Messages

Shot with Canon T5i/700D utilizing Magic Lantern in-camera software.  Magic Lantern RAW 2.5k Crop Mode Module @12bit depth/lossless compression, and SUHD overclock hack.  Workflow:  MLVapp to expand individual CDNG files from .MLV container, Lightroom to color grade extracted frames, After Effects to compile into video codec, Premier Pro to edit clip length and add soundtrack.  All footage saved in uncompressed .AVI codecs whilst in Adobe, then I used AnotherGUI and FFMpeg to transcode final footage into Apple Prores 185bit format for uploading to Youtube.  I used a Newer brand slider to pan the camera, then used Warp Stabilizer in Premier Pro to smooth out panning further.   Not too bad for a $400 camera and kit lens setup. 

By locking the registers between x5 and x10 you can bring x10 real-time preview into x5 vice versa in Canon 700D, here is the registers between the two:
https://docs.google.com/spreadsheets/d/1-3PlMmQHfUx7fh_u61p4v2O9rmP0madKaQjlkGWsBdo/edit#gid=0

I have started to look into it again; I am trying to get real-time correct framing LiveView in x5

Pull this off BilaFakhouri, and you've accomplished one of the greatest feats ever in the realm of Magic Lantern development.  This feature would literally transform the 700D into a cinema camera.  They will talk about your greatness for generations to come. 

Wait, what?

What actually happened:

- 1x3 readout removed "line skipping" aliasing artifacts present on most Canon EOS models (by reading out every single line, at the expense of rolling shutter "delay" being 3 times higher)
- regular Dual ISO causes aliasing in under- and over-exposed areas (because of 2 lines read out + 2 lines missing, in areas covered by only one ISO, ON TOP of Canon's line skipping)
- with 1x3 readout - which is pretty much oversampling -  Dual ISO aliasing artifacts were minimized (i.e. video recorded with this technique no longer suffers from extreme aliasing)

So, it's not Dual ISO that reduces the aliasing artifacts. It's the 1x3 readout. It reduces the aliasing so well that Dual ISO video starts to be usable.

This makes perfect sense.  Thank you.   

Dual-ISO was introduced as a feature increasing DR. See https://www.magiclantern.fm/forum/index.php?topic=7139.msg59687#msg59687
Aliasing artifacts are listed as a trade-off.

Years later it came as a surprise that under certain circumstances Dual-ISO can reduce aliasing artifacts caused by other technique.*

I have no idea what is gained by a question "I used xy for abc and lately I learned xy can be used for def and I want to know which one is right". Tools may be versatile. So what?



* Disclaimer: Haven't tested it yet.

Thank you for the explanation.  I think I was mostly misunderstanding the fundamental purpose of dual ISO recording in the context of Magic Lantern features.  You're right of course in that one should use the right tool for the job.  Dual ISO ability makes filming more versatile and offer photographers more options to achieve the desired end result.   

Hello,

Recently, a new module for ML appeared that facilitates RAW video recording at a dual ISO setting.  Along with a 1x3 sampling rate, among the features mentioned by the developer(s) of the module is the virtual elimination of both aliasing and moire' artifacts that have previously plagued ML RAW video recording at non 1x1 sampling rates (previously requiring crop mode recording).   My question is what ultimately is the advantage of dual ISO capability in the context of Magic Lantern use?  Is dual ISO required for the elmination of artificats and implementation of 1x3 sampling rates at non-crop modes or is dual ISO's main advantage increased dynamic range and reduction of noise in shadow detail? 


Thank you!   

Magic Lantern: The Arri Alexa killer.        Really amazing what a talented cinematographer can accomplish with ML. 

It says that the video is unavailable.

Hooray for Youtube's linking reliability.  It seems to be working now though.   

(Watch in 1080p mode)  Shot with Canon T5i/700D utilizing Magic Lantern camera software, Crop 4k on Steroids build, along with Danne's SUHD overclock module. Lens: Canon EF 75-300mm zoom @300mm. Shooting resolution: 2240 x 1080 @ 12 bit lossless compression. Shooting Mode: RAW Crop 2k @ 5x Liveview Crop Mode (3x effective). Workflow: MLVDump to extract individual DNG frames from .MLV archive; Adobe Lightroom to color grade frames; Adobe After Effects to compile DNG frames to encoded uncompressed .AVI file.  Adobe Premiere Pro to edit encoded .AVI file and add soundtrack, export again to uncompressed .AVI file. FFMpeg with AnotherGUI Interface Software to transcode uncompressed .AVI footage to Apple ProRes 444 format. Note: The very latest versions of Adobe software will export directly to Apple ProRes format as of December 2018. My software is older, so I must utilize the steps above.

Lines x Columns

1x1 Sampling is read every line (vertical pixels) and every column (horizontal pixels) without any skipping or binning, no aliasing and moire and you will see more noise in this mode , also Movie Crop Mode and x5 Mode are using 1x1 Sampling.

3x3 Read one line skip two every 3 lines/Bin every 3 columns. And skipping lines causes aliasing and moire.

1x3 Read all lines/Bin every 3 columns. No skipping at all only Binning pixels in columns, no aliasing/no moire and better details than 3x3.

3x1 Read one line skip two every 3 lines/Read all columns. Better details than 3x3 but with same alisaing and moire since we are skipping lines in this mode.

Aha, this makes much more sense.  My previous understanding was a little skewed, but I fully grasp line skipping vs binning now.  THANK YOU. 

I'm still trying to fully grasp what pixel binning is and how it relates to the final image quality when shooting ML produced video.  If I understand correctly, 1x3 binning equates to 1 out of every 3 vertical lines of resolution being imaged from the sensor whilst 3 horizontal pixels are grouped together or "binned" as 1 pixel?   Therefore, 1x1 sampling equals 1 out of 3 vertical lines being read while no horizontal pixels are being binned?   In 1:1 mode using Movie Crop mode on the 700D, this would equal vertical line skipping but no horizontal pixel binning?

In 3x3 mode, this equals no vertical line skipping whilst binning 3 horizontal pixels?   

I'm ultimately trying to understand why 1:1 mode in standard crop mode recording on the 700D virtually eliminates aliasing and moire.  Thanks so much for any insight you can give me!     

Here's a quick shot using the 5K Anamorphic setting (1x3 sampling) and Dual ISO processed to ProRes 422 HQ with MLV App (and new focus pixel map files). Didn't see any pink highlights using Danne's latest build.

Yes, I know it is a bit of a cheat calling it 5K because the horizontal resolution is "only" 1184 pixels stretched out to 3552. What matters is the final results.

Wonderful, amazing stuff.  Question: What does Anamorpic setting accompish exactly?  Is this a way to get around the normal limitions of Liveview Framing Mode during crop recording? 

Any progress at all regarding realtime preview with proper framing automatically makes you a true hero among heroes on here, theBilaFakhouri.      This would allow action footage to be properly filmed, along with the ability to use a steady cam type support for stabilized moving and panning sequences.   This just isn't really possible at present with a horribly lagged framing preview which seriously debilitates all of the other wonderful accomplishments with Magic Lantern in my opinion. 

Curiously, with the settings you posted on this thread using the 700D, I'm able to go into 5x crop (3x crop mode technically) mode and record RAW 2240x1080 with one of Danne's built-in scripts.  I'm assuming that it is recording with forced 3x3 binning at RAW 2240x1080, but the footage looks fine and realtime preview seems to have near 1:1 framing with just a little bit of lag during record.  Compared to normal framing preview at 5X crop, it's the difference between night and day; one could most certainly use a steady cam type support device and record.  I'm just not sure I can live with 3x3 binning though, as this would really leave footage vulnerable to moire and aliasing.  Still, this is very serious progress on your part, wow, and thank you for posting your research!   

Hi Experts,

By mistake, I just formatted my SD card which was used to install ML in my camera. After that it is not booting up at all.

I tried removing battery, lens and connecting with computer through EOS utility. nothing is working.

My set is full dead.

Can you please help me on this?

Did you format the card in the camera itself?  Will the camera boot without the card installed?  Magic Lantern does not alter internal ROM code on the camera itself except for a bootflag option, so it's more than likely the card itself has been corrupted.  If your camera boots without the formatted card installed or with a different card altogether, this is definitely the case and you'd just need to properly reformat the card and reinstall Magic Lantern on it through the standard procedure. 

Hello,

I'm currently using a Magic Lantern Crop 4k build for the EOS Rebel T5i/700D that includes options for 2K+ resolutions in 5X crop mode.  I'm trying to gain a better understanding of what full resolution liveview recording is versus crop recording?  Does full resolution liveview recording facilitate 2k+ resolutions in a full frame rather than a cropped frame liveview?   My current understanding is that for the T5i/700D, one must always be in 5x crop view mode to record in 2K+ resolution and that full frame recording in high resolution is not possible due to hardware limitations. 

I'm wondering if I misunderstanding the concepts here and that I'm not fully grasping what full resolution liveview recording is.   I would greatly appreciate any insight that could be provided, thank you!   

(Watch in 1440p HD mode) This is footage was taken via Danne's experimental Magic Lantern SHD overclocking module build that includes auto-script settings for SHD write speed overclocking as well as Crop Mode 2.5K and 10-bit lossless compression format for RAW video. I utilized the freeware MLVapp program for color grading and directly exported from .MLV format to Apple Prores 444.

[Requirements:]

Since I had to dig a bit too much for my patience to find an up to date and or exhaustive guide for 700d raw video, this is mine.

Requirements: UHS I or better SD Card. I've a SanDisk Extreme PRO UHS II which is definitely overkill, gets me around 60mb/s.

Ten easy steps:
1. Get latest build from here and install as usual https://bitbucket.org/Dannephoto/magic-lantern/downloads/
2. Enable all modules and reboot.
3. Under Scripts, press Q on "enable SD overclocking" and enable "Autorun". You could do the same for any of the recording modes listed, if you plan to use only one in particular and wanna make sure it's always working properly. Reboot.
4. Set your exposure correctly (make sure ISO as low as possible, over 1000 you start artifacting a bit, shutter speed to 1/48 if shooting in 24 or double of any framerated needed, finetune it to exactly that through Movie - Image Fine-tuning - Shutter fine-tuning)
5. Set up MLVFS on Win/Mac, guide here : https://www.magiclantern.fm/forum/index.php?topic=13152.0
6. If you get any pink dots, chroma smooth with 2x2.
7. Use Adobe Camera RAW to grade the image sequence, by importing it directly from the MLVFS virtual drive. I usually just hit "auto" on exposure and white balance, and select my lens from the profiles to remove any distortion / vignetting.
8. Render uncompressed avis or dnxhd movs in order to edit, directly from SD. It's faster for me this way, than copying files and then exporting, since it happens at the same time. You need to prerender - on a 250gb EVO SSD I can't edit smoothly in Premiere or After Effects without prerendering, maybe Final Cut handles image sequences better?
9. Edit your clip. Add grading. Upscale to 8k when rendering, YouTube allows for higher bitrate and better grain display. I use .mov dnxhr hq 8 bit in After Effects to render 8k. H.265 works fine as well, renders much slower in my experience, and I have a good upload speed so I afford uploading gigabytes.
10. Enjoy near 4k quality from your $500 camera lol.

A couple of examples: https://www.youtube.com/watch?v=-fwEpsWXQ04 - clean footage. https://www.youtube.com/watch?v=fYVwqTO7Blo - heavily edited footage.

Thank you for sharing!  IMHO, your method is by far the most practical and useful way to accomplish 4k rendering on the 700D.  SD_UHD overclocking is the only way to achieve useful UHD footage with the 700D at present. 

[crop_rec with higher resolutions]

Experimental crop_rec with higher resolutions for 700D (it Should work for EOS M)

This is a great accomplishment in terms of unlocking these higher resolutions of course.  Realistically, from my own personal experience with the 700D and a class 3 write speed Sandrisk Extreme Plus SD card, everything is severely limited unless one overclocks the SD bus via Danne's Crop 4K build.  I can only yield maybe about 3 seconds of footage before frame skipping, even at the 2.5K preset.  Does anyone have any different experience? 

What ever happened to 4k raw and 12 and 10 bit raw coming to the t5i? if it's here already or if there's a test module how dose one find and or enable it? also what about losses compression?

There is a very experimental build that allows both 2K and 3K resolutions on the T5i/700D.  My best understanding is that it requires overclocking the SD card bus in addition to running a special module on a Crop_Rec 4K build of Magic Lantern.   

Info here:

https://www.magiclantern.fm/forum/index.php?topic=19300.msg202049#msg202049

and here:

https://www.magiclantern.fm/forum/index.php?topic=5951.msg202957#msg202957

[2520x1384 16:9]

Let's try the new resolutions on 700D

1. 2520x1384 16:9 @ 24fps
2. 3K  3072x1304 2.35:1 @ 20fps

https://www.magiclantern.fm/forum/index.php?topic=19300.msg202049#msg202049

I cannot get this to work on the 700D no matter what I try.     Is your module designed to be used in concert with the SD UHD hack?  Could you please be more specific as to how and when to switch from Liveview to Play mode during the process?  Most of the time I simply experience a lockup when trying your module, but the one time I actually did get it to almost work in 3x crop mode it immediately frame-skipped and stopped recording. 

I'm running the latest Crop 4k branch for the 700D with firmware 1.15.  Thank you so much for your efforts - if you're successful with this module it would be the biggest accomplishment yet with Magic Lantern and the 700D.