Canon 5D Mark III Magic Lantern - Looking Back

[Zack Birlew]

Hey, everybody, so I recently took a look back at some of the Canon 5D Magic Lantern RAW video developments and saw some more recent examples that I didn't see before with newer 3K/3.5K/4K recording and I couldn't help but be absolutely impressed. Even looking at my RED One MX/Blackmagic Pocket 4K/Sony A7S 4K external footage as I finish up my first feature by comparison, I can tell that the 5D Mark III RAW video is a step up in many instances and even does better than more recent 4K RAW camera examples. What was it that made the 5D Mark III's RAW video so good and filmic and why can't we get that level in newer cameras even with RAW?

 

[Phil Holland]

I think you need to test and look closer at newer RAW capable cameras IMO. 5D MK III is a good camera, but a lot has happened in 11 years.

 

[Zack Birlew]

Oh yeah, I know there are definitely better options, I thought about picking one up but for the price 5D Mark III's are going for, I might as well save up for a Blackmagic Pocket 6K. Like I said though, there are some instances where it's stunning how good it does look for such an old camera. The major problem I see in some tests and examples is that there are some cheaper lenses being used and, if not that, there are particular full frame lenses being used that aren't the best when cropped in when using some MLRAW video crop scenarios. Like any camera, if used right and with the right combination of lenses, someone can get fantastic results.

Looking at certain cameras out there like the Fuji GFX series and, of course, the usual high megapixel mirrorless cameras, I wonder if they're hiding much more capable RAW video cameras than what's being delivered. Much like RED's movement of bringing 4K and making it mainstream, the Magic Lantern movement revealed a good deal of possibilities even in older cameras that could have led to a much different state in the industry had things played out differently.

 

[Clayton Burkhart]

The RAW files coming from that camera and Magic Lantern are among the most accurate I have ever seen when it comes to color. Rarely do images from most cameras not need some form of grading. However, something about that camera and hack just works when it comes to this. I frequently have left the files ungraded from that combo. Where it falls down a little bit is in dynamic range (a little under 11 stops) and of course image size in video. Sometimes it can be a little "grainy" (noisy) as well, but that just comes with the image size.

 

[Mike P.]

It’s uncompressed raw and has even over/under DR distribution. Even at only ~12stops, it actually has a comparable amount of highlight stops to most RED sensors at equivalently cleanliness, and it rolls-off in a more pleasing (easily achievable) way because there's usable/recoverable colour in the extremes (which makes it "look" more than it is compared to r3d footage which is typically the opposite).

I'd be curious to see 10/12/14bit uncompressed raw vs 16bit compressed raw comparisons... something tells me even the 10bpp would look better. There are some DoPs that making the argument that uncompressed RGB is actually better than any flavour of compressed raw, and the only reason we don't see more of uncompressed RGB in-camera is because of how much processing it takes to convert (not even the data rate) *cough*Yedlin*cough*

Anyway, aside from all the positives, 5DML footage is still pretty soft looking 1080p and its rolling shutter isn’t the best…. But then again, it's VV ~2k uncompressed raw for like ~$700 in an out-of-the-box shootable 2lbs camera that can be powered all day by a single 45Wh battery (and take some pretty stellar photographs while not rolling).

 

[Leonardo Newman]

It’s uncompressed raw and has even over/under DR distribution. Even at only ~12stops, it actually has a comparable amount of highlight stops to most RED sensors at equivalently cleanliness, and it rolls-off in a more pleasing (easily achievable) way because there's usable/recoverable colour in the extremes (which makes it "look" more than it is compared to r3d footage which is typically the opposite).

I'd be curious to see 10/12/14bit uncompressed raw vs 16bit compressed raw comparisons... something tells me even the 10bpp would look better. There are some DoPs that making the argument that uncompressed RGB is actually better than any flavour of compressed raw, and the only reason we don't see more of uncompressed RGB in-camera is because of how much processing it takes to convert (not even the data rate) *cough*Yedlin*cough*

Anyway, aside from all the positives, 5DML footage is still pretty soft looking 1080p and its rolling shutter isn’t the best…. But then again, it's VV ~2k uncompressed raw for like ~$700 in an out-of-the-box shootable 2lbs camera that can be powered all day by a single 45Wh battery (and take some pretty stellar photographs while not rolling).

did you really just say a 5D mark iii shooting magic lantern raw, has comparable DR to a Red, has better or more pleasing highlight roll off and is just as clean? This is laughable. I own a 5d3 and unless magic lantern shoot better raw than that camera does in photo mode (which would be impossible) then you just might be spewing a lot of hyperbole. But I won’t just leave this comment, I’ll actually post some frames for you in a bit.

 

[Mike P.]

Highlight stops (above mid-grey) with HTP is comparable, yeah. Only like a ~1 stops off in the high-end (+4 vs +5), and it typically has better/stronger colour at its +4 before it clips, too. That’s what helps it look as good as it does in spite of having less overall DR.