I recently purchased an 5d mk iii and installed magic lantern on it. The 14bit raw video is AMAZING. Also, this particular camera can now record at, I think the limit is 3.5k vs 1080p from the factory.
This shows a pure comparison (it should link to the right time-mark). First shows some before/after full shots then does a split screen comparison in various situations:
https://youtu.be/WTr2L3uCYts?t=80
5D3 comparison:
It would also be nice to see it without Youtube's compression, if anyone has a better link.
(Your second video seems much more in line with what I'd expect, but it's still hard to see clear differences with Youtube's compression.)
Many choices must inevitably be made along the pipeline from estimated-electron-count-per-sensor-pixel -> image on a display. Default choices are not inherently more correct (or truer to the scene or whatever) than deliberate choices.
The fairest comparison is probably to find someone highly skilled at image editing and get them to try to make the best output images they can from both inputs. You’ll get to see places where the standard processing and compression actually lost data.
Often the processed-in-camera version still has enough data that a more-or-less comparable output image can be produced, but if you start peeping on pixels you might notice extra noise, banding, blur, ringing, ...
The problem with the comparison videos in this thread is that the person/software deciding what to do with the “raw” data has made a bunch of choices to allocate more contrast to shadow and highlight areas, etc., while the in-camera software made a different choice... and nobody tried to reconcile the two versions afterward.
Comparing the amount of usable data available
"I disagree when I shoot raw and h.264 my camera has the ability to shoot both at once and spit both files out, and these are exactly how it looks no color grading fresh out the camera. "
Here's my stupid question of the day: given that this color-grading of RAW streams makes _SUCH_ a massive difference... why don't companies like Conan and Sony just ship it with the product? Why do their cameras fall short of having something like Magic Lantern? In the end it gives the customer what they want... surely Sony/Canon are capable of producing something like this, so why not produce it?
They can charge this much, because they're competing with high end cinema cameras that are used to shoot blockbluster films, such as the Arri Alexa LF ($100,000) or Red Monstro 8k VV ($79,500).
Some video camera manufacturers have started offering raw at lower price points. The Blackmagic Camera Pocket Cinema 4k shoots 4k raw 60fps for only $1,300. Sensor size micro 4/3s, which is a lot smaller than a full frame sensor that the 5dmk3 has.
https://uk.reuters.com/article/tech-eu-cameras-trade-dc/eu-t...
It's not just the firmware that makes the difference between a DSLR and a pro video one, but ergonomics and hardware connectivity. Videographers tend to be more demanding on technical support issues too IME.
However, you can get very good cinema cameras for not very much money from Blackmagic that shoot raw video. I have both their 1080p pocket and 4k pocket cinema camera, they are awesome.
Magic lantern is an amazing project too, but reverse engineering the firmware is a lot of effort and never quite bug free
But it’s mostly a business decision I guess.
- e.g. more cost effective to use the same HW on multiple devices but limiting their capabilities through SW by segment => total cost covered by many sold med-end devices AND few high-end devices?
- and/or maybe what "magic lantern" allows to do wasn't certified to work on 100% of the HW.
- and/or "magic lantern" did some own SW development, which would mean "additional costs" if done by a company (and here is where open source hits one of its targets?). Meaning that maybe at a certain point Canon said "ok, we've spent all our budget for this price class of camera vs. the expected earnings, so that's it".
Cheers
I don't know if these are presets or something done in post-processing, but it almost looks like the video equivalent of filter plugins in some cases.
I'm by no means an expert but color grading video is often a detailed process with a lot of room for finding your own creative "vision". Think of the split toned, green/orange aesthetic that was popular in movies for a while. Doing the color grading can have a major effect (good, bad, or just different) on the mood of a video.
In these examples, I was definitely impressed with some aspects of how curves were adjusted to bring up shadows and make details more visible. But at the extreme end, it reminded me almost of the overkill seen in early HDR photography.
Some tweaking can make things look more realistic and bring out detail but it is very easy to go too far and end up with something that looks like an Instagram filter.
That's mainly why I was wondering if the video examples were in-cam presets when running Magic Lantern, or if it was something the videographer did in post with the extra data captured with the higher bit depth enabled.
It should be theoretically possible to do anything to each video frame that you can do to a still photo, including moderate/tasteful color & contrast adjustments.
Some photo printers love to allocate almost all of their available contrast to large-scale shapes, producing essentially silhouettes. Others like to allocate almost all contrast to local fine detail, leaving the image looking like a gray blur from afar but detailed and crisply textured from close up. Some photographers like their images to be a festival of competing intense colors, while others make nearly monochromatic images in one color or another, or stick to a pastel palette, or make mostly neutral images with a few intense exceptions. Etc.
When someone says a photo or video was printed badly, what they usually mean is that either (a) the printer had shallow aesthetic judgment or boring artistic goals, and/or (b) the printer lacked the skill to effect their artistic vision.
is this false for the factory default?