Monstro VS V-Raptor…….

[Andrew Reese]

I still believe wavelet compression to be superior to that of DCT. I personally made a big stink about this when Komodo first came out when I was on the stormtrooper list and I saw that Komodo produced larger file sizes than my Gemini or Dragon even at the same resolution. I immediately knew something was up with Komodo's compression and flavor of the RED codec. After enough complaining in the Stormtrooper group, RED finally did announce/admit that it was indeed utilizing DCT based compression over Wavelet in Komodo, which at the time people seemed to defend due to many having limited PC processing power in the prosumer market and liking the idea of easier editing.

I however edit Gemini/Dragon/Monstro wavelet based files just fine on several maxed out MBP's with little hiccup and can even edit 8K on the go, so I was a little disappointed when I saw Raptor is also DCT nonetheless. In my experience with my larger projects, file sizes are more of a limitation to the client than processing power is to the editor or colorist. This also greatly increases the cost of doing business for most clients on budgets where storage starts to ad up. Yes storage is becoming "cheaper" in 2021, but it is still not "cheap" by any means as a standard 1tb of SSD will typically run you about $300 if you want reasonable drive speeds. I try to keep most of my smaller projects below 2-6TB, which has allowed me to run the entire project off a few portable SSD drives (Samsung T5/T7), which is crucial for documentary work and editing projects while on location. Remember, not everyone is working on projects that have budget for a standalone 40TB raid set-up and a post-house editing and coloring there projects daily's. Many projects and almost all of mine, still require some level of on-the-go editing across several states or countries, with quick turnaround times and no time for offloading files half-way across the nation to a post-house. Smaller file sizes is crucial for some of us and that was a main reason I am stilling shooting 5k wavelet based REDcode with my Gemini and my clients keep asking for it. For most it seems to be the sweet spot of image quality versus file sizes.

Keep in mind "compression" was a huge marketing factor for 8k originally with Monstro/Helium user's who could shoot 8K with high compression (16:1) to later downsample to a 4k final in order to "clean-up" the highly compressed low bitrate footage with little compromise on file size due to higher resolution. Now with DCT, not only are file sizes higher to begin with, but we currently are stuck with HQ,MQ,LQ and have even less options for compression.

To me this kind of goes against the original methodology of RED who marketed shooting 8K 16-1 OR 8:1 and giving us OPTIONS and having it be visually loss-less in a 4k timeline.

Now that I am done with my rant, I believe the switch to DCT for RED was really more about dumping heat through the utilization of a less processing intensive codec at the price of larger storage volumes. Simply put, we keep asking for smaller camera bodies that shoot 8K HFR and there really just is no way to cool a camera capturing and drastically compressing that many 16-bit frames in concession.

Arri speaks often about how they could have given much higher frame rates in the Alexa Mini LF but at the cost of the camera body being larger than it already is.

DCT compression is a trade-off RED is currently willing to make in order to keep form factor small.

Maybe we will see different flavors of RED Raw (i.e. wavelet /DCT) and better compression options in Raptor XL or other future DSMC3 cameras to come.

 

[Ryan Sims]

DCT file size is the main reason I did not buy a Komodo, even though I had a Hydrogen discount option. Massive file sizes don't work for my current mode to operating. I'm generally the DP/Camera OP/DIT.
I shoot all day with 2 cameras rolling with enough cards on-hand to save doing all the DIT work of transfers that night off-set. Not ideal, but a reality of indie-film making. With DCT file sizes going up almost 3x to get the same quality, I now will need 3 times as many cards, hard drives, and then lose 3 times as many hours of sleep (and I'm already working too long each day)

 

[Brian Timmons]

It's very important that people look into the entirety of their workflow before deciding if they should get it.

Hey Christopher-

I absolutely agree with this statement. I've seen some productions get into a bit of trouble (and occasionally I've helped get them out of the deep end) for not properly test driving a post workflow before starting a project.

Below is stuff I'm positive you already know.
Just repeating to remind you of the solutions I'm sure you know exist.

As you pretty much said, RED is primarily a camera company that is concerned with all image acquisition issues in production and the follow through of that footage into post. I always felt that REDCODE and any cinema camera codec should be primarily focused on getting the best images that best stand the test of time.
Provided that tools are given to help aid the variety of potential post production needs (i.e RED CINE X, compatibility with a variety of post production tools) this support for post can and has been successfully done by RED and other camera manufacturers.

The new REDCODE flavors are definitely an adjustment in terms of data size but I do think think the tools already exist to adapt to current limits in post infrastructure.

The biggest workaround, being that you transcode your 8K REDRAW (now DCT based) footage to a codec that plays nice within your infrastructure for your editorial needs and go back to your original camera files for final color and output.
Nothing revolutionary there as it is a tried and tested workflow used by a huge majority of productions for decades now.

I've worked several years in broadcast post production and can definitely say the vast majority of post houses prefer working with offline codecs like ProRes proxy or DNX 36 and the like for editing. Many editors and some assistants don't want much to do with RAW files. Besides conforming at pic lock they leave RAW and camera originals up to the colorist.

Most servers are also used more for the offline edit. Unless it's a project with a huge budget, it's quite unusual for most post houses to store camera original files on a server usually due to the very storage and data rate issues you mentioned. This has always been the case and is not unique to Raptor, Komodo or any other heavy camera codec.

My experience is that most online bays for color correction tend to have their own local drive setup and or dedicated server so they can deal with the needed storage and bandwidth to push those files around. That's a cost that's a usual part of the online bay along with the high end monitoring, souped up finishing desktop, cappuccino machine and other client amenities.
Even then, many colorist may still prefer transcoding and grading in higher end flavors of ProRes, DNx, and the like and go back to the raw file only when needed.

The approach that is picked fits within what the post house can handle and what the client pays for if they want more.

Yes - grading in RAW is easily the most satisfying approach but it's not always the most used and there are alternatives that fit a variety of budgets, approaches, etc.

Again, I'm sure you already know all of this. I'm reiterating that the storage and bandwidth issues have always been with us but the tools and workflows have been well established to deal with those issues.

Progress is a double edged sword. I understand your frustration but this is a workable problem.
In time, I think post infra structure will surely catchup .... and then greatly fall behind again once RED releases 16K 32 bit color cameras using holography and a new bleeding edge codec with a time sliced fractal image engine.

Wash, rinse, repeat.

Brian Timmons
BRITIM/MEDIA

 

[Phil Holland]

I still believe wavelet compression to be superior to that of DCT.

I would posit that 300MB/s of RedCode (JP2K variant) is roughly equal to 800MB/s DCT.

I think you're both missing where each method shows it's strengths.

The actual advantage of Wavelet encoding comes at lower data rates/higher compression ratios. But DCT's strengths come on the opposite end, i.e. better quality at lower compression ratios.

I've been a rather vocal advocate that you shouldn't be rolling say 22:1 REDCODE RAW on DSMC2, but what you really don't want to do is get near that with DCT. In fact I don't roll beyond 12:1 in the modern era on DSMC2 as that is my visible and testable tolerance for most content, and honestly I'd prefer to keep it higher data rate than that at 60fps 8K.

I’m merely pointing out the existing ability of generating an R3D still from an R3D (which I believe was always wavelet in the past) - means the structure exists to already output a wavelet R3D format in the software.

You're missing the part where an R3D Snapshot or R3D is not re-encoded at all. It's a slice out of the REDCODE RAW material. It's not outputting, converting, or encoding. It's just a slice.


Thoughts and Theory Post incoming about what is going on.

We're in a weird time where all filmmakers are looking for cameras to be swiss army knives, which is cool because most of the premium tools truly can be. Compressed RAW with variable data rates, ProRes in multiple encode types/data rates, and occasionally other desires to respond to workflow demands, data footprint, or roll time considerations. In terms of making much, much smaller encodes RED could explore H.265 RWG/log3G10 or other Color/Gamma encodes for every small direct turnaround encodes. But these cameras aren't exactly that targeted at that world. Could be cool though.

The key selling point outside of the sensor technology and hardware with these cameras really is what their RAW performance can lead to. I know. I know. Some films have explored ProRes and such, but if you want the best the cameras have to offer, RAW and Compressed RAW is where it's at. There's been a lot of smearing of that mindset mainly due to other cameras trying to punch up in that playground, but it's still true no matter the manufacturer.

Another big thing and oddly one that seems to have been forgotten, but is perhaps the biggest one as it pertains to filmmakers and what a digital cinema camera can and should do. Original wavelet based REDCODE RAW 3:1 has been considered "visually lossless". We typically roll at 5:1-8:1 in the world of higher end productions and anytime I dipped down in resolution or aspect ratios I was exploring 2:1-4:1. Yes. the concept of visually lossless quality is compelling mostly to the world of high end productions, but that is "the quality bar" when it comes down to it that we make decisions based on.


Here's how I see it as of now. V-Raptor with newer REDCODE RAW and increased data rates are better than or close to visually lossless (i.e. HQ = better than visually lossless, MQ = visually lossless, LQ = a smidge below, but likely undetectable or even perhaps visually lossless). The reason I'm saying this is I just showed LQ material side by side with HQ material on a rather large 8K display (pausing, replaying, and A/B'ing) with high frequency detail and I'm having a hard time seeing any difference up close. This performance across the board is a notable difference in my mind and I suspect one of the reasons this new hardware was created.

I'll dig into this further in regards to how productions should roll (i.e. best practices), I have a whole line of brutal encode tests lined up, but right now I'm not seeing much reason to "decide" between HQ, MQ, or LQ as we have done in the past. At least without making notable visual concessions or compromises. Now REDCODE RAW essentially looks the same no matter what encoding strength you choose which is floating around better than or visually lossless, which I think in the back alley spit contests of compressed versus uncompressed RAW uniquely gives RED an advantage once again when it comes data footprint and image quality.

 

[Brian Timmons]

Maybe we will see different flavors of RED Raw (i.e. wavelet /DCT) and better compression options in Raptor XL or other future DSMC3 cameras to come.

I do hope a 10:1 or perhaps 12:1 version of the new REDCODE gets released.
Just hope they don't call it "Very Low Quality". That won't go well with any client.

Brian Timmons
BRITIM/MEDIA

 

[Phil Holland]

I do hope a 10:1 or perhaps 12:1 version of the new REDCODE gets released.
Just hope they don't call it "Very Low Quality". That won't go well with any client.

I'm not a fan of the "Q", but logically it makes sense.

Data Rate or Bit Rate could work, but:

HDR is taken!

But perhaps HBR, MBR, and LBR make sense. Or even RC-H, RC-M, RC-L.

But not two characters really and that deal read sexy.

Maybe HR, MR, LR. But that has weirdness to it.

 

[Brian Timmons]

DCT file size is the main reason I did not buy a Komodo, even though I had a Hydrogen discount option. Massive file sizes don't work for my current mode to operating. I'm generally the DP/Camera OP/DIT.
I shoot all day with 2 cameras rolling with enough cards on-hand to save doing all the DIT work of transfers that night off-set. Not ideal, but a reality of indie-film making. With DCT file sizes going up almost 3x to get the same quality, I now will need 3 times as many cards, hard drives, and then lose 3 times as many hours of sleep (and I'm already working too long each day)

I hear you with the new file sizes. I think we were spoiled a bit with wavelet.
The media (even with CFAST and CFEXPRESS) is at least much cheaper than before.

I also think LTO tape for storage and archive can play a role with managing these new file sizes.
I've been backing up to LTO since getting the RED One back in the day but it is still one of the cheapest per GB and
more durable means of storing footage.

New ones are still pretty pricey but you can find some older new and used LTO drives on eBay for a decent price.


Brian Timmons
BRITIM/MEDIA

 

[Mike P.]

I dunno... When/if it's between a used Monstro at ~$12.5k or Vrap at ~$25k, it'll definitely be a difficult choice for a lot of people. 60 to 120 fps wouldn't be enough of an incentive. Nor would cleaner shadows (when Monstro is already good to go at ISO1600). On paper, there's nothing about Vrap that would really make my work easier/faster/better.

More than the mount or DCT vs Wavelet, the biggest concern for me is the power draw. What's been stated is a bit of a dealbreaker (and a shame given the strides made with Komodo). After years of using heavy-ass 100~220wh batteries, I love getting ~1.5hours on DSMC1 and 2 with 1lbs 100wh mini-v-mounts (which can recharge as fast as they deplete). I shudder at the thought of going back to traveling/lugging around a bunch of big batteries. It's one-step forward, two-steps back.

I think a Komodo VV for ~$10k would be a better camera overall (with or without GS). Limited framerate would be fine (8k/40p, 6k s35 at 60p, 4k/120, etc), just so long as it retained that 3hours runtime with ~1lbs worth of batteries. Alas, by the time that happens, Canon will have the R52 and BMD will have a Pocket8k for $2500 (with no pre-order or 6mth wait time).

 

[Blair S. Paulsen]

Each codec has its "sweet spot". With wavelets, which are entropy based, one can push beyond that and the primary artifact is general softening of high frequency detail. With an 8k, 6K or even 4K capture - down scaling to 1080 HD typically yields as much detail as the final viewer will ever see. Score one for wavelets.
DCT, OTOH, has pretty nasty artifacts - I particularly hate macro blocking - when bit starved. If such artifacts are induced at capture, getting rid of them is almost impossible and smoothing them out kills detail. I'm assuming RED got that memo and decided to keep the bit rates high to avoid the nasties. I believe that's why even LQ has a fairly high data rate.

If the higher data rates of DCT based R3Ds are an issue for how you work - and/or those you work with - then the DSMC2 lineup; Helium, Gemini and Monstro are still excellent cameras. The latest GPUs have made decoding wavelets less of a burden. New architectures like Apple's SOC designs are making portable solutions more capable. Etc.

Phil is seeing DCT at LQ hold its own with HQ. Impressive, but not surprising. Cosine codecs do well when given enough data rate.

Cheers - #19

 

[Björn Benckert]

I sort of have a preference for wavelet myself and think the performance on most modern systems generally made playback a non issue.

Wonder if RED is taking longer view of things using DCT particularly where it relates to getting higher performance (resolution, frame rates) down the line.

One simple question is what would it take RED to get similar V-Raptor performance if they stayed using wavelet?
What would be the effect on the size, weight, power requirements, and heat dissipation/cooling requirements? Would it be anywhere near as quiet?

I think that would likely answer why the move to DCT.

I doubt Komodo would get 90 minutes on a single Canon BP battery or would even have such a responsive user interface and experience if it was based on wavelet. Despite the small size of Komodo, it's clearly the most quiet camera RED ever made.

It's likely that the overhead to perform wavelet compression and the hardware needed creates other performance issues and limitations that RED felt a need to move away from.

Very much agree Blair with your statement on trade offs.

Brian Timmons
BRITIM/MEDIA

Cost is also a factor, DCT compressing cameras are likely way cheaper to bring to market. People want less costly cameras / the competition demands it.

DSMC2 compression creates more heat which is equal to drain more batteries. Basically heat is pretty much all a camera creates, sure pictures as well but the heat created is pretty much equal to the poser consumption.

The sluggish GUI of DSMC2 cameras has likely nothing to do with what compression is used. More likely it's a whole new operating system in Komodo. The DSMC2 stuff looked quite old allready when launched with Epic. Don't know what is the back bone of it, but likely komodo and Raptor got something completely new as there must be some sort of reason not only to change the gui but also most of the functionality is at least temporary lost with komodo and raptor gui functionality also seams to build more on the komodo interface then the interface of DSMC2.

 

[Brian Timmons]

I'm not a fan of the "Q", but logically it makes sense.

Data Rate or Bit Rate could work, but:

HDR is taken!

But perhaps HBR, MBR, and LBR make sense. Or even RC-H, RC-M, RC-L.

But not two characters really and that deal read sexy.

Maybe HR, MR, LR. But that has weirdness to it.

They might take a page from Panasonic's old DV tapes
and do Master Quality, Premium Quality, and Standard quality.

Perhaps "Budget Quality" could be a 4th category.

It's a good subliminal statement to a client that you're filming in "Budget Quality"
because they're too cheap to buy decent storage for the project. :smile:

We've all had that conversation at least a few times.
Hard enough to get people to make additional LTO backups.

Brian Timmons
BRITIM/MEDIA

 

[Mike P.]

What about ditching LQ and going Standard Quality (SQ) and having a lower one called Extended Quality (EQ)... Bringing back VHS baby! Ultra Quality, High Quality, Standard Quality, and Extended Quality all sound better than Medium and Low Quality to me (and I presume clients).

Regardless of what they choose, pretty sure people will complain anyway. But judging by the DCT discussion, it makes sense why they're unwilling to go too low (cause that will make the image fall apart in gross ways [re: macro blocks]).

 

[Ryan Sims]

I also think all the heat and power savings of DCT processing have been loss and more to the new thermoelectric cooling. It may have reduced the fan noise, but at the expense to burning the batteries much faster.

 

[Nick Morrison]

DCT file size is the main reason I did not buy a Komodo, even though I had a Hydrogen discount option. Massive file sizes don't work for my current mode to operating. I'm generally the DP/Camera OP/DIT.
I shoot all day with 2 cameras rolling with enough cards on-hand to save doing all the DIT work of transfers that night off-set. Not ideal, but a reality of indie-film making. With DCT file sizes going up almost 3x to get the same quality, I now will need 3 times as many cards, hard drives, and then lose 3 times as many hours of sleep (and I'm already working too long each day)

Keep in mind we found that with LQ the Komodo files sizes were pretty comparable in size and quality to Dragon 6:1.

It's all we shoot on Komodo.

 

[Tom Gleeson]

If someone has access to a Raptor it would be best to do actual tests in regards to the Wavelet vs Cosine. No point talking about it or even sharing impressions. We need hard data. From what Phil has said I suspect he is already on to it but I would love to see Raptor vs Monstro (Wavelet vs Cosine) compression test. I would be good news if MQ and even LQ were good enough for general shooting and HQ only for VFX or heavy processing work. Happy to do these tests myself if I can get my hands on a machine

 

[Brian Timmons]

RedUser seems to turn into an interesting social experiment whenever RED announces a new camera.
The range of emotion from excitement, elation, fear, doubt, frustration, anger, and gradations in between is kind of fascinating.

I have a hard time seeing Monstros, Heliums, Geminis, Dragons or most other REDs being somehow lost and forgotten with V Raptor.

Those cameras each have state of the art sensors that still perform extremely well if not best in class.
The older wavelet version of REDCODE they use is already shown to have some potential advantages that make it easier to work with for some compared to Raptor and other cameras.

I don't know everyone's financial commitments and the nuances of their work and clientele but I truly think if people were doing well with their cameras in the past they will likely continue to do well for the next few to several years.

We just had a recent thread where Epic MX turned 10 years old and some users stated how they are still getting work done with them.

I'm still shooting with a 13 year old RED One MX along side Komodos. Komodo is definitely a huge improvement but the RED One MX STILL holds it's own.


Regarding V-Raptor- the camera shoots 2-4x the resolution, has 2-4x the frame rate, is smaller, lighter, and is more than half the cost of other competing cameras (Alexa LF, Venice). The camera file sizes are either comparable if not considerably smaller than the codecs used on those other cameras.

I don't see anyone who already invested in RED cameras or a future camera like Raptor taking an "L" on this.

To me it's pretty much an all around win.

One thing I've learned (part of the reason why I was able to survive with an older RED) is the establishment of RED as a brand.
It obviously varies based on the client but many people don't know the specifics of each camera model. All they tend to hear is that it's a RED. I still occasionally have people take selfies with my bruised, battered, scarred, but hard R1 for this reason.

Whatever people's anxiety (as there seems to be a bit in this and other threads) I think all will work out.

Every RED camera made gives a professional what's needed to get excellent images from it.
The kind of images people notice to get further gigs and advance the craft.

.... I'm done.

Brian Timmons
BRITIM/MEDIA

 

[Brian Timmons]

Keep in mind we found that with LQ the Komodo files sizes were pretty comparable in size and quality to Dragon 6:1.

It's all we shoot on Komodo.

With 6K and 5K that's about where I'm at as well. LQ seems to work well though I have not gone into full on deep post work with Komodo just yet.
A part of me thinks 4K and 2K might benefit from MQ and perhaps HQ though.

Brian Timmons
BRITIM/MEDIA

 

[Russ Fill]

I find it kind of interesting that there is a lot of talk about the post process and heaver files and all with Komodo and Raptor. But when the R1 came out it was almost the same sound of big files and how will post handle all of this and etc....
At the rate of technology progress I don't think it will be very long before we will all have laptops that handle these heavier files. And if Red has shown us anything its that you will always be surprised by some update that shows up and your camera seems new again.

Here is a question Id be interested to see some of your comments on.
RedCine X pro has been around for quite a while. Would there ever be a possibility of a new version of RedCine that handles the new cameras files much in the same way but Exclusively for the DCT camera files.
Love these threads.

 

[Nick Morrison]

With 6K and 5K that's about where I'm at as well. LQ seems to work well though I have not gone into full on deep post work with Komodo just yet.
A part of me thinks 4K and 2K might benefit from MQ and perhaps HQ though.

Brian Timmons
BRITIM/MEDIA

Another thing to keep in mind too is we've found that that with a dedicated GPU or EGPU that Komoodo files have been faster/easier to transcode.

For example, even on an older 12 Core Mac Trashcan with two D7's, the render times were as follows:

5K Gemini to Prores HQ: 21.5fps
6K Dragon to Prores HQ: 19fps
6K Komodo LQ to Prores HQ: 37fps

 

[Brian Timmons]

Another thing to keep in mind too is we've found that that with a dedicated GPU or EGPU that Komoodo files have been faster/easier to transcode.

For example, even on an older 12 Core Mac Trashcan with two D7's, the render times were as follows:

5K Gemini to Prores HQ: 21.5fps
6K Dragon to Prores HQ: 19fps
6K Komodo LQ to Prores HQ: 37fps

Cool test. I figured this but these are the 1st tested number I've seen.
Was this with RCX?

Brian Timmons
BRITIM/MEDIA