty funny how so many of you act as if there's a blatantly noticeable difference between the color of one's skin throughout different HD format cameras. To this day I usually have trouble discerning whether or not a film is even digital or film when I'm at the theatre, let alone what kind of digital format and what "skintone" I'm viewing.

But I guess a lot of y'all are better versed for this kind of stuff than, say, a Soderbergh or Fincher.
Click to expand...

the difference may not be blatant but it is definitely there, and yes, quite discernible to the trained eye. I suspect Fincher and Soderbergh may notice it too but choose to ignore it in favor of all the other things digital capture brings to the table. Its what we all do.
 
Nigel Stanford said:
FYI, FilmConvert grain scans are 6k, for $199.
Click to expand...

Dear Nigel,

When the new 6K sensor hits the street, will Film Convert be able to apply the 6K scanned grain to the Dragon footage as well?
Will we be able to work in 6K in FilmConvert?

Thanks!
 
James_Mills said:
I think it's pretty funny how so many of you act as if there's a blatantly noticeable difference between the color of one's skin throughout different HD format cameras. To this day I usually have trouble discerning whether or not a film is even digital or film when I'm at the theatre, let alone what kind of digital format and what "skintone" I'm viewing.

But I guess a lot of y'all are better versed for this kind of stuff than, say, a Soderbergh or Fincher.
Click to expand...

I personally saw a huge difference when Mad Men moved to digital. It looks completely different. I just said to myself, well that's a bummer, and moved on. But a quick google confirms they moved to Alexa. This is not a bash of Alexa... but it's vastly different to the earlier seasons.
 
Medavoym said:
Dear Nigel,

When the new 6K sensor hits the street, will Film Convert be able to apply the 6K scanned grain to the Dragon footage as well?
Will we be able to work in 6K in FilmConvert?

Thanks!
Click to expand...

The FilmConvert standalone is limited to 4k output right now, but we can easily increase this to 6k once the dragon is out. If people want to export to 6k. I think most will capture 6k and stick to 4k export.

As for the plugins, it will depend on their architecture, but I don't see why not.
 
Nigel Stanford said:
The FilmConvert standalone is limited to 4k output right now, but we can easily increase this to 6k once the dragon is out. If people want to export to 6k. I think most will capture 6k and stick to 4k export.

As for the plugins, it will depend on their architecture, but I don't see why not.
Click to expand...

I agree.

But people will probably color correct natively at 6K (directly from the R3Ds) and only then export to 4K, as a final mastering step. So I guess it makes more sense to apply the grain in 6K, and then export everything to 4K?
 
Graeme Nattress said:
Well, film ISO works by changing chemistry. We don't do that. ISO setting for us is just electronic gain in development, so it doesn't impact the RAW R3D file at all. When there's a physical change, like chemistry, or analogue gain (or in some cases burned in digital gain) there is a sense of a certain setting being native, or that particular stock is native to a particular ISO value. So I think your argument above more applies to us not having a native "do least harm" setting, as compared to a DSLR say where some settings of ISO are done with pure analogue gain, (and hence the one with least analogue gain could most likely be called "native") and some with burned in digital gain which are often compared to the ones which don't burn in digital gain by saying they're "not native".

A confusing topic for sure.....

Graeme
Click to expand...

Hi Graeme,

Can I ask if you ever think there will be an option for burning in the digital gain on Epic? It's really my only complaint now. Being RAW means that shooting at high ISO's produce more compression artifacts because the signal is not boosted prior to compression. If the R3D had options for baking the gain before it was compressed, then for me that would eliminate my urge/need to go and buy an external RGB recorder.

I did a side by side test the other week with my friend's SONY NEX-FS700 and his camera performed better in low light than the Epic did. However the live output on the EPIC was much better than the FS700. So really, the sensor on the EPIC was more than capable, but when played back the R3D was unusable at such high ISO's. To compete, I would need an external recorder. Is this such a hard thing to implement on the Epic or does it simply go against the RAW philosophy?
 
Karl H said:
However the live output on the EPIC was much better than the FS700. So really, the sensor on the EPIC was more than capable, but when played back the R3D was unusable at such high ISO's. To compete, I would need an external recorder. Is this such a hard thing to implement on the Epic or does it simply go against the RAW philosophy?
Click to expand...

In this case you are subsampling a 5k sensor output, greatly reducing noise...5k red footage scaled to 1080p is clean, 1:1 is not @ relatively high isos.
 
Tommaso Alvisi said:
In this case you are subsampling a 5k sensor output, greatly reducing noise...5k red footage scaled to 1080p is clean, 1:1 is not @ relatively high isos.
Click to expand...

Sorry Tommaso you're missing the point, or I didn't explain myself. Downscaled to 1080p the EPIC R3D is still noticeably noisy compared to the FS700. The reason is simple, an R3D does not have baked in gain before sent for compression so it introduces more artifacts the more you 'boost' it in post.

Imagine shooting on your canon and you need to up it to 3200ISO to get a nice image. Now, you would never record at 100ISO and then boost the image post would you? But this (in a very simplistic way) is how Epic works, ISO is metadata and so the signal sent to the compression system in low light falls in the worst part for any compression algorithm. If it was boosted first you'd have noticeably cleaner footage in low light.

If you don't believe me do a test. Shoot high ISO on EPIC, say 4000, and record an R3D , meanwhile take the live output from the EPIC at 4000ISO and record to any external recorder.
Downscale the R3D from 5K to 1080p and compare. the R3D will look worse, (I hate to say it, but much worse the higher the ISO)

Yes we need to light, but the frustrating part is that the sensor Is capable of better low light ability and it's the RAW recording that limits that right now.
 
I agree Karl, that Redcode does have a hard time with underexposed footage compared to what you get through the SDI which bypasses the REDCODE compression. Oh the dilemma, clean output but 1080, or use 5k and neat video?
 
Thanks Bob, but couldn't Epic simply have an option to bake in ISO prior to recording (for those who want it) or would that takes us out of RAW space?
 
Karl H said:
Thanks Bob, but couldn't Epic simply have an option to bake in ISO prior to recording (for those who want it) or would that takes us out of RAW space?
Click to expand...

It's not about raw or not it's about compressing noise. Downsampled noise is = to less noise. high noise levels does not do well when compressed. r3ds are big files but compressed. So basically if you would record the sensor without compression into a r3d and then downscale vin post it it would be just as good as the SDI out but it would be raw. The issue is that compression is applied before downsampling when doing the onboard recording.

Still I rather have it as it is. The solution to lowlight is not to record the sdi... the solution is dragon. 2000 iso native then when you are asking for more... I guess a flashlight would do :)
 
Hi Bjorn,

Understand that, you're saying the R3D is worse because it's trying to compress larger noise at full 5K, than the external recorder at 2K.

Their may be some truth to that but it's impossible to quantify with the EPIC as there's no way to record a 5K signal out of the SDI and compare. But another major difference between the two files is one is recorded post gain (SDI external) and one pre gain (R3D). I would guess even if recorded at full 5K that having an option to record post gain and baking in ISO would produce less compression artifacts.

What I'm noticing is that the MX sensor does rival some of the best low light cameras out there, but R3D is not the way to collect it ATM
 
Thanks Bob,

That's really helpful. :

"Canon Raw is a 10-bit format with baked-in ISO and white balance. That’s right, unlike the other Raw formats, Canon is baking-in gain adjustments (ISO/white balance) before outputting the Raw data. You may be scratching your head as to why, so here’s a little bit of the logic. Adding gain adjustments at the sensor level produces a consistent stop range above and below the middle grey level, even at high ISOs, and reduces the overall noise in the image."

Love to hear Graeme's thoughts on giving us some kind of option like this, it's the noise level I'd be interested to see improve.

I'll shut up now.
 
I would welcome this option as well. Why not help get us the best results possible. Those who wan't the option can enable it in the menu.
 
Karl H said:
Hi Bjorn,

Understand that, you're saying the R3D is worse because it's trying to compress larger noise at full 5K, than the external recorder at 2K.

Their may be some truth to that but it's impossible to quantify with the EPIC as there's no way to record a 5K signal out of the SDI and compare. But another major difference between the two files is one is recorded post gain (SDI external) and one pre gain (R3D). I would guess even if recorded at full 5K that having an option to record post gain and baking in ISO would produce less compression artifacts.

What I'm noticing is that the MX sensor does rival some of the best low light cameras out there, but R3D is not the way to collect it ATM
Click to expand...

What I was saying is, that if the raw signal was not compressed , then it's essentially the straight data that is coming from the sensor. Where in the chain you "gain" that data is not really relevant. but downsampling it and not compressing it makes quite a difference. Especially as the way the compression work it is really sensitive to noise. The more noise the harder the picture is to compress. So most of the noise you see in a dark r3d file is not sensor noise, it's compression noise that comes like a multiple of the little sensor noise thats there. Hence the reason for shooting lower compression rates in lowlight. Using 3:1 instead of 6:1 or so, makes huge difference in dark shots.
 
Björn Benckert said:
What I was saying is, that if the raw signal was not compressed , then it's essentially the straight data that is coming from the sensor. Where in the chain you "gain" that data is not really relevant. but downsampling it and not compressing it makes quite a difference. Especially as the way the compression work it is really sensitive to noise. The more noise the harder the picture is to compress. So most of the noise you see in a dark r3d file is not sensor noise, it's compression noise that comes like a multiple of the little sensor noise thats there. Hence the reason for shooting lower compression rates in lowlight. Using 3:1 instead of 6:1 or so, makes huge difference in dark shots.
Click to expand...

Hey Bjorn, I agree with most of what you said, especially that compression has a hard time with sensor noise, but I think it makes a very big difference where in the chain you apply gain.

Bandwidth priority is given to information in the highlights and midtones, even in Redcode. So, if you've just got a bunch of information sat at the left of the histogram it's still going to be in the danger zone and plastered in artifacts. That's partially because of sensor noise yes, but also there's a point of no return on the histogram that anything below will be artifact riddled because it's less relevant to the compression engine. As you say this is evident because if you shoot at 3:1 instead of 7:1 you see less artifacts. I think Graeme once said that the further you moved from the the top of the histogram, the less bandwidth it was given. So if you've got a super low light shot, then you have a signal thats getting no love from Redcode at all, even at 3:1. To the compression engine it's all still in the 'irrelevant' zone. Or at least 'far less relevant' zone.

If however, you could take that signal, apply gain till you had a full histogram (even if that was 6400 ISO), the recorded image will look much better because those shadows are now pushed into the midtones and highlights (at least technically) and given greater bandwidth. Using an EX1 you have options of digital gain when shooting. That's there because if you tried to shoot at 0db in low light and add gain in post, it would look mush. Now, of course there is a world of difference between Redcode and an mp4 RGB codec, but the same rules apply to a lesser degree. You need to send the 'fullest' histogram you can to the engine and in very low light that can only be done by applying gain first.

I agree that it would still have a hard time compressing noise, but I think it would still be vastly superior. At the very least it would mean you could shoot at a higher compression ratio in low light and get a cleaner image. Unfortunately we can never do a practical test, because we don't have the option.
 
Patrick Tresch said:
Sometimes after a 7 hour color correction session I walk in the street and feel people have strange skintones ;-)

Reality isn't always what we imagine.

Now I hope no professional camera, nor post-process, will make thinks look good just to get weird DPs (customers) happy.

Pat
Click to expand...
Ha!! Me too.....
 
You must log in or register to reply here.
Back
Top