KOMODO Compression Qualities

[Elliot Lucas]

Aloha! I just spent the last few days with the Komodo and I have to say I love every bit of it. If I were to add one thing to this camera (this may be something you will add in future updates) it would be a "LOW QUALITY" R3D compression. I am coming off of using my scarlet dragon for many years and I loved how I could switch to 22:1 compression on projects that I didn't care too much about quality but still needed the resolution and RAW flexibility. Also great for being on a budget with memory cards and HD space. When shooting with the KOMODO I'm burning thru cards very quickly...especially in interview situations that I am mastering in 1080p and don't need things to be that high end. My only option is ProRes but Im a r3d guy now and cant go back!!! Thank you guys for everything you do!

[Blair S. Paulsen]

Bringing up such issues to help the RedTeam make optimum choices is one of the best aspects of this forum.

Best advice I have for the short term is to make at least one copy for safety, then do a one light grade on the R3Ds and bake out to ProRes. Alternatively, you could fuss with the look being sent out the SDI pipe and record that directly to ProRes on an outboard device. If you're happy with the ProRes version created during capture, you can just backup the R3Ds and move on. If you don't like it, you can still go back to the R3Ds and do a more careful treatment. The downside is having to add a ProRes recording device to the mix, the upside is the potential to have a ready to deliver ProRes version immediately after shooting.

Cheers - #19

[Elliot Lucas]

Thank you for the advice Blair! I really like staying r3d all the way. With the scarlet 5k it was like that. I could shoot 3hrs on a 512gb card all day. I like the small file footprint from memory card to backup and everything in between.

[Andi Loor]

I have also expressed the need to be able to do long recordings without having to change CF-cards every now and then. And by long recordings I mean around two hours. I have never gone below 8:1 compression so I have no idea what 22:1 would look like, and I prefer not to go that low. The Komodo is using another kind of compression, DCT, than other RED cameras. That has the disadvantage to use more drive space to achieve a certain quality, so to be able to record one hour there is need for 1 TB. Using 8:1, one hour would fit into 128 Gig on the Epic. That is a huge difference. Even though as Jarred claims, media is cheap today, this will impact workflow in a big way. Not only is the media cost to record one hour the same today as it was when the Epic was introduced, it will also make a big difference during the editing process. I am spoilt with using my laptop and one external harddrive during editing. This will only be possible in the future by transcoding everything into Prores or DNxHD first, and I will have 10 harddrives with original media at home instead of 1. Having said this, I'm not criticizing, just saying that the need for bigger drives or harder compression is important to many.

[Andrew Reese]

This was one of my biggest issues with DCT over Wavelet as a decision for Komodo and something I brought up right from the beginning.... Considering it's supposed to be a B-Cam for many and what I was going to use to intercut with my Gemini (A-cam) it's somewhat alarming to think that it has a higher data rate than DSCM2 cameras which is concerning in terms of editing and long term external storage requirements of adding Komodo into the mix.

For mini-docs or talking heads I may record anywhere from 1-2 hours of an interview on and off and usually rack up a minimum of 45 minutes of usable interview footage to cover all the bases. In this instance, using Komodo as a B-Cam for a second angle I would need at minimum a 1TB Cfast card or would have to pause half way through to swap 512gb cards out and that is just for a 1HR interview alone. On Gemini I have been able to shoot a 1HR interview and almost a half day of B-roll before I would have to swap a 480gb Mini-Mag. Similarly, for music videos where I do 3-4 takes of every lyrical scene (at various focal lengths) and may record continuous for up to 2-3 minutes for each take (sometimes at 48-60fps), the data is really going to start to add up on Komodo.

I already come away from most jobs with around 500gb - 1TB (1-2 480gb mini-mags) of RAW footage from my Gemini and I edit exclusively on my 15" & 16" MacBook Pros with 2TB Samsung SSD's as my go-to editing drives (already own 10 of them which travel with me to most jobs). If I plan on adding Komodo into the mix for a second angle for interviews and for some gimbal or specialty scenes I can expect to come away with at minimum an additional 1-1.5 TB of Komodo footage. For 90% or more of my jobs I have been able to fit each individual project on it's own 2TB SSD and the last thing I try to do is break projects out onto multiple drives as that makes editing an absolute nightmare. If I am going to rack up approximately 1TB of footage just from Komodo from a 1HR interview as a B-cam plus the interview and B-roll footage from my Gemini then I am going to be breaking the 2TB threshold for most projects and will need to resort to investing in larger drives or a RAID system which is not exactly cheap and absolutely not a mobile solution for a traveling DP. If Komodo is your first RED camera and your A-Cam then this is a non-issue but the minute you start using it as a B-Cam next to a DSCM2 camera then data rates are a lot more meaningful in terms of a limitation for a project which you want to keep below 2TB.

I was really looking for Komodo to compliment DSCM2 as a smaller, less power-hungry, and equally data friendly option, not for Komodo to become a beast of its own in which it would have to be treated for it's own needs for additional media storage due to higher data rates. The benefits of global shutter and its form factor are obvious wins but considering its other limitations such as DR, Low-Light Sensitivity, and Frame Rates I wouldn't have dreamt that it would be pumping out data rates much higher than my Gemini. I can obviously cut down on how much I would use Komodo in the mix, which to me kind of defeats the fun in owning one if I constantly have to asses and limit where and how much I can use it on a job as a B-cam. Komodo will still serve its place for me in providing all the unique hard to achieve shots that were more difficult for its larger DSCM2 brothers (i.e. car mounts, gimbals, inconspicuous hand held shooting or travel) but for using it as a B-cam on long interviews, doc's, or event shooting where multiple camera angles are a necessity this is somewhat of a deal breaker for me and i'm almost wondering if sticking with DSCM2 as a B-cam is the better option to keep most projects at a reasonable data rate. Keep in mind you can pick up a used Scarlet-W for about the same price or probably even less than a built out Komodo right now which comes with much higher frame rates at higher resolution with less crop and with the benefit of lower data rates and better compression options. It's definitely not as fun and compact as Komodo and there's no global shutter but the Dragon sensor is indeed beautiful in the way it renders images.

Personally I like to shoot at low-compression as IQ is important to me and I find Gemini does indeed produce more beautiful images at lower compression but for most of my work I can't see having projects turn into greater than 2TB of footage per day as the cost to benefit analysis start to come into question. On large productions with huge budgets this is a non factor but for run-and gun crews and smaller commercial DP's we can't have our projects turning into heavy hitters in terms of data and storage requirements as I can't justify any benefit to the client other than increased time and cost in terms of file management, editing, and color. If I was a documentary shooter looking at Komodo I would take a long hard pause and question will you be able to manage the data rates realistically for all day shooting (i.e. 2-4TB).

For example, if you shoot say 3 projects a month at 2TB each (2 X DSMC2 cams at 1TB per project) you have to back-up, edit, and offload 6TB of footage (i.e. 12TB with back-up) per month. If Komodo as a B-cam takes each project to even an increase of 3TB instead of 2TB (1TB DSMC2 + 2TB Komodo) this not only reduces your ability to use anything but large volume editing drives and RAID systems (removes your ability to edit on-the go) but it also increases your monthly storage requirement to 18TB vs 12TB and essentially doubles your long term storage requirements. At first glance the higher data rate seems like you may just have to buy an extra CFast card but when you really start to think about managing the files from multiple projects on the go throughout a months workflow the data rate from Komodo becomes a lot more important as many of us already have a small fortune in media storage.

[J. Orlandi]

After playing and pushing the new compression, I would say that MQ is the lowest you'd want to go for anything important. I can noticed quite a bit of compression artifacts in MQ when looking and pushing. HQ is quite nice. I'm not sure going below MQ would be wise for anything critical. That being said, having the choice is still nice. I would actually say if you're in a situation where you need long recording times, just switch to ProRes, it gives you TONS of time on the cards, and you can still have lots of color info in 422HQ to grade.

[M Harvey]

Phil started a thread you guys might want to check out that explains some of the basics of the new compression scheme:
http://www.reduser.net/forum/showthr...-RAW-Explained

Among other points:

Phil says that an "LQ" (more compressed R3D) option is on the way, but that they're still working on it.

Jarred also chimed in to say that they're still tuning this stuff, but that it's expected that MQ will eventually record at a lower bitrate.

Part of what all this is about is achieving the same visual quality with the significantly lower power draw demanded by Komodo.

A lot of people also raised similar concerns as you guys have about storage space and record times.

[Robert Hofmeyr]

Jarred also mentioned in the thread linked above that the new compression format in Komodo will also be used in DSMC3, so it's not all about limited power and lower cost of Komodo. I'm waiting to see where they land when they have optimised the compression, and for some comparison tests showing HQ vs MQ vs LQ (when it is available). If the data rate ends up being significantly higher for similar quality, as compared to DSMC2, then the additional media, storage, and transfer times need to be taken into account.

[Christian Jadot]

After playing and pushing the new compression, I would say that MQ is the lowest you'd want to go for anything important. I can noticed quite a bit of compression artifacts in MQ when looking and pushing. HQ is quite nice. I'm not sure going below MQ would be wise for anything critical. That being said, having the choice is still nice. I would actually say if you're in a situation where you need long recording times, just switch to ProRes, it gives you TONS of time on the cards, and you can still have lots of color info in 422HQ to grade.

Can you show some samples of the MQ compression artifacts, and comparisons between HQ and MQ?

[J. Orlandi]

Can you show some samples of the MQ compression artifacts, and comparisons between HQ and MQ?

Sure, give me some time to shoot some new tests here. It's been a crazy few weeks since I got the camera.