Messages

Dig the video (two of my favorite cities). What was your workflow?

Cheers.

Most shots are 1280 1:1.39 ratio 23.976fps and shutter as close to 180º as possible (174º usually) + some cheap ND filters. Extracted with Batchelor (to save time) and imported to After Effects with ACR. No further colour correction.

The wider shots are 1728 at 11.988fps and 90º shutter, twixtored to 50%. I discarded footage with obvious Twixtor artifacts (most of it).

PS. Which cities? Cairo should be there as well, but I stayed at the airport because of the current situation .

Here's some stubborn testing. I think the results are quite useful, with some issues here and there (moiré, obvious twixtoring...).

I haven't tried upscaling it to 1920 yet (can't keep the res on Vimeo anyway).

I'm on a trip testing the Raw functionality in real situations.

The only thing I miss is the option (that other builds have) of turning GD off automatically when you start recording. Other than that, I'm quite optimistic about the results I'm getting at 1280 2.39 23.976fps and 1728 2.39 11.998fps (crossing fingers for Twixtor).

Not really.  The less frames to interpolate the better.

That's why. When twixtoring from 12 to 24 fps you are interpolating 50% of frames, because all the original ones match a new temporal position. If you twixtor 18 to 24, only 1 or 2 6 original frames will match their new temporal position, and 22 12 won't, so you will have 22 18 interpolated frames.

*Update: I've double checked (18 to 24), and it appears that 6 of 24 (25%) frames will be original. Still not as good as 50% (12 to 24), but much better than other combinations

People using twixtor - is interpolating 18fps to 24 a big deal? Will the result look acceptable?

12 to 24 should be better. I wouldn't use it at all if the image has much movement.

I tested some methods. To do so, I basically downscaled an image to 50%, upscaled it again with different plugins, and compared the results to the original image. Here's a detail of the comparison.

https://dl.dropboxusercontent.com/u/15481036/upscale.jpg

And here you can download the tiffs: https://dl.dropboxusercontent.com/u/15481036/images%20to%20compare.zip

Blow Up 3 is the only one that does something to aliasing. It does so by "destroying" the image and "creating a new similar one", in my opinion. I don't like the plastic look it gives, and I feel like it is inventing too much information and applying sharpness where there shouldn't be any. However, I wouldn't mind using it for some shots with bad aliasing and then mask out the non-aliased area. To my eye, the winner is S-Spline from Photo Zoom, the best results in both sharp and blurry areas, while respecting the look of the image.

I wrote some code to compare the images in a measurable way. Basically, the code tells you either how many pixels in the original image match the ones in the processed ones, or how different the different pixels are. The results were really disappointing. Basically it looks like only the Nearest Neighbour image shares around 30% of the information with the original image. The rest of them are mostly "invented". This might also be because there seems to be a small offset (less than a pixel?) between the original and the processed images.

Does anyone now of a method to make this comparison useful?

Guys, is there a guide for noise reduction? I am getting a lot of noise by default, even at ISO100

A certain amount of noise in dark areas is unavoidable. Maybe you are keeping the shadows too bright in post. ACR does a nice job reducing noise, and also ETTR should help in many cases. If you really wanted bright shadows, you could try filming HDR, but that's a whole extra workflow on top of RAW.

so when u go into 5x mode and u change the raw resolution, the bar that pops up showing u what your recording  area is. Is it showing u the real form factor. I notice its there when your in normal mode, does it adjust for if your in normal mode or 5x mode.
or is it guessing?

That area is not useful in cropped mode. The recorded image is actually larger than what you see in the whole live view. You can compose the frame by selecting ML or Hacked preview in the Raw menu.

Does someone a have a recent and complete build of Tragic?

EDIT: solved: http://www.mediafire.com/download/96zgrdjaubamjuh/ML_Fileman_Raw_Tragic.zip I can't use the more recent ones, don't know why.

strange happened to me now.... it seems i cant get the raw video size past 1728

someone have any idea?

i was able to do it yesterday playing with it... (even with 1fps)
damn i still have the raw file in my laptop for 2512x1080 @14fps
now i it keep saying more than 1728 is not supported in video mode

The max resolutions that can be read from the 550D sensor are: 1728*1156 in canon's 1080 mode. 2512*1080 in cropped mode. 1728*694 in canon's 720P mode. That's if I'm not missing something, and of course at really low framerates. You must have recorded that bigger file in cropped mode (x5 zoom).

on my 550D, when I click Don't click me! the LV starts panning left and right

Those of you struggling with getting the higher write speeds with the new hack... Try formatting exfat and post your results

Same or worse results with exfat

Come on, man. are you serious? )) What's the reason for trolling?
Specially for you:

Sorry for my English, have to practice more.

Interesting. I wish I had found this info when I ordered my new cards. A very small increase in speed makes a difference with the new improvements.

ML on 550D works best with Sandisk Extreme (new) Series, 30MB/s & 45MB/s, don't bother getting a 95MB/s because it is both expensive and 550D has 21MB/s bottleneck on SD controller

I know, I actually meant Sandisk Thanks

This is starting to look quite good! Thanks devs!

I've just found out that my fastest cards are not the Transcends (as I thought, while I wait for the Sandisks*), but one called MFC that I hadn't even bothered testing. The Transcends write around 15-16mo/s, while this one gets to 18-19mo/s if the recording lasts a bit.

The problem is that the speed is very inconsistent. Sometimes it starts at 15mo/s and goes up, but some other times it shows 4mo/s or lower and the buffer gets filled quicker. I have the impression A1ex's last update (speculative start) might help a bit... don't know how to check this for sure, as the problem is randomish. Has this been studied? I remember this card was really cheap back when I bought it.

This version seems to run good for me, things I do with this one that helps is.
Turn Global Draw Off
Use FPS to set fps to 23.976 but 23 seems like it will go on forever
I was testing 1152x460
I use Canon Preview mode
And I used Magic Raw

But first try it with your normal settings you are use to and see if this version does better or worse.

If first record does not do good, try again and see how good next record does.

http://www.mediafire.com/download/25py137awp0s7x9/ML_550D_June_19_2013_Update.zip

UPDATE: After more tests I found this one was no better and recorded corrupted frames.

I'm getting pink frames with this one (*too). Never experienced this before, I think.

A1ex, if that patch works you'll have to tell us where to build you a statue.

In another order of things, would it be possible to keep the 10x crop for focusing purposes and avoid the camera crash? Something like ML turns Raw off the moment you go 10x and then turns it on when you're back to normal or 5x.

Right click on the clip you imported to After Effects -> Interpret

Then you can choose any framerate you want.

Here is my workflow step by step:

• 1. Import DNG Footage / Interpret at 12fps (or whatever your FPS was)
• 2. Drag into a 16bit Comp and change Comp fps to 24p
• 3. Apply Twixtor, set the Source Footage FPS to 12p
• 4. Precomp that into it's own Comp
• 5. Add Instant HD: ReSizer onto the new Precomped layer, set to fit to Comp
• 6. Render to 10bit 444 DNxHD or Cinestyle

Hope this helps! Also, it helps to limit the amount of camera movement and shoot at a higher shutter speed and use the amazing Reel Smart Motion Blur to put natural motion-blur back into the shot.

the trees' movement looks twixtored to me. Unless the scene is almost steady, I think we will get better results by shooting at the speed we need, even if we have to upscale much more. Especially considering that Canon's compression messes around with the resolution as well (and we didn't know till now). We just need to optimise the upscaling process.

Is there a way to record audio while shooting raw?

Yes, check the Raw menu (I'm not sure if that's working on every build we have). It produces a wav file and can emit a beep for syncing purposes.

(Remember that fps override turns sound off!)

Can someone explain what is the difference between 5x crop mode and normal recording? When I set a resolution without 5x crop engaged I get resolution which dont correctly relate to ratios. For instance, ratio is 16:9 and the resolution is 1216x416. And when I convert such footage it is stretched.

5x crop mode gets only a portion of the sensor. So it's equivalent to having a much longer lens on. The other benefit is that the sensor sampling does not skip lines, so unlike other modes, you will get no aliasing.

If you are recording on Canon's 720P mode, the camera skips some more horizontal lines than vertical ones. You will have to stretch the image in post (550D is around 1.67). The benefit is that you can scan a bigger area of the sensor while keeping the data rate (closer to the vision angle of your lens). This will generate more aliasing, though.

Thank you for the new version!
I get 1280x544 with 24 fps with Tragic RAW in 720p mode.

REALLY? You mean 544 after stretching from 326, don't you?

Native 1280*544 is the maximum I expect from our camera if we ever get 10bit Raw

Yep ! Many thanks ! I solved by setting "Hi Jello, fast Tv" ! Now I can do the experiments I want to do. If someone wants to follow me, it's good ! So:

[1] Set raw video resolution to 1920x1080, 16:9 - 1080p FullHD standard - (you can achieve this by entering 5x zoom Live View BEFORE enabling RAW video)
[2] Set override fps to 5fps (the maximum our little 550D can do at that raw resolution without loosing frames) and enable, in the same menu, "Hi Jello, fast TV"
[3] Set shutter speed to 1/300 or so... however keep it high
[4] Film... try not to pan too fast, at least to do this experiment... because too jello will ruin the video

Then, you have a quite Jumpy video, but at least the frames are not smeared down by the 1/6 sec shutter speed that you are forced to without enabling "Hi jello, fast Tv" in fps override menu. But no problem: now fire up Twixtor and interpolate up to 24fps, then sit back and render. Then... good vision of the final product !

If someone will beat me publishing his results, he is so much welcome ! 
(If it will work, I give you the permission to call it "dlrpgmsvc 550D RAW fullHD method) ! 

I think twixtoring 5fps to 24 fps is waaay too much. I'm getting better results upscaling, rather than twixtoring from 12 to 24fps.

Thanks a1ex

Sorry if this has been discussed before. I could not find a clear answer anywhere.

I've noticed that the way the Raw module measures the aspect ratio in 720P mode does not match my results. The calculations are made stretching by 1.40, but I have to use 1.67 instead to get a proportional result. Does this change between cameras or is there something I'm missing? (I'm using a 550D).

The thing is that the Tragic Lantern build does display it correctly to me: it changes from 1.67 to 1.71 depending on the resolution.

Excuse my silly question, but: how to record raw in 1:1 crop mode, so no line skipping ? And how to record raw in no-crop mode (with line skipping??) ? Do "no crop mode" equals to line skipping entering into play ? many thanks ! 

I'm not sure I understand the question but: In crop mode (5x, press focusing zoom on the camera and record from there) there is no line skipping. When the camera is not zoomed in (no-crop mode?), Canon's firmware reads some lines only.

Anyone having hard time deciding which record module they like best out of Magic Unified vs the Tragic version, this will help you.

http://www.mediafire.com/download/9v3l9lxfnmvbffg/Raw_MODULES.zip

Just use my latest bin to make sure your up to date on everything, but go into the ML/Modules folder on your card and delete raw_rec.mo and add these 2 files that are in this zip. You will have 2 new option within Magic Lantern.

When you decide which you like best, just delete the other module or just keep both, your choice.

The more resolutions are available for our slow cameras... the better. Thanks!

The last build it's giving me better results for the first 200 frames, but then I get drops more often than before. I'm using a Transcend class10 16gb

From what I remember, the main problem about that function was that it does no seem possible to keep the camera on when opening the card slot. I don't know if there's been any progress since then.

I don't think there is anything to gain by down scaling any bit depth of raw as it would entail discarding pixels and squeezing the rest into your chosen resolution. Standard EF lenses should work at any resolution on the 5D2 and if anything, this workflow increases the performance of your lens by utilizing the sharpest part of the glass. All you need is a 1000x cf card and you are able to record 1880 x 840 at 14 bit, which covers most of the horizontal on my LCD. Yes they are expensive, but if Canon offered an upgrade to unlock raw video capabilities, I wouldn't think twice about shelling out. I was also getting reliable 720 with a 30MB/s Sandisk Ultra.

I think there is a lot to gain. Maybe not for the fast cameras, but the slow ones will not be able to get much more than 540 lines in 24fps Raw. Knowing that, it's much better to get most of the sensor and then discard lines or pixels, rather than getting all the pixels in the central area (we've got the cropped mode for that).

Now, from all I've read, I understand that you cannot just read whichever lines you want, developers can't change sensor sampling modes (although g3gg0 seems to be working on that? http://www.magiclantern.fm/forum/index.php?topic=5558.msg38266#msg38266 ).

What I struggle a bit more to understand is why we can't read the whole sensor and then save only half the lines, or half the pixels, to have low bitrates but whole sensor images. Maybe the Raw information does not go through ML code at all? Non-developer dummy explanation: are we just telling Canon's firmware: get that much of the sensor and throw it to the memory card?