Canon 5D2 – RAW vs. H264 In Real-World Production

When I was in middle school, I saved my pennies until I had enough money to buy a Gameboy. I played Tetris until I could close my eyes and see falling blocks. All the game graphics were gray/green monochrome, because that was the limitation of the technology. About 15 years later, my wife got a Gameboy Advance SP. It was billed as being backwards-compatible with original Gameboy games, so I dug out my old Tetris cartridge, and plugged it in. To my amazement, the falling blocks were in color! The programmers of that game put color data in the game code, even though they thought nobody would be able to see it.

The way I felt when I read that the Magic Lantern team had discovered a way to record RAW video from a Canon 5D Mark II reminded me of the way I felt when I saw those colored Tetris blocks. How is it possible that a five-year-old DSLR can have state-of-the-art functionality? It leaves me shaking my head in awed disbelief.

I didn’t spend a lot of time in disbelief, however. I immediately downloaded the Magic Lantern RAW module, shot some test footage, and decided to use it for a real project.

So, I did. Here’s what I learned.

The Concept

The best way to assess a new tool is to use it for its intended purpose. Shooting scenic beauty shots is fine, but I knew that I’d learn a lot more about what this 5D2 RAW can and can’t do by using it for a short but full-featured production. I had been planning to do a small promo video for a friend’s tennis program anyway, and the project was a perfect fit.

In order to do a meaningful evaluation, and because I wasn’t confident in the technology, I shot enough footage in both RAW and conventional H264 formats to finish the spot, even if the RAW files didn’t work. As it turned out, only a couple of the RAW files malfunctioned, so I was able to create two complete versions of the spot. This allows for some interesting comparison.


RAW files gobble up enormous amounts of storage (about 4GB per minute, depending on resolution), so I needed a short project with simple shots. A web video promoting kids’ tennis was perfect. The kids were only available for about an hour, and I couldn’t afford to waste disk space, so I decided to shoot everything off a tripod, to maximize usable footage, and to avoid the complexity of moving camera shots. Also, I was working as a crew of one, so logistically I had to keep things minimal.

On the 5D Mk II, Magic Lantern RAW has a maximum resolution of 1880×720. It’s an interesting, super-widescreen aspect ratio, but it caused some problems in my initial tests, so I decided to shoot at the standard 720p resolution of 1280×720 instead. Magic Lantern decreases resolution by cropping the image, so the fairly severe crop factor from 1080p to 720p meant that I would only see the very center of the image. I wanted to use a wide-angle lens to give a sense of immediacy and to limit rolling shutter “jell-o,” so I decided to shoot all the action footage with the widest lens I have, a Sigma 12-24mm, and the on-camera segments by the coach with my Canon 24-70mm.

The Sigma 12-24 is not a particularly fast lens (maximum aperture f/4), and it does not accept filters, so I knew I’d be shooting at a pretty high aperture, but for sports, a more generous depth of field isn’t necessarily a bad thing.

tascam_dr-100mkII-5409Magic Lantern RAW does not record audio, so I knew that I would have to use true dual-system sound. I normally use my Tascam DR-100 Mk II for DSLR audio anyway, but I’m used to being able to record a backup track on the camera, which makes sound sync a breeze. In this case, I would have to sync everything manually, so I would have to remember to slate each shot of on-camera audio.

I also decided not to worry about recording the ambient “nat” sound of the tennis activities. I planned on using voiceover and music throughout, and I suspected that I might be able to use some of the camera audio from the H264 files for background flavor.

With these preliminary decisions made (and a script and shot-list in hand), I was ready for the shoot.


For the most part, my prepro decisions worked well. I was hoping for a nice blue sky, but the day was overcast. This wasn’t ideal, but it diffuse the sunlight enough to minimize dark shadows, while still allowing enough directionality of light to provide for decent contrast.

I did make three mistakes:

1) I mentioned the substantial depth of field associated with shooting a 12-24mm lens at a high aperture. Well, I did take the precaution of running the 5D2’s sensor-cleaning function, but I didn’t bother to check the result on a monitor. Consequently, when I got the footage home, I found the image of a thin hair in the lower right-hand corner of all my H264 footage, along with a couple of small spots in the center of the image. Fortunately, the hair was outside the crop zone of the RAW files (there’s an upside to everything … even severe image cropping!), and there was enough motion in the shots to disguise the other sensor spots.

2) I forgot to slate two out of the three RAW shots that needed to be synced with audio. This was a major hassle, and was further complicated by the fact that the RAW files would glitch occasionally, throwing off sync. Mercifully, I only had two lines of on-camera dialog to sync, and only a few takes of each to sift through.


3) Magic Lantern RAW recording is triggered differently than conventional recording. A couple of times, I instinctively hit the record button, while RAW recording was already engaged. This corrupted the RAW files. I also had some glitches and Magic-Lantern-flavored pink noise in a couple of my H264 files, which surprised me a bit.

Overall, I was quite pleased, as only three of the 20+ RAW files I shot failed to work in post.


I won’t lie to you: post was a bear. RAW files require a three-stage conversion (RAW to DNG image sequence, DNG to TIF image sequence, TIF to ProRes) and a two-stage color grade (preliminary RAW processing, and then normal grading within editing software). This results in multiple versions of each (huge) file, and is very time-consuming. It’s also a process replete with “gotchas.”

The RAW processing stage tripped me up a bit. I’ve processed photos in RAW for years, but doing the same process for video was quite different. I used Photoshop to batch-process the DNG files, and I tried to do sort of a preliminary grade with them, while maximizing detail. This turned out to be a rather unhelpful approach. I should have either gotten the files as close as possible to the final result I was looking for, or I should have made them as clean as possible. As it was, my artificially warm color balance (I wanted a warm, summery look to the footage) caused me headaches when I started using Colorista II on the ProRes files.

DNG process

The dynamic range of the RAW files is great, but by maximizing detail in the shadows and highlights, I allowed the midtones – the heart of the image – to get muddy. In retrospect, I would suggest grading for what’s important, rather than what’s most technically impressive.

I must say, I was a little disappointed in the color rendition of the RAW files versus the H264 footage. Although the RAW files had more of the “RED look,” the punchy colors of the 5D’s native format really look good by comparison, both before and after grading.

Here’s a still frame of the RAW footage, with only preliminary processing. class_raw_ungraded

And a still frame of a similar shot from the H264 footage, ungraded. The blue shirt, the green court, and the green bushes all look really nice.class_h264_ungraded

Here’s the RAW file, graded with Colorista II. I’ve brought up the saturation a bit, and adjusted the total balance to give the image some more contrast.

Here’s the H264 file, graded with Colorista II. I’ve warmed up the shot a bit, and brought up the midtones considerably. It now looks very similar to the RAW shot, but it still seems to have punchier colors somehow.


There’s also considerably more moiré in the RAW files than the H264 files. Apparently, this isn’t an issue on the 5D3 and 6D, but I sure saw it on the 5D2. Look at the coach’s shirt, in the RAW image …


… compared to the H264 image.


The RAW image quality is very impressive, and the flexibility of RAW post-processing is awe-inducing, but it’s hard to imagine using DSLR RAW for more complex projects until the files can be integrated more seamlessly into the editing software. RED’s “R3D” files can be imported directly into Premiere, and adjusted and re-adjusted as necessary, without hogging harddrive space and wasting time. We need something similar for these Canon RAW files.

The Magic Lantern interface is clunky and a little buggy, but it’s a work in progress, so that doesn’t bother me a bit. Is 5D RAW ready for primetime? Probably not yet … But the fact that it exists at all is incredible, and it is surely a harbinger of great things to come. More to the point, these developers are giving the fruits of their labor to the rest of us for absolutely nothing in return, so they deserve our heartfelt gratitude. On behalf of all the 5D users, I say “Thank you!” to Trammel Hudson (the original Magic Lantern creator) and all the brilliant men and women who have continued his work.

  27 comments for “Canon 5D2 – RAW vs. H264 In Real-World Production

  1. Jesus
    May 21, 2013 at 6:03 am

    Very stupid article. Don´t waste time doing bad videos like yours. Go to the cine.

    • May 22, 2013 at 12:03 am

      Thank you Jesus, now go back to turning water into wine.

    • Jesus is an idiot
      May 22, 2013 at 4:57 pm

      Very stupid reply, Jesus. Don’t waste time typing one sentence condemning someones informative article. Go back under the bridge.

      • Alexander
        May 22, 2013 at 5:49 pm

        Wow … The very first hostile interchange in the comments section! I feel like my blog has passed an important milestone now. Thanks for sticking up for me, folks.

    • Doug
      May 27, 2013 at 6:37 pm

      Somebody completely missed the point of the post.

  2. May 21, 2013 at 6:42 am

    Great feature … thanks for taking the time to shoot and share.

    Do you think the 5D2 will ever have 1920 x 1080 pixel raw video resolution ?

    If you work from a larger image, does this reduce moire and aliasing ?

    One final point … what do you usually do to keep dust spots to a minimum when shooting video ? Blow brush and auto sensor clean isn’t always enough … 🙁

    Fingers crossed for the 5D2 ! 😀 And wish i still had my Gameboy 😉

    • Alexander
      May 21, 2013 at 11:23 am

      Thanks for the kind words. I don’t know if the 5D2 will be able to do full-frame RAW. That might be asking a bit too much!

      The moiré and aliasing seems to be a byproduct of the additional sharpness, but it only affects the 5D2 (so I hear!).

      Regrettably, I suffer from sensor spots too!

  3. Jag
    May 21, 2013 at 7:24 am

    Is magic lantern able to get RAW footage from 5D mark II? I see lots of article on mark III.

    • Alexander
      May 21, 2013 at 11:20 am

      Yes … This was shot on the 5DII.

  4. May 21, 2013 at 10:25 am

    Thanks for writing this up. You’ve given me some perspective on how usable this will be for my future projects.

  5. May 22, 2013 at 9:38 pm

    how can I that in 5d2 please could you post the link of ml^thanks

  6. David Keller
    May 22, 2013 at 10:03 pm

    good write up alex.. kind of makes me regret selling my mkIII, as that seems to have the most functionality with these new raw developments. a few months ago, who would’ve guessed all this would happen? i still have an old mkii, so maybe i’ll try it with that.

  7. May 24, 2013 at 4:58 am

    Hi Alexander. Thanks for the effort you’ve put into making this article. I have one advice for you that will make your life easier: Open the DNG files in After Effects as an image sequence. Experience the awesomeness. Thank me later. 🙂

    • Alexander
      May 24, 2013 at 10:43 am

      Saad, thanks for the kind words. After Effects was the first method I tried to convert the DNG files. Everything was set to 23.976 fps, but the files came out jerky and glitchy. I’m sure I did something wrong, but I’ve only recently switched to Adobe CS6 after over a decade with Final Cut Studio, and I’m not up to speed with AE yet. Any tips??

      • Rob
        July 30, 2014 at 8:46 pm

        Ok the real deal is to shoot raw with dual iso (MLV). Those pics are amazing. The workflow is simple but time consuming.
        I use freeware for RAW to DNG (window interface). I use another freeware for the DNG dual iso processing (its a DNG to DNG, very slow). In After Effects CS6 “Import” then colour correct first pic, and the rest is batch corrected into a sequence. Ok they call it a composition in AE. Export, lossless, select H264 and you can generate an MP4 for each clip. Final sequence edit in Premiere Pro CS6. Its the easiest I found. You have lots of tools in both AE per clip, and overall in PP.

        However RAW is just for editing. The final product is normally an H264 (mov or mp4). So in the end you achieve the same as can be done without raw. That’s the interesting bit for me. Since RAW has such a work flow, and syncing audio is not the simplest, I am wondering about H264 as an alternative. I generally notice the biggest difference in low light, but a well lit 5 point scene can have very nice colour depth on an H264, so that’s the dilemma. RAW work flow or well lit H264. Hmmm … I am thinking to work with the lighting to improve the picture quality. Also colour grading with a rather flat H264 is another good option. AE can be quite cinematic. So your article to me is very interesting, as its brings to bare the idea of when to use RAW and when to use H264. Howabout H264 for sound, and RAW for non-sound….

  8. May 27, 2013 at 8:37 pm

    Interesting article, thanks for taking the time to do the test and post your experience.

    What it suggests overall is that far from the revolution people are talking about, the ML hack is a long way from being usable in any professional situation.

    Any gains in picture quality, which seem to be debatable, are substantially outweighed by a) the instability of the firmware, b) the eye-wateringly inefficient post workflow, c) the storage space required for the various versions of the files, and d) the lack of audio – it was noticeable by its absence in the tennis school promos.

    The usual defence is that these are early days and given time, the ML team will stabilise the hack. That may be so, but it only solves one of the above.

    It’s hard to see this ever becoming a viable production device. Using it experimentally for a friend’s web spot would be about the limit – it doesn’t look like money changed hands and if it was a complete disaster it could be re-shot without much hassle. But on a paid, one-chance-only gig, no frikkin way!

    Even on a low budget feature, why would you risk it? It’d probably turn out to be more cost-effective overall to rent in a Red!

    What it probably will do, however, is create enough of a buzz around the whole ‘raw’ concept that will force manufacturers to sit up and take notice. As an FS700 owner, I am looking at dropping several grand on new hardware if I want to shoot raw. Magic Lantern may just take that price tag down a few notches.

    • Alexander
      May 27, 2013 at 9:40 pm

      Neil, I agree with your assessment. The RAW advantages are (reportedly) more pronounced with the MkIII and 6D, but since the footage, when graded, winds up looking much the same as H264 footage, it’s hard to justify the additional hassle and storage requirements. If anything, this is a vindication of just how good the 5D’s H264 image capture is. Not perfect, not RED-quality, but very, very good.

  9. Chris
    June 3, 2013 at 10:09 pm

    What light did you use for this video the one in the picture of you filming and how did you power it while shooting in the field?

  10. PiDstr
    July 13, 2013 at 10:00 am

    Interesting read there Alexander, nice to see a ‘real world’ use of the Raw output, and I see what you mean by the “Red look” in the ungraded Raw sample image at the top.

    All this hype about Raw our of a DSLR, and that work flow is a pain, and you can’t get 1920×1080 out of 5D2 ??

    I think I’d prefer the wide dynamic range of the sensor put in to a more efficient CoDec, or a video file that doesn’t need to stick to REC709.

  11. dude
    July 26, 2013 at 12:07 pm

    in after effects, right click on the imported dng footage and go to ‘interpret as’ , when that is different to your composition frame rate, that is a common cause of jittery playback

    • Alexander
      July 26, 2013 at 4:25 pm

      Great tip, thank you!

  12. Rahul
    September 12, 2013 at 8:01 pm

    The article and shoot was very informative.

  13. January 24, 2015 at 11:33 am

    That Genaray seems to have quite a bit of punch. How is it working for you today, Alexander?

    • Alexander
      January 24, 2015 at 11:35 am

      The Genaray lights are great! I’ve actually bought a couple of their newer ones, with adjustable color balance. I’ll do a review soon.

Leave a Reply

Your email address will not be published. Required fields are marked *