Music video: An'Hedonya - Purple Death, shot with 5D3 & ML (Danne's latestbuild)

Started by a.sintes, June 12, 2023, 01:18:50 PM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

a.sintes

Hello dear ML enthousiats,

Just released this music video for a dark folk band (with Leo Margarit from Pain Of Salvation on the drums) that I shot entirely with a 5D3 & Magic Lantern


camera: Canon EOS 5D mark III
lens: Canon EF 24mm f/2.8 IS USM
cards: 64GB CF (Komputerbay Professional 1066x UDMA 7) + 64GB SD (SanDisk Extreme PRO SDXC 200MB/s, UHS-I, class 10, U3, V30)

firmware:
Canon firmware 1.1.3
Magic Lantern Danne's crop_rec_4k_mlv_snd_isogain_1x3_presets build (2023-02-03)
cinematographer-mode module to deal with focus sequences

Magic Lantern configuration:
3.5K 1:1 centered x5 (x1.7 crop factor) preset
3072x1728 resolution (16:9)
14-bits lossless data format
fast grayscale framed preview
240MHz SD overclock
card spanning

camera settings:
23.976 fps
100-200 ranged ISO
f/5.6 aperture
1/48 (180°) shutter speed

lighting setup:
2x Aputure Amaran 200x (key lights)
6x construction lights (back lights)
Neewer RGB LED (eye light)

post-production:
Magic Lantern Video File System
Adobe AfterEffect (DNG to Apple ProRes conversion via Camera Raw: white balance, exposure, highlights & shadows, micro-contrasts, dehaze, color mixing, color noise removal, masked sharpening...) applying the ML-Log profile
Topaz Video AI (4K upscaling via Gaia engine, 2x slow-motion via Apollo engine)
BlackMagic DaVinci Resolve Studio (initial review, editing, crop, stabilization, wires removal, clip & timeline color-grading...) with Neat video (noise removal) and Dehancer (film print emulation) plugins
Apple ProRes export via Voukoder
It's too bad she won't live, but then again, who does?


yourboylloyd

Join the ML discord! https://discord.gg/H7h6rfq

a.sintes

It's too bad she won't live, but then again, who does?

vastunghia

Congrats, awesome edit!

And thanks a lot for sharing your detailed workflow... very interesting.

Personally I switched from ACR to DVR for debayering in order to have a leaner, faster and all-in-one solution. However I would be lying if I told you that I'm fully satisfied with final image quality. Highlight recovery, for one, is barely decent if compared with ACR magic.

There is also the fact that I find ACR to be problematic in some situations — in particular, where overall average brightness happens to change dramatically (say e.g. when you tilt your camera towards the sky at some point, even after having carefully ETTR).

In my experience, the same recipe applied in ACR to different frames of the same clip (so of course with same aperture, shutter speed and ISO) yields inconsistent results (mainly in terms of overall perceived brightness) in these cases.

Looks like the way ACR applies the same recipe is somewhat content-aware — which makes perfectly sense in the photography world.

There was one clip where I had to "ramp" ACR recipe settings between frames in order to get consistent brightness through the whole clip.

Wondering if anyone stumbled into this too.

5D3 for video
70D for photo

a.sintes

Thanks!

You summarized it very well: I tried also to do the switch from ACR to DVR to rely over a way more straightforward workflow but I really missed the tools quality of the Adobe solution to deal with RAW images...
Of course the highlight recovery which is incredible on ACR, but also the microcontrasts adjustment panel and the masked sharpening.
It's also true that being more a photographer than a cinematographer, maybe I just feel more confortable with ACR, using it alongside Photoshop to prepare primary adjustment LUTs to be applied later at color grading stage under DVR.

And yes, sadly I also observe what you describe (inconsistent results - sometimes brightness, sometimes unwelcomed color switches) on some of my clips: was unable to figure out why and I'm pretty sure you nailed it about ACR being kind of "content-aware"!
It's too bad she won't live, but then again, who does?

a.sintes

Regarding the post-process workflow applied to this video, please note anyway I would no longer recommand to use Topaz AI do do the 4K upscale over the media before creating the timeline under DVR...

I currently see three big advantages of using it the way I did:
- working with a proper 4K-conformed workspace (no strange 3072x1728)
- avoiding potential (proven) color shifts occuring when doing the upscale on the final export, after the color grading stage
- consistency of the processed image grain, that will be altered during the upscale (OR you need to disable it and use Topaz grain, which is awful compared to Dehancer...)

...but also two big drawbacks:
- the computation time is just intolerable (almost 3 continuous days of computation for my laptop...) when it can be reduced to a couple of hours if done over the final media
- the upscale (or slowmotion) being applied over log-based (ML-log) data, its result is sometimes less beautiful than if applied over a properly contrasted image

This is why I may currently recommand to use Topaz after the export, maybe using the following workaround:
- export output video in its native resolution with color grading applied but without the Dehancer node
- do the 4K upscale with Topaz (no grain addition)
- import/export the upscaled media on Resolve by adding just the Dehancer stage for proper film print emulation/grain addition, plus an initial color deviation fix node to compensate Topaz color shifts

Sadly, it currently leads to encoding quality compromises, Topaz being currently unable to export 444 ProRes formats (only 422), but I'm pretty sure it's acceptable when dealing only with an additional Dehancer correction stage.
And of course, this workaround cannot be applied when doing slowmotion (maybe it's better to do it over source media, but without the usage of the ML-log profile to rely over a proper contrast).
It's too bad she won't live, but then again, who does?

vastunghia

Thank you for your notes a.sintes, personally I find that sharing and discussing workflows is always helpful.

At the end of the day, it takes a few seconds to record one clip, but it can take days to deliver it!
5D3 for video
70D for photo

Nick_Andr

The picture of the video is very interesting. Color etc.
but very fast transitions. flickers.

a.sintes

It's too bad she won't live, but then again, who does?

a.sintes

It's too bad she won't live, but then again, who does?

Damiafix

Hi, well done! Lights and colors are just magic and the cut is extremely dynamic!  I am amazed by the results in terms of noise considering the dark environment. You managed it very well.

BTW  I am surprised you have been able to push the SanDisk Extreme pro 200mb/s up to 240Mhz  keeping the camera stable enough to film.  8) My 128Gb SanDisk exemplar doesn't even accept 160Mhz.

Best.

Canon 5d Mark III, Firmware: 1.1.3, Lexar Professional 1667x SD card 128 GB, SDXC UHS-II, up to 250 MB/s., Lexar Professional CF 128 GB 1066x (160 MB/s) UDMA 7

a.sintes

Thanks Damiafix!
Just saw your message, here's photo-captures of my settings, hope it helps.
It's too bad she won't live, but then again, who does?

vastunghia

Couple of questions:

Why Kill Canon GUI?

Why use Raw histograms? My understanding is that they are not reliable with this mode. Plus, they are limiting total bandwidth.

Thx
5D3 for video
70D for photo

a.sintes

vastunghia:

Canon GUI conflicts sometimes with the ML one and adds actually no useful information to me (current x5 zoom position), where I prefer to have my live view 100% focused on the image itself (cannot use external HDMI monitor with crop modes).

For the histograms, it was indicated that YUV histograms are based over JPEG data (used then as a fallback when RAW data are unavailble) while RAW-based histogram shows basically your whole 14-bits RGB data range (in a Log fashion in my case), and I'm unaware of any unreliability of it?
And yes it seems to limit a little the writing bandwidth when activated, anyway as this specific 3.5K 16:9 configuration requires approximatvely 104MBps I can currently afford this extra cost while keeping continuous recording (mostly thanks to card spanning and 240MHz SD overclock), the histogram being an extremely useful information (with both ETTR hint and clip warning).
It's too bad she won't live, but then again, who does?

vastunghia

Quote from: a.sintes on July 08, 2023, 01:46:58 PM
For the histograms, it was indicated that YUV histograms are based over JPEG data (used then as a fallback when RAW data are unavailble) while RAW-based histogram shows basically your whole 14-bits RGB data range (in a Log fashion in my case), and I'm unaware of any unreliability of it?

Compose your shot in picture mode. Now activate LiveView. Use ETTR indicator in Raw histograms to achieve ETTR. Now switch to video mode: ETTR indicator in Raw histograms will give you a very different, unreliable reading with crop modes. This is what I'm referring to.

Hope so hard I have always done something very wrong and actually there is a way to have consistent Raw histograms  with working ETTR indicators in crop modes.
5D3 for video
70D for photo

a.sintes

ho that's very sad actually, I'll check it.
is there any dedicated forum thread related to this issue with explanations about what's happening and if a fix is feasible?
It's too bad she won't live, but then again, who does?

vastunghia

Never stumbled upon any post related to this. If you find anything, pls report ;)
5D3 for video
70D for photo

Damiafix

Quote from: a.sintes on July 08, 2023, 09:40:12 AM
Just saw your message, here's photo-captures of my settings, hope it helps.

Thanks so much for sharing your ML settings!


[Full quote removed by mod. Please respect the netiquette]
Canon 5d Mark III, Firmware: 1.1.3, Lexar Professional 1667x SD card 128 GB, SDXC UHS-II, up to 250 MB/s., Lexar Professional CF 128 GB 1066x (160 MB/s) UDMA 7

Skinny

I like everything it's very cool but you already said the editing is very agressive, it is hard to watch sometimes...
Don't get me wrong I don't want to say anything negative, I love production and the music is also great, but I just miss the ability to focus my attention on a girl there :D you cut everything so fast, or maybe I'm just getting old..

Huge respect to everyone involved in making this song and the video, it is great guys! It is really nicely produced

a.sintes

haha I'm quite old myself and it's also a bit too much for my own taste with a little more hindsight now (my preferred parts are the contemplative Tarkovsky's levitation plans, by far), so there's no offense at all :p
note anyway these stroboscopic cuts are a kind of nowadays standard in this musical genre (metal affiliate) if you want to catch people's attention...
It's too bad she won't live, but then again, who does?

Skinny

Quote from: a.sintes on July 11, 2023, 08:00:22 AM
if you want to catch people's attention...
yeah that's unfortunate.. tiktok generation want everything fast :) It is really good to see someone who still find an inspiration in Tarkovsky movies :)

a.sintes

who doesn't? it's an endless and timeless source of poetry and visual inspiration...
It's too bad she won't live, but then again, who does?