Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Feature request for Dragon! In-camera scaling to 2K, 1080p and (most important) 4K

We could shoot at 3:1 all the time (and do slow mo at very high fps that use the whole sensor)!



Do you actually shoot 3:1 compression?
Yes this will eat your hard drives like no tomorrow.
If you do not use 3:1 compression what is your set up on compression.

i shoot 10 cameras 10 EPIC cameras on a TV show ultra low budget and have tested compressions in various ways under all sorts of circumstances.
Ive seen them projected on REDS theater screen from compressions 3:1 all the way up to 18:1.

Whats your compression settings at this may or may not be hurting you.

Best of luck
 
Meizler Module is going to give you the same workflow as the Alexa supposedly. 1080p DNxHD or prores proxies on mini-mags while still recording full res R3D's to the RedMag's.

...it's just not going to do it for free. The module does look fairly light and compact though. Been waiting on mine since it was announced. Alexa is an exceptionally popular camera here in Toronto in large part due to the simple workflow, the Meizler should level that field.
 
As far as I know the Meizler module will not record slow motion proxies however. What was the maximum frame rate 60fps or so?

Myself I don't really see the problems with red Raw. I just go into Redcine, do first color tweak and then generate proxies in Final Cut Pro X. It works pretty problem free. If I'm in a hurry and really need to edit fast I can work in raw directly on an Apple Retina laptop. It's fast enough to work with, but of course not as smooth as proxies. If I need to hand over work to others i just give them the proxies from Final Cut Pro X.

/Andreas
 
Karim D. Ghantous said:
BigLu has made a good point. Otherwise, Patrick Faith has solved your problem:http://www.reduser.net/forum/showthread.php?99014-Dragon-Update&p=1195891&viewfull=1#post1195891
Click to expand...

Honestly, no offense, but i am deeply familiar with compression rates and how they affect image quality depending on circumstance and what kind of grading is likely down the pipe. if you are doing strong grading, it's not that simple as what is described here - not by a long-shot. B+W compression is not the same as color, and the math is very different
 
BigLu said:
We could shoot at 3:1 all the time (and do slow mo at very high fps that use the whole sensor)!



Do you actually shoot 3:1 compression?
Yes this will eat your hard drives like no tomorrow.
If you do not use 3:1 compression what is your set up on compression.

i shoot 10 cameras 10 EPIC cameras on a TV show ultra low budget and have tested compressions in various ways under all sorts of circumstances.
Ive seen them projected on REDS theater screen from compressions 3:1 all the way up to 18:1.

Whats your compression settings at this may or may not be hurting you.

Best of luck
Click to expand...

i have advanced training in compression trade-offs and have spent many hours testing grading and other post matters and the effect compression has on them. I know exactly what I'm doing. 3:1 has its place, not all the time, but certainly it has its place.
 
I don't own an Epic but it would be cool to get 1080p RAW AND Uncompressed, Maybe even with 2K!
 
I'm probably missing something very obvious but on the Canon still cameras but it does look like they are doing RAW to RAW sizing with the M/S RAW options ?
On the Mark III size are (RAW) 5760 × 3840 (M-RAW) 3960 × 2640, (S-RAW) 2880 × 1920, and i'm using the M-RAW option quite often for timelapses with very good result.
if the same thing could be done on the Epic i would definitely use it on a few shoots..
 
Loic Parent said:
I'm probably missing something very obvious but on the Canon still cameras but it does look like they are doing RAW to RAW sizing with the M/S RAW options ?
On the Mark III size are (RAW) 5760 × 3840 (M-RAW) 3960 × 2640, (S-RAW) 2880 × 1920, and i'm using the M-RAW option quite often for timelapses with very good result.
if the same thing could be done on the Epic i would definitely use it on a few shoots..
Click to expand...

They are not doing a proper re-size - thus why there is still quite a bit of aliasing and over-sharpening - very unflattering to women BTW.

They are doing what could best be described as a "smart line skip" - not quite as bad as the MKII's hideous line skipping and rainbow moire, but not top-of-the-line image scaling either.

It won't always be obvious, sometimes subtle, sometimes very obvious - as a difference in texture quality and definition. I find it particularly ugly on skin pores, as mentioned.
 
The ML raw thingy is not using the algorithm/chip discussed here, which cannot work in real time at 24fps. The linked discussion applies to still photos only.

ML is taking the video stream before it is compressed to H264, and seemingly before it is also slightly blurred. Others have figured out that on MkIII, line skipping debayer cheats were improved, but were still imperfect, and that for some reason the video image was also receiving a 1.1 pixel or so gaussian blur, which ML bypasses completely.

My own tests with skin under different lighting comfirm this.

ML is OK, but it's not superb. Side by side a shot of a the same scene looks suspiciously very, very similar to a a 1/4 Red debayer, which I would never use for final output.

Besides an initial lack of creaminess, which might be subtle or not, the additional problem is that such faked resolution based on line-skips is not consistent frame to frame - it may sometimes look like it is, but technically it's not. Therefore compression algorithms like those used in blu-rays, digital broadcast, youtube and even some theatrical deliverables get fooled and confused, and can't work effectively. They depend on recognizing the exact same pixel pattern moving from one frame to another. WIth line-skipping, the pattern changes slightly on every frame, so the compressor thinks it's a new object.

I did a test here using x264 and VBL that allowed a Max rate of 50 000mbits. Same frame, same lens (cameras side by side).

5K downscaled to 1080p in DaVinci vs ML 5D MKIII Raw 1080p - both became a 1080p AVI which I compressed in X264 - using the exact same settings.

The 5D X264 output files were almost three times the size of the downscaled Epic files.

This may well be why Canon opted for blurring - because it helps compressors recognize moving patterns.

These always changing patterns also reduce the ability of VFX trackers, especially planar trackers, to remain perfectly stable.

You can see here

https://vimeo.com/66974538#

At 1:30 the line above the awning - that is "smart line skipping" and here it is obvious, but it isn't always to the naked eye, although I find it tends to be on women's skin, which my wife and I are always so careful about in terms of lighting and makeup - so this kind of sharpening would be the last things we would ever want.

Look here:

https://vimeo.com/channels/529954/66962637

Look at the bricks on the wide angle buildings around 0:40 - they look like a mild form of what happens with insufficiently antialiased 3D animation, and it's creating visible luminance-flicker.

Here it's even more obvious on the slow pans a few seconds in, despite it being shot at 1/50th second.

https://vimeo.com/channels/529954/66955398

Most RAW MKIII test videos have such poor focus - or no pans or camera movement - that the issue is less obvious, but it's always there in anything moving that is in focus, and once you see it, you can't un-see it.. I don't know about you guys, but I'd like the option of panning in my work, at any speed I like, thank you very much.

These issues cannot be resolved by a new version of ML - because real-time proper scaling chips able to shrink 21 mp to 1080p in real-time at high quality could not fit into a body like the 5D's without serious heat issues, and anyway they are not included with the camera.

Remember also that Vimeo is forgiving in terms of compression levels - youtube much less so - so if you send a compression-unfriendly video to youtube, it will look far worse.

The above issues, we'll call them "inferior texture rendition" along with :

- very short recording times possible on CF cards and huge file sizes compared to redcode - it always uses 1:1 compression Data rates of 1080p = 5K at 4:1
- lower dynamic range (11.5 vs 13.5 stops on MX)
- Much less nice curve-to-white / shoulder than Epic.
- Skew/jello issues far worse than Epic
- unreliability (dropped frames can and do happen)
- poor ergonomics - no real EVF, etc.
- Occasional issues with hot/dead pixels
- occasional overheating issues
- tedious post-production workflow - Red has very great Raw settings panel in many software, and debayer is FAST.
- impossible to reframe
- less resolution for VFX planar trackers, green screen, etc.

Would make me completely uninterested in ever using an MkIII + raw for any shoot except if absolutely necessary as crash-cam, helmet-cam, hidden cam etc.. This EVEN if the project was only destined to 1080p or even 720p YouTube.
 
Last edited:
You must log in or register to reply here.
Back
Top