By James DeRuvo (doddleNEWS)
Development of the Magic Lantern Raw video hack has gotten so stable and sharp that according to further testing being done by ML users, it now not only surpasses the official H.264 codec, but completely blows it away in low light and at high ISO. And looking at the image above, can you argue why anyone wouldn’t want it?
“The ISO test was amazing,” Tolman tells Planet 5D via email. LI didn’t know what to expect at first, I really just wanted to do a video that I hadn’t seen yet within the ML RAW arena.”
First, let’s look at what Dustin Tolman, the ML user who conducted the test, used:
Canon 5dmk3 @ F8
Canon 70-200 2.8L IS II @ 200mm
Magic Lantern Raw set to 14 bit
What Tolman did was shoot a candle in low light, starting at ISO 100 and then doubling up the ISO all the way through the spectrum up to 6400. Almost immediately, you can see that the image begins to degrade with H.264 as more and more noise gets invited to the party. There’s also a more pronounced halo and the colors begin to fade as the speed of the sensor increases. The H.264 image also gets far more soft as it goes up in ISO, while the RAW image stays good and sharp throughout most of the test. Additionally, there’s very little noise and the colors remain bold, especially the surrounding blacks.
Tolman did notice an interesting phenomenon as the test progressed, however, and that dropped frames with the camera once battery life dropped below 30%. He also noticed the color magenta became more pronounced as the ISO range increased, but he was able to color correct for it rather easily using Adobe Camera RAW.
And Tolman references other low light tests, like user “Squig” in the Planet 5D forums who stated that increasing ISO in increments of 100 lends itself better to the Magic Lantern Raw setup. “Keep in mind that the “clean” ISOs are the opposite of what they were with H.264,” writes Squiq. I’m seeing much less noise with multiples of 100.” This is something that Tolman finds rather interesting.
But while Squiq’s aggregate image post of noise throughout his ISO test is pretty comprehensive and revealing, I didn’t find that much noise coming from the ISO test that Tolman put out. So I would agree with him that it is interesting how noise is affected when you bump the ISO up while using the Magic Lantern Raw hack.
Regardless, the images are absolutely stunning and will only get better as ML Raw matures. So the real question is, what does Canon think of all that? And on that note, Tolman is philosophical …. “Imagine if Canon were to let developers and testers create products the same way ML is,” he writes. “Imagine if the world was run this way…”