Description image

Luke Neumann on Canon 5D Mark III RAW Video and a Comparison with the RED EPIC

Luke Neumann - Aviv Vana InterviewWe’ve seen a few tests putting the Canon 5D Mark III with the RAW video hack up against lower-end cameras, and only one, the Blackmagic Cinema Camera, is capable of shooting RAW as well. Now, we’ve got a comparison with another RAW camera, the RED EPIC, from Luke Neumann (who has done a few Magic Lantern tests before), as well as an interview with Big League Film School talking with Luke about how this is impacting the DSLR community.

Luke has shared some sample DNGs and a .R3D file since comparing them compressed to Vimeo isn’t as useful (head on over to Luke’s site to download them). Top is EPIC and bottom is Mark III RAW (again, download them and try them out for yourself, there are many variables here):

RED EPIC - Compared to Mark III

Canon vs RED Comparison - Canon

I tried to get them as close as possible, but obviously it’s a bit of judgment. Keep in mind that a program like Adobe Camera RAW will add a bit of sharpening either way to the image (unless you manually set it to 0), and it’s unclear how Canon is getting the full image down to 1080p. Regardless, this is very, very impressive, and the detail coming out of the Mark III RAW is just fantastic. The fact that it can keep up with the much more expensive EPIC downscaled to 1080p shows how good the sensor is on the Mark III (and all of it would not have been possible without the work being done by the people over at Magic Lantern).

Here is Aviv from Big League Film School talking with Luke (he apologized for the audio on his site):

We’ve shared a lot of things from Luke on this site before, and that’s because he’s has usually been at the forefront of the Magic Lantern testing, and he’s produced some incredible results. I think his advice is fantastic, and it’s exactly what I would tell people starting out: maximize the current tool you’re using and always keep learning. I have no doubt every professional learns something new every time they step on set, even if it’s something small.

If you missed his Mark III RAW tests, check them out below:

See more at Luke’s site and at Big League Film School using the links below.



We’re all here for the same reason: to better ourselves as writers, directors, cinematographers, producers, photographers... whatever our creative pursuit. Criticism is valuable as long as it is constructive, but personal attacks are grounds for deletion; you don't have to agree with us to learn something. We’re all here to help each other, so thank you for adding to the conversation!

Description image 71 COMMENTS

  • The write up confused me. What is meant by, “Keep in mind that a program like Photoshop will add a bit of sharpening either way to the image…”? If it means what I think it means, it is wrong. Photoshop doesn’t just do things without you doing it. If the images are resized in photoshop then an algorithm will be used (and CS6′s default should be changed) but no auto-sharpening.

    • That should have read Adobe Camera RAW. Camera RAW’s default is set to 25% for sharpening, and the demosaic algorithm can change the apparent sharpness of an image.

    • Actually some downsampling algorithms like bilinear and bicubic typically have actuance artifacts that show up as a halo or fringe similar to an unsharp mask effect.

      For comparison, lanczos is a better (but more computationally intensive) algorithm…Premiere’s GPU Mercury engine uses lanczos, whereas the software mode uses bicubic and there’s a very notable difference in quality. As far as I know Photoshop only has bilinear and bicubic.

      • For sure, which is why I made the clarification about image resizing being a possible cause but was unsure if that was going on in this.

        And forget exactly what it is but the auto bicubic setting in PS CS6 is common to have an over-sharpened look and should be changed.

  • RebelPhoton on 05.23.13 @ 7:03PM

    How does exactly a 1080p no-one-knows-how-the-photosites-were-sampled-RAW feed “keep up with” a 5K compressed RAW feed?
    Please be serious.

    • One has a solid post workflow and file sizes that aren’t insane, and the other is a DSLR. People will latch on to the RAW part of the comparison and think that the two cameras are evenly matched.

      • I agree with those that are baffled by the Epic/5D comparison. On my site I noted “I think the key is not to see who wins here but to see just how close the 5D3 raw footage can get to something on the caliber of an Epic. I for one am blown away!”

        This is the point for me.

    • At 1080p, should have clarified that.

      • And what exactly is the point of that? Scaling one to match the resolution of the other?
        That’s like comparing a Bugatti Veyron with a Mustang and saying they’re both pretty good at 60mph.
        Ridiculous comparison.
        This is a great, great mod for the M3, but lets not get carried away.

        • Because where are you going to be showing that 4K or 5K video? The valid comparison for the cars compared to cameras would be 0-60 time, something you can actually do without breaking the law.

          If your final resolution is 1080p, which it is for most people right now, 4K or 5K has a very specific use, either as future-proofing, or for the ability to reframe in post. Otherwise, if you’re delivering something in 1080 and there is no need to move the frame around significantly, the differences come down to usability rather than picture quality – and picture quality at 1080 is the topic of discussion.

          • Aaaand you got carried away.
            The 0-60 comparison analogy is silly because there’s a rev limiter on the Veyron.

            My point is, one machine is being deliberately crippled to better match the lower end model it’s being compared to, and that can never be a fair comparison – for either model.
            If you were strictly comparing low light performance or dynamic range that would be one thing and I’d get behind a test like that in a heartbeat, but when it comes to sharpness and image detail I think it’s a pointless exercise. Compare it to the 2K RAW output of the F5. That’s a 2K RAW output derived from a higher pixel count on the sensor so that would actually be a valid comparison to this hack. But scaling a higher resolution image down to gauge resolution is a silly exercise – and that’s without even mentioning the various different scaling algorithm options.

  • Erwin (Netherlands) on 05.23.13 @ 7:05PM

    +1. Photoshop won’t do sharpening unless you tell it to.

    • how can you prove that? just because the slider says “0″, it doesn’t actually have to mean that no sharpening whatsoever is applied. don’t forget, we’re talking about raw here. the image has to be interpreted (debayered/demosaiced) anyway, before you can actually get something usable from it.

  • On the RED I can tell there is a door behind the screen, 5dmkIII not so much. But ML has done some great work!

    • I can see other things too like how brilliant the flowers look with the Red, less so with the mkiii.

      But considering the cost difference between the cameras this comparison should make lower budget film makers happy. It should also confirm to those with mega budgets that they really are getting a more brilliant, more beautiful, more detailed picture for their money in the Red. I think this comparison is a win-win.

      • The difference is so negligible that the receiver will not perceive the difference between the two.

        What I see is that the Canon footage is a bit more sat than the RED, but again… tiny differences.

        If you are delivering for HD, and when ML is finalized and when they have fixed the DNG workflow (which I have a feeling is a matter of days rather than weeks or months) then you would have a HARD time justifying using an Epic for delivering HD work. It’s nothing but faboi BS….

    • Did you mess with the files at all? The 5D you can pull the door out easy – it’s RAW, that’s the whole point. On the embedded pics. the RED shot is just exposed a bit hotter, look at the sky for example.

  • vinceGortho on 05.23.13 @ 7:20PM

    Can HDR televisions display 14 bit color?
    I was told they were only capable of 10bits.

  • The RED is a bit flatter and has better colours but the “No Trespassing” sign is unreadable.

    • I wonder what the original 4k Red image looked like.

      • Go to Luke’s site and download the samples to see the original files.

      • Seriously…?

        Who the hell cares? I’m pretty sure both images received the same tender care, and in the end it’s all about the delivery.

        When we start having 4K tvs at home in general then you can worry about the 4K acquisition…

        • Sure seriously. If the Red is not shown at its best that makes a difference in how it looks.

        • There aren’t many Red users commenting on this site. There’s quite a few Canon users here. And I don’t think there’s a Red RAW channel in vimeo, or a lot of people running comparisons of their Red to the MkIII. So there is no defense of the Red going on. I don’t think Red users feel the need to defend their camera. It makes better video than the MkIII. The only thing that is happening is MkIII users are seeing an improvement in their video. And they needed it since even the $1300.00 GH3 and the hacked GH2 were making better video than it. (will there be a hacked GH3 soon?) I don’t think big budget Red users are contemplating switching to the MkIII.

          Just trying to keep things in perspective.

          • For sure it’s a great image from the ML hacked MkIII but there’s a ton of reasons this is gonna be a hobby tool rather than a cheaper sub out for an Epic or Scarlet or even a C300. It comes down to user friendliness, proven established workflow, ergonomics, accessories, rolling shutter, sound recording, metadata, reliability, timecode, professional connections, servicing, warranty etc. Still amazing, and very cool.

          • Red and Alexa aren’t losing sleep over this hack. I don’t think they’re even losing sleep over the 1D. Canon must feel the GH breathing hot down their neck though. If Panasonic is able to get 4k video into the GH4 (even if it means making a bigger body, which some GH users won’t like, because after all, the GH is all about being more convenient, and better, but maybe they’ll sacrifice some convenience for 4k in return, and end up with the better) they will move beyond giving 5D3 users fits to giving it to 1D users.

            Have you seen this GH3 video? I think it’s a beauty! Make sure you view it in 1080p!

  • At 1080p they’re definitely close enough to intercut pretty seamlessly between the two. Very impressive work by Magic Lantern! But once 4K UHD televisions are being bought by regular users, that’s where the 5K Epic footage will truly outshine all other 1080p cameras.

  • I’m not a Canon user but I actually think the Canon looks better in this. Look at the red NO TRESPASSING sign. It looks a lot better on the Canon. On the Red its basically unreadable and you can only tell what it says from reading the Canon one.

    Red is the first colour to blow out in chroma so to me that little sign makes the Canon the clear winner.

    • There are so many variables, so I would encourage you guys to download and play around with them yourselves.

  • Ben Howling on 05.23.13 @ 8:24PM

    Anyone tried shooting some real world shooting situations in RAW on the Mark III yet? I’d love to see some basic set up shots (tracking/dolly/steadicam) in a drama situation, as well as some documentary style run-n-gun shots. Anyone got links to some tests in this mould?

      • That group seems a little confused. A lot of it isn’t RAW.

        • Lou Goetzmann on 05.24.13 @ 5:19AM

          check out this channel:

          i manage it myself so there are onlly ML Raw videos ;-)

          • guys, this is awesome. i really appreciate it.

          • The quality of some of the shots blows me away. Cinematic with a capital C. the dynamic range, the detail, on a sub $3K camera, with cards that are very affordable, I think I’m sold on adding a 5D, regardless of the BMPC. There’s no reason you can’t have both (apart from the most obvious case of affording them).

            Once ML get this stuff locked off and reliable, it’s gonna be the standard for small projects and online content, IMO. And with that market already being such a growing force, it’s only going to mean better work and higher standards, which will open the door to more work (as opposed to traditional broadcast media).

  • Austin Burke on 05.23.13 @ 11:15PM

    Man I want this to work well on the mk 2, then I might switch back to canon. Just switched to the d5200 and I like it but man I miss magic lantern. Now to just see if they can get the mk 2 stable at a good pixle ratio and maybe it will work well with vaf filter?

  • This looks great. There is a ton of chatter about it in our offices and on the sets I’ve been on the last 10 days.
    We’re all waiting for ML to ‘finalise’ it as they tend to do, and then a lot of people will probably pick up 5D3s.
    For me it will be between this and whether Blackmagic can get that 4K cam out on time. If they do, I’ll go that way to upgrade/add on from my D800.
    On the RED thing, I have to say that apart from music videos and shorts the amount of footage we see for high-end post has declined to a trickle. We were also discussing that the amount of MX we see now is zero, from maybe 50% of ALL footage only 18 months ago. That Dragon sensor needs to happen yesterday.

  • For me, it’s not about RAW or sharpness etc, it’s the LOOK.
    The REDs and BMCC look so much better in motion that the MKIII. Very filmlike.

    People like Andrew at EOSHD are going bananas over this ML hack, and I can’t really see why.
    Its the same look as the MKIII, just in RAW.

    • One. Maybe because of the superior low light capability if the hack is stable. Two. Maybe because the hack will be free and the camera is only $3,000 compared to $15,000+. Three. A 5DIII is so small and easy to use and walk around with compared to a Red. Four. If the postproduction workflow is simplified (maybe even a “drag and drop” app to convert to ProRes HQ), and prices for CF cards, SSDs and other external storage media solutions get cheaper, this mushy camera will be a beast.

      • I’m curious when a hack for the GH3 will come along. Will the same people that are excited about this hack improvement in MkIII be just as excited if the GH3 hacked turns out to have a better picture than this?

        • RAW Video hack is almost certainly not going to happy on the GH3, so don’t hold your breath on anything more than a higher high bitrate hack.

        • Also, I don’t think nearly as many people would be as excited about a RAW GH3 as a Mark III. It’s not quite the holy grail that the Mark III is, because it’s MFT and not full frame, big difference. People are thrilled with the Mark III being RAW because it’s the first full frame RAW video camera anywhere near even $30,000….if Canon released a full frame RAW camera like this it would probably be over $30,000. There’s no crop on lenses and it has SUPERB low light.

          The Gh3 recording RAW video wouldn’t be able to record 1080p or probably even near it at 24fps continiously since it uses SD cards and not CF cards and is said to have a small buffer. If it could use CF cards and had a decent sized buffer it would be a compelling RAW camera.

  • This post is only to show us how good of an image we’re actually getting out of the hacked MkIII. It’s not really to say one is better then the other. Of course the RED camera is a better camera for certain things. I’d love to have one if I could afford it. However, with this break through, I think my MkIII just got a longer life. Being able to do this with a $3k camera is amazing, as is the BMCC. I just hope they figure out a way to either get a CF->SSD adaptor going or someone comes out with some cheaper CF 1000x cards. I know Komputerbay are a lot cheaper then the competition already, but it still is more expensive then SSD. Since there’s SD-SSD adaptors out there, I would assume the same thing could be done easily with CF. Nerds, get on it!!! Make some $$$! If I had the knowhow, I kickstart it and make a killing.

  • funny thing is i downloaded RAW Files of RED Epic and converted Jpeg it looks waaaay better than exampe that you posted..No Trespassing sign clearly visible unlike the Canon.Also R3D. files always needs sharpening

  • it all looks great… “looks”, but there is a big difference in the way the movement is rendered… it never accomplishes to deliver a full cinematic movement rolloff like an alexa, epic or BM … it looks super hype, full of dynamic range, but the movement still fills electric (anti-organic) not filmic at all.

  • looks like andrew reid has had a humbling experience after his initial overhype of ml raw. i also think its the bomb, but his over the top statements given as fact were not all accurate and rightfully pointed outas such on several occasions to him by dps who know better. He then lost more credibility by taking advantage of the situation to slag off bloom and laforet, even though much of it was valid, it was not very professional nor helpful. Reid is eloquent at projecting the technicalities, which is great, but then you later realise its often all a bit delusioned by ego and innacuracies.

    ML Raw potentially solves my camera dilema in terms of budget and stills requirement.. i was all set for bmcc but it meant i was gonna have to stick with my 550 for stills, which i was a bit sour about. i Just really really hope they get the motion sorted. Appears that the verticle lines and 4gb limit have been solved!

  • So Luke or Joe, I’m wondering if there has been official/unofficial comments anywhere from the Canon folks? Seems like damned is they do, damned if they don’t. If they say nothing they clearly will sell more units of the Mark3, until something in the price range perceived to be better will come along…….


    • I doubt they will say anything. They don’t benefit from talking about it, because they are going to upset people one way or another.

  • Joseph Birch on 05.26.13 @ 11:35AM

    Played around with the canon raw .dng files in camera raw and I must say – I am impressed. Love level of detail and the latitude.

  • I’d love to see some tests with fleshtones and movement. It seems like most the tests I see, there some inexplicable obsession with shooting buildings, trees, rocks, all cut to cool music. I test cameras constantly and we always use a human in front of the lens because that’s what we get paid to shoot most of the time. It’s what audiences focus on more than any other pictorial element. There’s a reason, Alexa at 1080 or 2k is the number one camera in episodic and features – it’s rendering of fleshtones are hands down better than all the competition by a significant margin. How human faces look influences the audience far more than pixel count. It’s one of the reasons I sold my Epic. I was never satisfied with its renderings of skin. It’s why I prefer the image of a 5D Mk2 over a GH2/3 (and perhaps why I’ve yet to have a client request them), despite the higher rez specs.

    • Most people probably have not heard of the GH. Here’s a comparison of GH3 with 5D3. If the average person saw this video they may opt for the GH3 if they wanted some video work. It’s from 4 months ago so it’s pre ML RAW. But a hack for GH3 hasn’t come along yet either. And that would change things too. Watch the video twice, first with captions off to see if you can guess which camera shot each segment, second with captions on to find which camera shot which segments. Make sure you click 1080p in the quality button—if you do want to watch it:

    • This may sound odd but the Arri Alexa didn’t fare well in the Zacuto Revenge of the Great Camera Shootout. I know that it sounds odd because going in starting to watch it I expected the Alexa and the Red to blow away all the other cameras. It was odd to me to see the Alexa not stand out for most of the comparisons. It did have a few good moments. But it was a big disappointment for me. From what I had seen of the Alexa before I thought it was great. But then seeing it next to other cameras burst my bubble, hard. Both the Red and Alexa were not too much different than a few of the others, like the Sony FS100. The actual best moment, to me, in all the comparisons came from the Red though with the burgundy color paint in a corner of the room. It looked gorgeous. Maybe how both the Red and Alexa were handled in post, or rather, not handled, made them not stand out. I don’t know. I don’t want to be critical of the two teams that did the work with those cameras.

      The two biggest surprises in that documentary were how well the GH2 Hacked did, and how poorly the 7D did—I mean awful really.

      If you haven’t seen the docu and want to it’s in 3 part in vimeo. It’s VERY MUCH worth a watch.

      Here’s part 1:

  • I think ML is letting the 5DIII produce a fantastic image. The sharpness and clarity is not something most people are used to, coming from a Canon DSLR. I think the 5DIII honestly has some real potential now, and I’ll be looking into using it when the occasion calls.

    However, I kind of feel like there’s still an awkward DSLR type motion / movement going on that makes it easy to identify as DSLR footage. This is the only downside for me… the cinematic or filmic 24p motion that other cameras like RED and Alexa have, just isn’t replicated in the 5D series very well.

    The hack is amazing though, and will be following the development very closely.

  • When are we gonna have RAW for Nikon D800 and the GH3?

  • Refreshing dialogue between Aviv and Luke! Loved the wine palette analogy. Also good advice is to buy the equipment when you need it. With all these cool toys and accessories it’s hard to put a brake on the shopping cart. Sound advice is to only buy when you need it, and are able to justify the expense. The business side of our industry is all too often underestimated…

  • Too late on the post, but surprised to read comments that people are complaining “How did you compare 5D to the Epic”. Please, when most projection is 2K (yes there are 150+ countries other than the US) it matters! And no matter what the naysayers say, it is a fantastic option to have. Mindboggling in fact.