Testshoot with Alexa and Epic Dragon in germany

[Hans von Sonntag]

So personally I think that some, like my self will continue to work straight from trimmed r3d's. Flame/lustre allow for this workflow and I find it to be a very good setup not for everyone for sure but, I think that a solution where the comps are floating in terms of grading until final render is a real plus and for certain applications and VFX work especially I find it to be the way of the future...

Bjoern, I guess you are hard importing the R3Ds, meaning they get transcoded to RGB DPX by Flame's Backburner using Red's SDK. Working off soft-imported R3Ds that requires a permanent debayering on the fly is tedious unless you hard-commit your stuff some wehre in your workflow. Is this a true RAW workflow...?

Collaborative working off R3Ds requires that everyone uses exact the same metadata and build of the SDK. This is possible to accomplish in a small team but close to impossible with larger teams. Hence the need to transcode all shots to something meaningful in RGB from 1 source feeding all the seats that work on the project.

----

On-Set LUTs as a pre-garde for the final finishing work to show the director's intention. But I always change my mind later in post. Hence personally I find on-set LUTs not as valuable as they might look at a first glance.

----

According to Red by transcoding to RedLogFilm you maintain the most possible information from 16 Bit linear RAW to 10 Bit RGB. It doesn't matter where in the pipeline you transcode to RGB as long as you don't screw up the SDK's colour interpretation by selection funky settings such a curves, gamma etc...and/or introduce further compression artefacts (R3Ds is already compressed) by using a mediocre codec. In the RAW vs. RGB workflow discussion the point is often missed that RAW is ALWAYS RGB if we want to look at the pictures we shot. So wether the GPU/CPU does the transcoding on the fly or we work off already transcoded RGB does not change anything. So yes, IMHO Jake is right when he says that there exists a RAW-myth on these boards.

Hans

 

[Gunleik Groven]

Hans... Even with Epic, there were advantages of going 16 over 10 bits. These will be amplified by the Dragon.

Linear EXR or log DPX is pretty sweet for larger-shop workflows...

 

[Hilmar]

Hey Guys,
there is no reason to get harsh. Even since I know jakes posts are often a little provocative I do agree with him on the most matters of this pipeline subjects.

It is important to do Pre-Grades before sending footage to VFX houses. They should be able to work close on the colors the film will look like in the end. Otherwise there would be a chance that footage and integrated FX differ from each other during heavy color corrections. There are just 2 things where I'm asking for a change:
A: I would like to get more bit depth than 10-bit of the very common DPX-Workflows. The actual cameras deliver much more. EXRs are very fine with me. In the last project we were working on some underwater scenes (millions of blue tones). I asked the post house if I could deliver them 32bit EXRs instead of the DPX files because I wanted them to have more flexibilities in grading before color bending starts. It was not possible since they convert everything to DPX-Sequences for their Lustre...
B: I heavily recommend to additionally deliver the Raw-Files to the VFX houses. And I don't mean them to grade with the files. The Pre-Grades are done to go for the Look of the film. Not to shift the colors in areas were we vfx guys could work better. The best example is greenscreens. After Pre Grades the greens very often are not greens any more. Another example could be the tracking department. For tracking you could shift the contrast of your shot to the best spot to see your tracking points the best...
You don't even need the LUT for that stuff, since you don't effect the colors of your shot. When you finished working on the greenscreen you can just render out a mask and import it into your Nuke script with the EXR imports...
But for sure it would be good to get the LUTs as well. And if you deliver the LUTs and the Raw files to the vfx house they could render the EXRs by themselfes if you want to save transfer space.

We did a similar thing on a project with Alexa Log (not Raw). With some easy retouching shots we just opened the clean Alexa Log files in Nuke added the lut, did the job, exported back in Alexa Log (without the LUT) and done. This way we could work in the highest possible image qualitiy there was and the grader was not bound to work with the pre grades since they were not baked in. I'm pretty sure that not many people used this kind of technic since I found a bug in Nuke's good old version 6 that shifted the colors a little when opening Alexa Log files in the correct colorspace. And I don't belive that noone ever recongnised that before if someone would have worked that way. (It effected the skin tones and the green noise in lower floors got lighter... But since version 7 its fixed. )

The Filmlight approach you mentioned, Jake, sounds very interesting! I would like to hear more of this kind of solutions. I also asked the graders in some projects (espacially with CG characters integrated in live footage) to go very far with their pre grades (to about 90%). But it was a little bit risky since the colors and contrast were "baked" in the DPX-Sequences and when the DOP would have asked for a bigger change in the end we would have had a problem... But of course he was involved in that decision.
Would be nice if with such plugins it would be possible to get more flexibility. I just remember the open Adobe DNG Raw format... Was it ever meant for writing Raw files in workflows we are just talking about?

Have fun with what you are doing!
cheers,
Hilmar


PS: I know, that you can get really nice pictures out of the Alexa shooting with Raw. I was just a little bit down cause so many projects were shoot on Quicktimes and that we had such bad quality with the shots in 120fps in Quicktime mode.

---
Jan-Hilmar Petersen
VFX Supervisor

Farm Studios Germany

 

[Gunleik Groven]

Hilmar.

Agree on all points.

 

[Hans von Sonntag]

Hans... Even with Epic, there were advantages of going 16 over 10 bits. These will be amplified by the Dragon.

Linear EXR or log DPX is pretty sweet for larger-shop workflows...

Gunleik, 16Bit 1080p DPX is for most postproduction pipelines hardly manageable. The size increase does for most jobs not justify the gain in perceived quality.

Of course you are right, the boundaries will be pushed further and further. But the question remains: how limiting are those boundaries? I never surfed on the bleeding edge of technology, chances are too high to get cut. In my world of client-payed jobs security and dependability are mandatory. Hence no technical experiments that don't broaden the creative range.

Cheers,

Hans

 

[Hans von Sonntag]

A: I would like to get more bit depth than 10-bit of the very common DPX-Workflows. The actual cameras deliver much more.

---
Jan-Hilmar Petersen
VFX Supervisor

Farm Studios Germany

Jan, Epic has a 16 Bit A/D transducer and delivers 16bit linear RAW. With 10 Bit Log RGB you get the sensor's data pretty well stored in a 10bit RGB DPX container. I know that there is a huge difference between 16 Bit and 10 Bit regarding graduation from black to white. But perceived there is for the human eyeball no difference to see, especially if the audience are not only kids but also viewers who passed the age of 40.

For the eye and production value it is more important that the camera delivers 1 stop, or better 2 stops more DR than the question wether post production is 10 Bit or 16 Bit based.

Hans

 

[Hilmar]

Hans, I see what you mean. And with the dynamic range I totally agree.
The thing why I like to work in a better quality than the eye can see is on one side vfx software can produce better results in its calulations (the most of these programs work in 32-bit linear anyways) and I like to give back to the grading department the most of information I can, cause after we are done there is still alot happening to the pictures. I just think that file sizes shouldn't be the biggest problem any more. And I would like to find more solutions to get better workflows...
eg. What you said about collaborative working off R3Ds. It shouldn't be a big deal to program a pipeline which tracks the Raw xml-files along with LUT-files associated to a shot. I like the idea.

 

[Gunleik Groven]

A good pipeline that did not involve transcoding when you have 2-3 VFX parties involved and 2.-3 graders and possibly a few editors working in parallell... would be nice.

I still fail to see the advantage of editing from the R3Ds for longform, though. (after a bit of thinking)
But for post, and especially as the GPU acceleration trickles down the SDK, I can see that happening more.

G

 

[Hans von Sonntag]

Hans, I see what you mean. And with the dynamic range I totally agree.
The thing why I like to work in a better quality than the eye can see is on one side vfx software can produce better results in its calulations (the most of these programs work in 32-bit linear anyways) and I like to give back to the grading department the most of information I can, cause after we are done there is still alot happening to the pictures. I just think that file sizes shouldn't be the biggest problem any more. And I would like to find more solutions to get better workflows...
eg. What you said about collaborative working off R3Ds. It shouldn't be a big deal to program a pipeline which tracks the Raw xml-files along with LUT-files associated to a shot. I like the idea.

Jan, 16 or 32 Bit float as a precision scale in VFX applications make a lot of sense. But to derive from there that the camera data should have the same precision is questionable. Let's assume that you get the data off an Epic into your VFX machine in 16 Bit linear, meaning max. possible precision. What do you have? You have 2 stops of very accurate noise in the blacks, 3 stops of max. precise usable information that eventually will get crushed later in the grading process, and last but not least 3 stops of very precise information in the whites which precision is hardly usable because no one can see it. Ok. Perhaps slightly exaggerated but you got what I mean.

In my opinion it makes a lot of sense to have as much DR as possible for post production. But Log encoding in fact makes it possible to include much more information/stops in a container which has less bits than stops for convenient handling. 16 Bit DPX or EXR are not viable for many because of the sheer data rate and huge space which is required to handle the files, think also about backup etc...

But we will have that in 10 years or so. Then cameras will have 24 stops of DR, setting the iris is not a question of exposure anymore and the information the 24bit AD transducer delivers are by then 24 Bit linear RAW. Then of course 16 Bit RBG Log will make a lot of sense to picture the 24 bits of RAW data. I also can imagine that developers in 10 years will store the assumed 24 stops in a 16Bit Log RAW for convenient camera storage.

Hans (still happy with log encoded Epic footage, AD boxes and 10 Bit DPX)

 

[Jeff McCoy]

Well, yes colorist hate me for grading in flame, but for a lot of short format productions (read commercials) I find it quite convenient DPX files and EXR does not come with built in proxies, they do not come compressed and even if they are capable to hold everything that was captured they also add weight, as in the case with RED the original source is compressed raw 16 bit. So personally I think that some, like my self will continue to work straight from trimmed r3d's. Flame/lustre allow for this workflow and I find it to be a very good setup not for everyone for sure but, I think that a solution where the comps are floating in terms of grading until final render is a real plus and for certain applications and VFX work especially I find it to be the way of the future... Especially if you think of 6k source then even more so. I love to fiddle around with r3d's and and be able to suck them straight over san network and do my sketch composting at a quarter res or so, then it's like melted butter and when I like what I got in my comp I can simply just set the inputs to full debayer, adjust my settings and render the final, especially when doing 4k work then the r3d's shine in comparison to EXR or any other uncompressed formats. As most things we do are shot at 5/1 or even higher compression then bumping the files to uncompressed 16 bit makes things...well, a lot slower. So even if you do not like it I think others will... But fine with me, I like the idea of my workflow being a myth.

Here is a screen shot of my myth of the day... 29+ layers of r3d's that gets multiplied and comped together into a crowd dup shot.

Dude, this is amazing. I am one that really appreciates those that are doing the work the way it SHOULD be done and not the way it HAS been done. "good enough" and "what works now" aren't good enough in my book. This is similar to the talk about why D-tap connectors on cameras aren't great, and why 19mm is better than 15mm. It's tough to be forced to deal with things they way they have always been done in the past, when there is clearly a better way. Anyway, I could go on forever about this stuff, but wanted to express appreciation for your methods and not being afraid to work with new/future/better methods. The majority will argue with you about making any change, but not everyone loved Steve Jobs or Jim Jannard, but they have forever changed the game- for the better.
Cheers!

 

[Robert Hart]

Personally myself, although it would be anti-competitive to a degree, I would like to see ARRI and RED foreswear their fisticuffs and employ the best of both worlds even if it came down to some sort of cross-licencing thing. - Bring on the "ARAGON". Wouldn't happen of course but writers of fiction are self-endorsed in creative licence. It is personal entertainment after all even if in this instance it might be delusional indulgence.

 

[Shadi Abu Sharar]

Thank you for the share and am sharing it with my clients coz am having the same Problem they all ask for Alexa and i hate it when they ask for it and the Reason is that they don't want to spend time on converting altho am offering a free DIT Work Station with RED Epic to Transcode there files

 

[shashbugu]

Lets make it very clear, it's not end user clients that ask for Alexa cams, it's freelance editors, colorists, producers and DPs. The reasons vary from person to person. Thy cant really in point the benefits in image quality of an Alexa over Red. Frankly there are none. As for vfx, the type of vfx will determine the pre grade or post grade. It often always smart to pre grade and have a predetermined look where all vfx should match.

 

[jake blackstone]

Lets make it very clear, it's not end user clients that ask for Alexa cams, it's freelance editors, colorists, producers and DPs.

I have never heard of editors or colorists dictating the choice of the camera. I hope it doesn't come to that…

 

[shashbugu]

I have never heard of editors or colorists dictating the choice of the camera. I hope it doesn't come to that…

Not dictating more like recommending. Easier workflow. Mostly freelancers. I rent out equipment on a very small scale, it's Definetley not the end user clients that seek Alexa's. Honestly the Red is more popular around the world to end user clients but the professionals just like the easier way out.

 

[shashbugu]

I have never heard of editors or colorists dictating the choice of the camera. I hope it doesn't come to that…

How come you have over a thousand more posts than I have? I know I post twice as much as you do. You've been spending your extra post fighting with Rob Ruffo. Lol. I kid, I kid. Rob as usual just yanking your chain...

 

[Bérenger Brillante]

I had editors and colorists ask to shoot epic in 2k. Several times. And some other asking small files and demanding 12:1. They didn't understand their own workflow and even less the camera.

The reason, To do like with the Alexa. So yes, it happens all the time. Sadly.

Jake, The problem is you think every colorists and editors are as talented as you. Most are not.

The two workflow you are speaking of, bjorn and jake, are valid. I use both of them depending on the project.
R3d for projects without vfx, Dpx for vfx projects.

 

[Peter Moretti]

Hans... Even with Epic, there were advantages of going 16 over 10 bits. These will be amplified by the Dragon.

Linear EXR or log DPX is pretty sweet for larger-shop workflows...

Yes, in a "perfect world" 8, 10 and maybe even 12 bit codecs would be log and not linear. It's pretty crazy how those eight bits are being allocated, IMHO.

 

[jake blackstone]

I had editors and colorists ask to shoot epic in 2k. Several times. And some other asking small files and demanding 12:1. They didn't understand their own workflow and even less the camera.

That's because those DP's, that shoot 2k R3Ds thinking, that it is the same, as shooting 2k Prores or 2.8K RAW on Alexa, should be ashamed for not doing their own homework. And frankly, they should know better, than listen to an editor or a colorist on the subject of camera choice, resolution and compression. And yes, I have scolded few DPs, that did shot 2K and even HD on Epic. Personally, i'm camera agnostic. Said that, I have my personal camera preferences, but that has nothing to do with my work as a colorist.

 

[jake blackstone]

How come you have over a thousand more posts than I have? I know I post twice as much as you do. You've been spending your extra post fighting with Rob Ruffo. Lol. I kid, I kid. Rob as usual just yanking your chain...

Yep, even occasionally answering Rob's annoying posts helps my web site SEO to keep my web site at the top