General 5D MkIII + ML + RAW queries

Started by matthewt, May 01, 2014, 07:33:12 AM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

matthewt

Hey there,

Just doing some research for a Music Video project and a few questions with regards to using ML and RAW for commercial use on a 5D Mk III. I have done a fair bit of looking around but have struggled to find clear up-to-date answers - any advice would be tops.



  • Is it stable to use the latest nightly (Apr 29 approx) ML build for 5D Mk.III running the updated 1.2.3 firmware? Is there a 'most stable' recent build? Is it stable enough for commercial use or should I downgrade to 1.1.3?

  • If 1.1.3 - which is the most stable build? Is there a build that has simple installation like the current ML 1.2.3 installation.

  • I have read a lot about RAW vs MLV here. Is there a definitive winner/difference other than MLV being self contained and allowing simultaneous Video + Audio recording for guide tracks? Whats the recommendation?

  • How long can you realistically record for without drop frames/other issues in RAW? Would you be capable of doing takes of 2min+ in length?

  • Is formatting in disk utility to ExFat the best way to handle the file spanning / card performance issue?

  • What about applying a LUT in camera to the RAW footage. Is there a simple way to do this? Or something similar to shooting in a cinestyle or VisionTech colour profile for RAW/ML as opposed to .h264

  • Can you playback your RAW rushes in camera?

  • Is 1.2.3 / 1.1.3 ML compatible with the Small HD ac7 monitor?

  • Can you get HDMI split out for client monitoring?



Oh and any advice for workflow? We're shooting 25p in 2.40 (1920x800) thinking of going for something like this:

.RAW out of camera > RAWMagic to convert to .cDNG > Lightroom for a basic RAW touchup and denoise (apply c-log like curve) > export out as .tif > take image sequence into Davinci Resolve to grade > export as .mov ProRes444 file > Premier Pro to edit > Out as .mov ProRes 422 (HQ) for client and .mov .h264 for web.


Will Davinci accept the .tiff image sequence, or should we use something like Media Encoder to turn the image sequence into a .mov ProRes 444 file? Will we lose the benefit of working with the RAW footage?


Will be doing some testing over the next couple of days and will post what I work out. Again, any advice/guidance would be killer! Thanks everyone for making this possible. DREAMZ.

Edan


SteveScout

Oookay, let´s start. ;)  Can´t answer everything, though.

- The 1.2.3 is definitely not as stable yet as the 1.2.3. Problem is that most people did not shoot hours of footage on 1.2.3 firmware yet, on the 1.1.3 with latest ML builds they did. On my tests 1.2.3 worked fine, but I did not use it yet on set and I won´t very soon as I´m just about to be satisfied with the MLV implementation´s "buggy-ness" of MLV on 1.1.3 which you can fully use - but RAW is still more reliable. Get your cards, your build, your camera - and shoot tons and tons of footage. Long takes, short takes, everything inbetween, that´s the only way to find out if 1.2.3 works for you.

- On 1.1.3 there has been a time where new builds introduced new quirks, but at least for my testing abilities this has long been gone. Feels safe to always use the latest. But never go into a shoot with a untested build. Before my last week of shooting a feature on ML raw/mlv I did not update the build for two weeks prior to shooting and was only testing the stability of the actual one.

- mlv vs. raw - if you don´t need metadata and audio it´s still faster and more safe to use raw. On Komputerbay cards (which not always give you the full speed they are advertised for) it´s especially tricky as video+audio recording needs a little bit more overhead. On mlv you could theoretically record to SD simultaneously, but I did not find this feature useful in the field for recording.
If you need audio - for syncing with an external source for example - then go with mlv, but all the time if I just need the raw videodata for a project, I still switch to raw.

- You can realistically record the full 12 minutes until the card is full - on a 64GB card. I do not own 128 GB ones. Shot a lot of 10 to 12 minute takes WITH sound and had only two corrupt files in a week of shooting.

- Yep, exFat is the way to go. Everything else (merging the spawned files) works, too, but you´ll need that extra step.

- No LUT. You can´t apply a LUT to raw data, it just does not work with the concept. The camera´s hardware is not fast enough to debayer the images AND apply a LUT. It can´t even debayer in realtime. That´s the concept of ml raw - you push all the processor heavy stuff into post, on the computer side. Feel free to add any LUTs in DaVinci or ACR.

- Yes, you can playback your shots in camera. And even delete them. Not realtime, though, but now even in color.

- On 1.1.3 you might not want to use an external monitor for focusing etc. as the resolution of the HDMI stream is too low. 1.2.3 fixed that but I read a lot of problems that the image on the screen was shifted, don´t know if this is fixed. Anyone?

- As for the HDMI splitter .. never used it. Believe me - you don´t want to use ML with a client involved. If we talk about a real client. I´m a commercial director and I LOVE having my 5DMK3 with ML raw with me on set, but I would never ever use it as the 1st camera. Not because of the quality, but for safety issues. Unless you shoot only 5 minutes of footage a day you WILL definitely have stopped takes at one point, some errors here and there and it will need some sort of acting to show to the client that this is normal and you´re just swapping cards.

ML makes a 5D the perfect stealth highend second camera which you can easily intercut Alexa oder RED (or blackmagic or whatever) material to .. but it´s not an A-camera for a paid job where the client watches every of your step and you might have to explain delays. Might be all fixed one day, but it´s still not as easy as pushing a button and stopping and playing back. What if the client wants to review a take? How do you tell him it´s not 25fps he´s seeing on the monitor?

If there´s clients involved, get a A camera and equip someone with your ML camera of choice to get closeups, additional footage, whatever, and you will have a much happier shoot than with the fear in your neck that your camera might stop in the middle of the take.

Then your workflow .. this looks .. well .. strange. First of all - don´t shoot so heavily cropped (2.4). What if you want to stabilize? Or reframe? Not saying you should shoot 16:9, but think about 2.0 at least.
For my taste you´re doing way too many steps which might be fine for a small project, but that´s definitely not the smartest way. Why not going directly into Resolve with the DNGs? You can apply your log curve (CineLOG, see the thread in the forum here) there as well. From there you can go to ProRes files for editing and you saved the whole Lightroom thingie.

I do it differently .. I don´t use Resolve but ACR in After Effects to create quicktimes (DNxHD, on MacOS I´d use ProRes) and apply the log curve at the same time. Have a script for importing, render times are long but done overnight. Then I edit the quicktimes with a LUT/fast color corrector on top and use whatever I want for finishing - Speedgrade for example.

I would consider the TIFF files you wanted to make an unecessary step. You loose the benefit of the raw file the minute you´re converting the DNG to a TIFF. You´d need to convert it to linear EXRs to hold all information, but there´s no way editing with that. That´s why 10bit LOG video files have been invented, to store the full data like in the 10bit cineon files. If you apply a log curve to your DNGs and export 10bit ProRes HQ or 444, you basically still  have all information - you just need grading to bring it back to a "viewable" form.

cheers,
Steffen

dubzeebass


DocMartensLutherKing

Solid post!  I too am wondering how an HDMI monitor looks with 1.2.3 - I really wanna get one for my rig.  Anyone have any info?  And can anyone who's using a monitor give their recommendation?

SteveScout

Glad you like it! Learned so much in this community that it was about time to give something back! :)

jrcreative

After reading through a few posts i thought we should be formatting all cards in the camera, is this not correct?

Walter Schulz

Formatting a card outside a camera will delete bootflag and ML won't start. And ML files are gone.
Formatting a card inside a camera where ML is *not* running will just be the same.
Formatting a card inside a camera where ML is running will give you 2 options: Default is "Format card and keep ML on it" or "Format card and wipe ML from card".

Ciao
Walter

jrcreative

Thanks, so to start when i format the SD card before putting ML on it I'll do it in the camera as suggested on previous posts.

GreerPhoto

With the new high speed CF cards like the ones from Toshiba hitting the market will we ever have the ability to record RAW on the 5D mark III at higher resolutions than 1920 x 1080? The new EXCERIA PRO™ CF cards  achieve a read speed of 160MB/s and write speed of 150MB/s. It would be great to have the ability to record 2K with an aspect ratio of 2.35:1.

Thanks
Jerry Greer
Greer Photography/EnviroStock Media/Mountain Trail Press

SteveScout

So far I don´t see an advantage in formatting the card in camera. Especially on a bigger shoot day with several people involved the DP or camera operator should not think about card formatting etc. - he gives a full card to anyone dealing with the data/copy process and gets a fresh exFat formatted card back once it´s properly backupped. Everything else - in my opinion - can just cause problems.
And never problems with a card that came formatted from the computer.
The SD card with ML on it stays untouched anyway. Of course this only applies to the 5DMK3 with its dual slot.

Regarding the question of the highspeed cards - I don´t own them so far, but as we´ve heard from numerous sources in this forum, after around 100 MB/s the camera´s internals seem to become a bottleneck, so you can´t expect it scales endlessly with the speed of the cards. We´ll see, hopefully it works with future faster cards, but I doubt it.

matthewt

Thanks very much for the unbelievably thorough answer steve.

Seems like the RAW hack (as great as it is) just really isn't suitable for any client inclusive workflow.

I've done some work with it now and I've found that RAW seems to be easier to handle in post than MLV (going straight into RESOLVE) so the workflow has simplified.

Had one follow up though:

Is there anyway to get playback of the RAW clips in camera (not using MLV)?

Seems to be fine to shoot to the 64 GB Komputerbay cards (ExFAT formatted) and the smaller size of the image (2.40) saves a lot of space in file size when shooting.

Thanks again
Matt