Welcome to our community

Be a part of something great, join today!

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 :)

Alchemy...

Status
Not open for further replies.
We're aware of the issue in RCXP3. I think that's now fixed in 4.

Davinci are "doing their own thing" with respect to ACES support - it's neither coded nor supported by us, so best ask them why it doesn't work right. The EXRs we export work correctly with respect to adhering to the colour space specification and any issues you see are indicative of the rendering issues inherent in developing an ACES image.

Graeme
 
To go along with this wonderful ability, is there any recommended tutorial material available on grading in general? I'd like a good extensive course.
 
Steve Wake said:
WOW!

I have a feeling this will have two major effects:

1. Any lens can get closer to that "Leica look" as far as microcontrast.
2. Scarlet 3K footage uprezzed to 4k will be comparable to or better than any competitor's 4K RGB codec, even 10 bit 4:4:4:4

Why?
  • Last I remember word of Graeme working on slider controls had to do with HDRx.
  • Having shot a lot of HDR, one of the more tasteful things one can do with it is adjust local contrast in small amounts, which gives an image that pops with a perceived increase in microcontrast. Not an actual increase in resolution, but looks like it. Different from sharpening.
  • Since Jim said this can be used on any RAW footage I'm guessing Graeme figured out how to use a microcontrast control to enhance a single stream.
  • Part of the unmeasurable Leica look, or look of other "special" lenses is their microcontrast-which has to do with coatings, QC, overall design, etc.
  • It's an effect that is simulated when you downrez a photo.
  • Now if you're going to uprez (say 3K to 4K) it would be nice to do it while still preserving as much perceived resolution as possible. A microcontrast control that worked as early in the processing stream as possible would help achieve the goal. (Or maybe RED will have a special Scarlet uprez in RAW option to apply before this step).
  • If I had a tool like that and was about to release a 3K camera, I would release the tool at the same time (or shortly before), then at the unveiling of the camera, show uprezzed 4K footage on my 4K projector and blow the competition away.

The only downside: you'll need 4K viewing ability to really judge the effect. A $$ downside.

The big upside: It's f-ing FREE. And you know it will be upgraded continuously. No wonder STORM bit the dust.
Click to expand...


Yup….

“Theoretically” there is a lot that can be done for correcting systematic deleterious effects caused by physical optics. It seems the Bayer pattern itself may offer a few tricks up its sleeve to correct some of these issues (as compared to a theoretical RGB device). I guess this does underscore that this is super clever AND as Graeme makes it out it’s just the beginning. That’s a big deal set to become increasingly so when the physical MTF vs. CA based issues come into play where everything is at the normal limits of expected “maximum” capability. I guess Alchemy lets you re-group and tune in this and that for a spatially and color corrected image where things start to normally fringe/break up a bit.

So in that sense it does turn lead into gold as Jim says (he, he).

For VR and s-3d type installations for public outreach type applications have been wondering about how 3k (let’s say Scarlety) footage looks on a 4k projector and if actual up/rezzing is necessary or whether there is some on board processor on the 4k laser projector that can help with that?

The other thing is in S-3d if you have “twin” 3k footage that in theory AND in practice the effective perceived horizontal resolution is actually higher. In essence for the most part you have two samples for every object point that are presented. Really hoping to see the s-3d implementation of the new 4k laser projector. Does anyone know if that beasty is being shown at the Nov 3rd gig?


Interesting thinking…

Cheers,


Eric
 
Jim, shouldn't it be Alchemy with a TM or an (R) after it?

Great name.
 
If this does indeed make uprezzed 3k footage look (nearly) as good as 4k, then I think that its all the more likely that scarlet will be 3k.

Still, if this was the plan from the beginning, they would have to have been working on this for a long time, right?
 
RED brings vertical integration to the digital age. Who knows the sensor and AA filter better than RED? And with Graeme, who better than RED to maximize the possibilities of processing. Digital imaging is truly still in its infancy. A stage of development when the real scientists and visionaries can shine. Not the best time for engineers held in thrall by beancounters - and the customers of those companies. RED didn't have to exist as the relatively well-oiled machine that it is (it could be a poorly financed garage operation) - but we should be glad it has a visionary billionaire behind it.

As for some old-fashioned vertical integration:
Jim, how about investing in lithium production in Bolivia? If Soderbergh doesn't already have a connection to Evo I'm sure that could be developed.
And rare earths - maybe in a safe country like Sweden. Or maybe US or Canada?
If this Scarlet thing really takes off you might need these old-fashioned raw materials.
 
Steve Wake said:
RED brings vertical integration to the digital age. Who knows the sensor and AA filter better than RED? And with Graeme, who better than RED to maximize the possibilities of processing. Digital imaging is truly still in its infancy. A stage of development when the real scientists and visionaries can shine. Not the best time for engineers held in thrall by beancounters - and the customers of those companies. RED didn't have to exist as the relatively well-oiled machine that it is (it could be a poorly financed garage operation) - but we should be glad it has a visionary billionaire behind it.

As for some old-fashioned vertical integration:
Jim, how about investing in lithium production in Bolivia? If Soderbergh doesn't already have a connection to Evo I'm sure that could be developed.
And rare earths - maybe in a safe country like Sweden. Or maybe US or Canada?
If this Scarlet thing really takes off you might need these old-fashioned raw materials.
Click to expand...


Thinking of VERTICAL integration...


I’m also musing about “skew” correction (after the fact), dial up or dial down ; maybe I’m not thinking about this very clearly or being just too “fuzzy” this morning. So conceivably if you crop in a smidge as the scan rate and pattern is known then you could spatially re-compose the image (I know that they have been doing something similar for some time with Aerial cameras). Maybe I need a slap to the head, …my espresso machine is broken…(need to get that fixed other wise we are in big “trub”.).

It’s interesting that the SONY effort claims a pretty global shutter BUT they claim a better one with a mechanical one interposed?

Maybe there is also a spatial and temporal component also to Alchemy where after the fact you can dial up or dial down the globally”ness” of the image. Maybe you can do that but what does that look like? Any good for professional work/look etc?

So it’s sort of spatial, tonal, chromatic and temporal correction…?

That would be "Wizzard..."?

But does it make the tea? (that’s the important question)…


Ta.

Eric


[I think a couple have dropped hints about "timing" (he he)]
 
Gabriele Turchi said:
In Davinci , Those EXR looks fine (no issue ) as soon i apply a Rec709 output transformation Lut (part of the aces engine system) the image gets banding in the Sky , and yes overall blocky noise

again stretching those EXR in ACES mode without that output transform Lut does not produce any banding , that Lut applied seems that translate badly the colors
Click to expand...

I don't know how Davinci implements ACES, so I can't comment in detail. But I would say that 3D LUTs do not lend themselves well to transforms with linear light input, and can easily produce a noisy output.

I tested in Nuke, and implementing an ACES sRGB output transform using a combination of matrices and a 1D LUT produces a good result with EXRs from RCX. Implementing that as a 3D LUT produces a very noisy inaccurate result with a 17x17x17 10 bit cube, and even with a 64x64x64 16 bit cube the result is better but not ideal, with very visible quantisation noise.
 
Graeme Nattress said:
We're aware of the issue in RCXP3. I think that's now fixed in 4.

Davinci are "doing their own thing" with respect to ACES support - it's neither coded nor supported by us, so best ask them why it doesn't work right. The EXRs we export work correctly with respect to adhering to the colour space specification and any issues you see are indicative of the rendering issues inherent in developing an ACES image.

Graeme
Click to expand...

Hi Graeme,

can you elaborate a bit on those EXRs?

Are these 16 bit half float?
Are these 16 bit int?
How are the colors mapped into that bit range and how can we make sure not to see any clipping taking place?

Is that available from the SDK as well?

As explained before we would love to get half float data out of the RED SDK that puts all colors according to the ISO settings between 0.0 and 1.0, but - remember its half float format - keeps the otherwise clipped colors in the value ranges above 1.0 and below 0.0 accordingly.
Will that be possible at some point?
Because THIS is exactly what is need in any compositing pipeline like Eyeon Fusion, The Foundry Nuke etc. when it comes to deep color processing.

Cheers,
Axel
 
16bit half float, with REDColor2 yes you'll find it hard to clip colours. No matter what the ISO it won't clip in the data, and it's totally compatible with REDLogFilm, so if you compare them both interpreted correctly they'll produce the same image. EXR support has been doing this for a long, long while in RC-X.

Graeme
 
Graeme Nattress said:
16bit half float, with REDColor2 yes you'll find it hard to clip colours. No matter what the ISO it won't clip in the data, and it's totally compatible with REDLogFilm, so if you compare them both interpreted correctly they'll produce the same image. EXR support has been doing this for a long, long while in RC-X.

Graeme
Click to expand...

hi graeme

what do you mean is totally compatible with redlogfilm ?
isn't the EXR exported as linear light ?

thanks
g
 
Meaning if you linearize RLF you get the same image as the EXR, or if you apply lin2log to the EXR you get the same image as RLF. Always EXRs are linear light.

Graeme
 
Is anyone else totally clueless as to what Alchemy is? This thread is much ado about nothing...
 
Status
Not open for further replies.
Back
Top