Welcome to our community

Be a part of something great, join today!

  • 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 :)

Epic, HDRx, and TV post-production

David Mullen ASC

Moderator
Joined
Apr 20, 2007
Messages
7,832
Reaction score
7
Points
38
Website
www.davidmullenasc.com
Perhaps Mike Most, Michael Cioni, Mark Pederson, etc. can weigh in on this... but assuming enough Epics hit the rental houses soon for television production to get a shot at using them, what do you think the workflow options will be for material shot with HDRx turned on? For "Big Sur" we basically didn't deal with the HDRx information for dailies, it will be used in the final color-correction, but a number of TV shows convert Red footage and dump it on 10-bit HDCAM-SR tape at some point, and dailies tends to be done quickly, so I'm wondering how different post workflows for TV will handle HDRx.

I assume that those facilities like Light Iron that can color-correction from the original R3D files for the final mastering session will just deal with HDRx then, and dailies probably won't reflect the HDRx information. But what about those facilities that do a HDCAM-SR workflow for dailies and then color-correct from those tapes?
 
At this point in time - and I repeat, at this point in time - to do HDRx "correctly" basically means treating it as a VFX shot. In other words, there are currently different approaches to dealing with motion issues, as well as mixing the two images, so the material itself to some degree dictates how it's handled. A static shot might require little beyond doing a mix of the two, but a shot with a lot of motion, all at different speeds, might require a bit of preprocessing or finessing in order to get an optimal result. Over time, we'll see better tools for handling this, but at the moment, it's pretty much in its infancy, so those tools are still under development. That's not to say that HDRx shouldn't be used when dynamic range extension is needed or necessary, of course it should, that's what it's there for. But there is no "automatic" way to get the most out of it at this point in time, regardless of whether you're using the R3d file or transcoding. And at this point, transcoding does give you more options as to how to handle motion blur in the "X" track. Of course, it should also be noted that television is a bit less demanding in terms of "perfect" motion rendition than a 40 foot screen is, so that should also be taken into consideration.

I'd also be curious as to what Michael and Mark think.
 
At this point in time - ......
That's not to say that HDRx shouldn't be used when dynamic range extension is needed or necessary, of course it should, that's what it's there for. ...

I think this is the point we should communicate about. HDRx is a tool that can increase the dynamic range of certain shot but thinking the EPIC has natively 18stop of dynamic range is, as of today, getting in trouble. That said, the A track is so amazing in exposure latitude that we stand on rock solid ground to get a great picture out of it. My impression, (not based on charts) is that it's even better that the ONE.

HDRx is another tool (a VFX one) for the dop to learn to work with.

Today I still see some "DOP" taking the RED ONE M @ 250 or less asa for exterior shots getting all highlights burned out. When Epic will hit the street, people will be shooting all HDRx footage even with the increased data flow. We will see a lot of bad exposed footage due to wrong thinking.

We can't make such a professional tool a dummie proof tool, but tell people not to use HDRx as The solution.

HDRx should be used only if post-production can be worked on shot by shot with the original R3D files. Or lay only the A track if you go to HDCAM-SR...

Patrick
 
I'll jump in the fray aswell. We recently outlined the method we would use in our DI process to deal with HDRx, in a 5 min video. Might be of interest : blog.localheropost.com

What David suggested is right, in our opinion. If you're grading from the RAW, you have the ability to start pulling in the HDRx material as part of your color correct, and we treat it as a "minor" fx shot. Minor in the sense that its done by the colorist as part of the color correct, as a basic luma key operation.

I would shy away from trying to incorporate HDRx into dailies, although it clearly can be done ( possibly by using a simpler HDRx blend method, like in RC-X ).

To us, HDRx is an exciting new tool that can take a great A track and polish it to perfection. It makes sense to polish when you're in the polishing stage, color correct. I wouldn't rely on it as away to create an entire look for an entire project just yet, but as a way to get out of some sticky dynamic range situations ( think interior car look out bright windows ) it makes a lot of sense.

Just my 2 cents.
 
Yes, I understand how it could work for a TV show that goes back to the R3D files for the final online and color-correction, but what about a show that wishes to transcode once for dailies and then final color-correction later? Seems like you'd have to use RedCineX and do a quick blend to create a flatter 10-bit Log version that would have to then be color-corrected for Rec.709 dailies, perhaps using a special LUT for just the HDRx shots?
 
Yes, I understand how it could work for a TV show that goes back to the R3D files for the final online and color-correction, but what about a show that wishes to transcode once for dailies and then final color-correction later? Seems like you'd have to use RedCineX and do a quick blend to create a flatter 10-bit Log version that would have to then be color-corrected for Rec.709 dailies, perhaps using a special LUT for just the HDRx shots?

"If" your workflow was to transcode to say a ProRes HQ (for example) and you were going to stay in this world through finishing it seems fitting that you would transcode both tracks on set separately using REDCOLOR 2/REDLogFILM and hold onto the "X" track until you were ready to finish. Of course, this "LOG" file will not be ideal for Editorial and since there is no way to attach a LOG - LIN LUT per clip in Editorial it may be nice to transcode a separate ProRes HQ file on Track "A" with a one light.
 
Yes, I understand how it could work for a TV show that goes back to the R3D files for the final online and color-correction, but what about a show that wishes to transcode once for dailies and then final color-correction later? Seems like you'd have to use RedCineX and do a quick blend to create a flatter 10-bit Log version that would have to then be color-corrected for Rec.709 dailies, perhaps using a special LUT for just the HDRx shots?

Assuming you were going to HDCAM and then never going back to RAW... the HDRx motion artifacts mean that you wouldn't want to just blindly have the blend turned on in all shots.

So I'd turn it off - but if I were going through dailies and found shots that needed HDRx, I'd play with blending it in. If it looked great, I'd output the blended result to HDCAM SR.

If there were artifacts that meant I'd need to do HDRx blending in a more sophisticated way, I'd lay off the normal track and the HDR track separately.

It would be nice if we could just use a funky LUT and say "okay, all color values past 80% are from the HDRx stream, all values below aren't"... but you can't use that to separate out the blending for trickier instances.

But seriously... HDCAM SR? I hoped that the damn tape shortage made everyone start to think properly about a tapeless workflow!

Bruce Allen
www.boacinema.com
 
I would test doing it as a 2-step process to tape, first doing a simple blend in RCX but using redlogfilm as the gamma space, mainly watching the histogram to see where you are retaining detail in your highlights. Then render that out and correct it in whatever workflow you correct everything else and see how that works. I know the blend in RCX isn't that polished yet, but it would be worth testing.

tim
 
Well, tape isn't the point, I was more thinking of the shows that transcode to 10-bit Log, then make dailies from that.

Then my 2c is that if you're not able to go back to the RAW, the most foolproof is to just transcode out three ProRes files / DNxHD files / dpx sequences:
- one with the normal stream all the way on
- the other with the HDRx stream all the way on
- plus a moderate blend

Usually you'd only use normal file, but it's handy to have the other options available when you go into online - and decide to do a virtual grad on the sunset or whatever.

Bruce Allen
www.boacinema.com
 
Then the online workflow would be:

1. Online using non-HDRx track.

2. If something's overexposed, take a look at the blended version

3. If the blending sucks in certain parts, get the operator to use a mask to combine the good bits of the blended version with the non-HDRx track

4. If you have a skilled online operator and are still having issues - or want to do something creative like add a lens flare or glow onto only the highlight track, etc.... then pull up the HDRx-only file only onto V2 and add that to the mix.

For the VFX dept (sky replacements etc): definitely give them the HDRx-only track as well since it gives us great points to lock our stuff onto because of no motion blur (only problem is it's half a frame out of sync!).

Bruce Allen
www.boacinema.com
 
Then the online workflow would be:

1. Online using non-HDRx track.

2. If something's overexposed, take a look at the blended version

The realities of television post schedules, not to mention costs, make that a bit impractical in most situations. Since both tracks are included in the original R3d files, it's much more practical to design a post path that allows for direct use of those files and thus eliminates the need for separate transcodes and more complex conforms. There are a number of color platforms that support or will support this, so if HDRx is part of the mix, that's a bit more of an incentive to use that approach. This still won't completely prevent the need for custom motion blur treatments on the X track (at least for now) in the case of some shots with a lot of motion, but it does allow for Leo Marini's scenario of the colorist combining the two tracks during grading where motion is not that much of an issue. Which, potentially, will often be the case.
 
The realities of television post schedules, not to mention costs, make that a bit impractical in most situations. Since both tracks are included in the original R3d files, it's much more practical to design a post path that allows for direct use of those files and thus eliminates the need for separate transcodes and more complex conforms. There are a number of color platforms that support or will support this, so if HDRx is part of the mix, that's a bit more of an incentive to use that approach.

TOTALLY AGREE!

The best move is to choose folks who use online / color platforms that support HDRx / the original R3D files.

I think David was just asking about some kind of outlier case where they just want HDCAM SR / ProRes / DNxHD files?

As in "as a DP... how on earth can I get any benefit from HDRx if they just ask for ProRes files and don't want to deal with R3Ds"?

To which I think the answer is "generally, you probably can't... but you could give them stuff split out if you have the time?"

Bruce Allen
www.boacinema.com
 
Would it not be fairly simple (albeit slower to render) to add "automatic" motion blur to the X track when rendering/transcoding out of RCX? That is to say, a simple toggle that approximates/matches/simulates the motion blur of the A track? X track will have a higher shutter speed at fixed increases depending on how many stops above A track it is, right? For quick turnaround or simple implementation it should suffice, only requiring manual adjustments on super-high-end productions that can afford the additional time required to get the most out of HDRx (as in, big budget features.) Or am I completely missing something?
 
Would it not be fairly simple (albeit slower to render) to add "automatic" motion blur to the X track when rendering/transcoding out of RCX? That is to say, a simple toggle that approximates/matches/simulates the motion blur of the A track? X track will have a higher shutter speed at fixed increases depending on how many stops above A track it is, right? For quick turnaround or simple implementation it should suffice, only requiring manual adjustments on super-high-end productions that can afford the additional time required to get the most out of HDRx (as in, big budget features.) Or am I completely missing something?

You said the word render.

In the future, I agree totally: Realtime HDRx with motion blur added via some smart GPU algorithms would be awesome, sure!

But by the time we get that, I'd damn well hope we wouldn't have to ever deliver ProRes baked out files anyway and the software could handle the HDRx tracks natively in the online.

Bruce Allen
www.boacinema.com
 
It sounds plausible... Could a motion analysis and smart X-track blurring algorithm for RC-X be what Graeme has been working on lately?

Red Rocket gives you fast transcodes. Half-speed transcodes for HDRx would fit into a lot of fast-paced workflows, especially since you won't shoot HDRx all the time.
 
While in theory a pure r3d centric online is best, I don't really see that as a realistic option for most productions attempting to make a complex show with limited time and budget. I think there is a trend toward more productions going the straight-to-ProRes-or-dnxHD-and-staying-there route. While there are some drawbacks to this for TV productions, most of the time the advantages of cost and speed outweigh image flexibility in the online.

For a show deciding to transcode straight to ProRes or dnxHD I would recommend rendering out the A track and X track as 2 separate files. The offline editor can just concern himself with the A track, then the online editor and/or colorist can still have 2 files available for comping. It would be nice if RCX made this easier by having a mode where you could automatically render out the 2 tracks as 2 separate files.

Generally, I'm not in favor of baking in a "Blend" as I feel it compromises the A track density for a somewhat minor benefit from the X track. The best application of HDRx today still seems to involve some type of masking.

As far as motion blur, I'm also not really in favor of RCX doing any sort of motion estimation retiming, that is somewhat beyond the scope of the application IMO. Motion estimation really should never be considered a blind process or an automated process; your vector detail and block size should always be tuned to what the content motivates. So really it's not quite practical to do this for dailies. HDRx is still 'best' done as a vfx I think.
 
As a local 600 DIT, my opinion would be best informed depending on what the post-production house, Post-Production supervisor, and cinematographer want to do and what the producers are willing to spend, given turnaround time necessities.

If the post house is Red Friendly and TV friendly I can imagine they already have workflows in place that would allow for a layoff to 10bit on tape (not for grading from the tape) for distribution.

If they want to grade from the tape, well, rendering out a DPX sequence with a cinematographer and Post-Production supervisor approved blend, and then laying it to tape seems preferable to me?

Maybe we can get someone from the major TV post facilities to weigh in on this?

Would the show hypothetically be shot "all" in HDRX, or would it be used for certain high contrast scenes?
 
i agree with many posters here, Bruce in particular. These days, there is no excuse to NOT preplan an appropriate path back to the r3ds in final grading. And more and more TV shows are doing just this. In fact, many are demanding it now.

But directly back to David's point, if creating a 'new master', be it SR tape, or higher rez ProRes or AVID files, the only real way to provide necessary info. that may actually be useful is to process out BOTH tracks.

As Mike said though, TV schedules and deadlines dictate much of what will and won't be done. I would personally not use HDRx in dailies as deadlines will be tougher to meet, and quality is just not there with one simple slider to control this process. I would highly recommend any series put thought into this earlier than later and ensure they demand apprpriate treatment of their material at the finishing stage. There are lots of options out there now that can handle this, and handle it well. No need to create 'new masters' in this day and age. Well.....there shouldn't be anyways, but often there still is.....for now.
 
It is interesting to see all of the handwringing about ways to use HDRx. Old entrenched players desperately trying to squeeze any remaining value out of their investment. That is why they're conveniently trying to steer the post flow through the tape recorders. The fact of the matter, using old recording technologies, like HDCAM SR, in it's present form has no place, when one is trying to use the multi-stream material, like 3D or HDRx. Yes, I understand, that HDCAM SR still has very important role to play in today's post process, especially for archiving purposes, but it is also important to realize, that not every project should use those VTRs. Attempting to use VTR for the multi-stream project increasingly resembles overpriced Rube Goldberg contraption. It is obvious to me, that one can't expect a lot of change in the near future. But it is equally important to realize the importance of having an access to the metadata manipulation, as well as the presence of "X" track, which pretty much directs manufacturers to come up with innovative ways to preserve and utilize that data throughout the post process.
 
Back
Top