Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Messages - masc

Pages: [1] 2 3 ... 66
Camera-specific discussion / Re: Canon EOS M
« on: Yesterday at 09:33:48 AM »
Thanks! I missed that one - works great, no more flickering in the lower part.

Camera-specific discussion / Re: Canon EOS M
« on: September 21, 2020, 11:23:04 PM »
That's indeed epic. Tried this mode: is it for you too continous at 14bit??? I get a green icon here... Awesome.
For the first test it crashed with this kind of console output. After that it worked nicely. Is it normal at the current dev stage, that the lower half of the screen does "anything"? Sometimes I get weird lines, or black white blinking, or a part of the last shown liveview image, or... any idea if it could be possible to make that just black for example?
Another finding: the histogram: while zabras already mark the overexposed highlights, the histogram shows 1..3EV space until overexposure. After a record, it fits again as it should.
But this RT liveview ... just wow!!! So much better than before!

H.264 uncropped is 3x1, correct? Is 1080 MCM 3x3?
H.264 uncropped = 3x3
1080p MCM uncropped = 3x3
Anamorphic modes = 1x3
Currently not available in most builds (only available for 5D2) = 3x1
Crop modes = 1x1

If so that could be considered a medium between 1x3 and 3x1 in aliasing and moire?
No. Reading all lines -> no moiree. Reading each 3rd line -> moiree.

Post-processing Workflow / Re: Remove aliasing and moire on 1:1 H.264?
« on: September 18, 2020, 08:22:51 PM »
Just record RAW and all will be easier. H.264 won't become much better than it is.

Yes. 5k anamorphic (crop~ x2).
For EOS M 4.5K anamorphic has a crop of ~1.18x.

Cool. How do the 1x3 and 3x1 modes compare?
1x3: no moiree.
3x1: a lot moiree.

Always just use what you need to use, and what the footage makes looking like you want it. If you need good quality, it will mostly be slow. If it is fast, quality will mostly be bad.

My bad! I thought both were made using Qt and FFMPEG.
So far, so right, but this is not used for processing.
I feel like the overall render experience on MLV App is a little funky,
with the limited preset panels and render settings...
What do you mean with funky? And what is limited? Other guys say there are way to much features... :D
And feels a little bit unnatural the way MLV app freezes while rendering, leaving the user with a progress bar and a Terminal/CMD running in background.
If you get freezes, please install a debugger and tell us where it freezes. This could help to fix such behavior. Never got any freeze in the past year on all my computers.
For the terminal/CMD window you must tell your thanks to Microsoft - it is a Windows only feature. All other OS are able to run external apps without this ugly window (and tunnel data to them).
These may be just nitpicking, but something closer to AME or DuMe, or a way to send the rendering file with the color corrections settings to DuMe would be nice.
As far as I can see in the demo, DuMe does no processing at all. It just configures FFMPEG to transcode into another codec. This is something very different.

Thanks for the links. But what do you mean with same tools and frameworks? Qt and FFMPEG won't speed up anything. MLVApp has a very unique processing engine. I don't think it will get much faster soon.

Raw Video / Re: QuickStart for 50D - Raw Video
« on: September 06, 2020, 09:12:33 PM »
Yes, but it is applied when using the camera matrix in MLV App is it not? Or at least it seemed that way to me. With my 'neutral' settings I seemed to get a nice flat-ish output visible in MLV App (with camera matrix applied) ...
Let me know if I'm wrong in my assumption of the camera matrix please. Thanks.
You're wrong here. Picture style is not recorded at all into MLV. There is one camera matrix per camera. Camera matrix is a kind of sensor calibration.

This gave me a curiosity to try Hunter's lut, and I think I will adopt another workflow : in mlvapp apply the alexa c log profile ( + set WB to 5500k° and 10 tint) & render as prores 4444. then in Davinci I apply this lut.
You would get better quality if you apply the LUT also in MLVApp. In MLVApp the LUT is applied on (3x)16bit color data. After exporting to ProRes it is 10bit, or 12bit with AVFoundation.

Please don't double post. The question was already asked here:
It seems the dependencies in Fedora are different than in Ubuntu. You'll have to compile the App on your own, or use e.g. Ubuntu. I expect this a bug in linuxdeployqt, which should include all needed libraries. Maybe this helps:

Share Your Videos / Re: Seattle | 5D MKIII | Danne's Build 3840x1536
« on: August 21, 2020, 07:56:52 PM »
I also shot some of those wide 16mm shots at 2.8 on accident, so it was probably my fault. ...
I hope to post more video soon, and I look forward to seeing more of yours!
Why by accident? I always use it at 2.8, because it is the most cinematic look this lens can produce. :)
Hope to see more from you too ;)

Share Your Videos / Re: Seattle | 5D MKIII | Danne's Build 3840x1536
« on: August 21, 2020, 01:21:25 PM »
... When you export a cinema dng from MLVAPP, what happens to the color space? ...

Does the color space just get interpreted by Resolve and the cinemadng file contains all of the original RAW data?
cDNG = RAW. It contains no color space. Resolve has to interpret the RAW data.

Nice video, nice mood, nice quality! Is this recorded with the first version EF16-35 2.8? The borders are veeery soft. I use the 2nd version and can't remember having seen this before. Is there a way to correct the CAs in Resolve? In MLVApp you could do this - at least if you don't use cDNG. I also used this nice UHD mode from Danne... you'll find it under "Share Your Videos" here.

Raw Video / Re: How to eliminate Vertical Stripes from DNGs?
« on: August 13, 2020, 08:25:55 PM »
You could try to import & transcode the DNG sequences back to MLV - in MLVApp.

General Help Q&A / Re: Some questions about specific terms around ML
« on: August 13, 2020, 03:35:26 PM »
It would be interesting to compare 5.7K 1X3 with Topaz video enhancer that upscales 1:1 to 5.7K as well
The biggest difference is the depth of field, and this can't be simulated by any software. Here just the used physical size of the sensor counts. And this size is just ~ the half for 1x1 reading, while the read pixel count is not sooo different. So 1x1 is just slightly sharper, while 1x3 looks way more cinematic (for my taste).

Camera-specific discussion / Re: Canon EOS M
« on: August 13, 2020, 11:51:54 AM »
...You should be getting a question if Mlv App needs updating I think...
EDIT: Maybe the official Mlv app version now still didn´t get the pixel updating automation included yet.
The latest official version can't do this. You need a inofficial newer version.

Share Your Videos / Re: 5.7K anamorphic 5D Mark III footage
« on: August 13, 2020, 09:01:30 AM »
@paulheran: Just copy both files into one folder. MLVApp recognizes card spanning automatically then.

SD slot needs to be overclocked for 5.7K mode, because otherwise it is by far too slow.

I've been shooting RAW for years with other cinema cameras without ANY such issues. BRAW from Blackmagic completely flies, for example. So is ARRIRAW.
And why don't you use these cameras if they work so fine? You really compare the most expensive professional solutions with open source free time projects? And then you cry because the professional solution is faster?

The only Blackmagic cameras you can really compare to ML record cDNG. This footage brings exactly the same speed using exactly the same applications (e.g. Resolve). BRAW is partially preprocessed and works differently.

But I'll stop now to answer and to explain, because it seems you still have nothing read I wrote, still have nothing understood I wrote, but you're still rumbling.

Share Your Videos / Re: 5.7K anamorphic 5D Mark III footage
« on: August 10, 2020, 04:18:33 PM »
While setting the shot up you'll have "Framing" and while recording you'll have "GreyScale".

I just told you that this is not what happens here. I open the file in MLV app and it's VERY slow. It does not playback real time, as you suggest.
Please read again my posts. Nobody is saying that MLVApp PROCESSES ( = playback ) MLV files in realtime. I think you should read a lot in the forum to learn what RAW footage is and how to handle it.

The "reading" of an MLV is faster than real time ...

That statement is NOT true here. I tried it BOTH on a Windows (fast Xeon CPUs), and a maxed out Macbook Pro (2016).

I shot 4k anamorphic MLV using an EOS M (non-stretched resolution is in the ~1400 pixels range, can't recall it exactly now as I write this as i don't have the files with me), in 10bit, 1:39 ratio. So I did not even shoot in 14bit, just 10bit. I DID NOT process the files in any way. Just a straight export to prores (using the default ffmpeg settings as provided by mlv app). It was *very slow* on BOTH my computers. So, yeah, explain that. Why is it so painfully slow here while you claim that is real time for you?!?

The mlv files don't playback realtime on the mlvapp either btw.
The statement is very true. See: just reading a MLV is for nothing. As soon as you see an image, it MUST BE PROCESSED! And this needs time. MLV data can't be viewed without processing, because it is RAW!
Download the code and disable processing in code just after reading, and you'll see that you'll get more than 25fps with that - even on a 10 years old computer. The problem is, nobody needs MLV reading without processing or transcoding, because you can't do anything with that.
What bitdepth you use don't cares at all. Processing is always done in 16bit.

Share Your Videos / Native 4K (UHD) footage from EOS 5D Mark III
« on: August 09, 2020, 07:10:08 PM »
This is the next preset of Danne's build I wanted to give a try on the Canon EOS 5D Mark III. The UHD mode records at 3840x1536 @23.976fps. At 12bit I got a minimum of 15sec, but often way longer than that. The crop factor should be somewhere around 1.5x, so you get a bit more sensor area than APS-C and again a bit more than you get with 4K from EOS 5D Mark IV (non RAW).

The biggest disadvantage with that mode is, that you have to disable the liveview while recording. Otherwise the cam will stop after 2 or 3 seconds. This means you have to plan your shot very well before recording. I tried to do so on my Glidecam - difficult but possible.

There was no rescaling done - this is the original resolution coming from the camera! And the footage was not sharpened.

Gear used:
Canon EOS 5D Mark III
Canon EF 16-35mm f/2.8 II USM
K&F Variable ND8-ND128
Audio-Technica Pro-24 (modded)
Glidecam HD-2000
Lexar 1066x 32GB
SanDisk Extreme Pro 128GB 170MB/s

When comparing this with my last small test video in 5.7K anamorphic mode, my conclusion is:
5.7K looks more cinematic to me, while UHD is without doubt sharper (but no idea if you see it, because of youtube compression). 5.7K mode uses the full sensor width and has a nice field of depth. In UHD mode you'll have to focus very very well and it shows the smallest weekness of your lens.

Do you have any idea what you're talking about? The encoding takes less than 25% of the time. The "reading" of an MLV is faster than real time (for 25fps), even on a 10 years old computer with HDD. MLV contains RAW data. You can try sending RAW data to the encoder and tell me what you see. Wait... I can also tell you: You don't see anything - maybe digital noise with a little luck.
RAW data must be processed! The comparison to FCP is misleading, because FCP can't process RAW data (reasonably or at all).
The way things are now, it makes ML not serious at all, not even for fun. In fact, it takes all the fun away.
If RAW processing is too slow for you and takes away the fun, you shouldn't use it. You'll have to decide between quality and speed.

Also, the speed of writing on the memory is limited in the 6D, it is only about 40MB, while the 5D iii reaches 90MB
Is that correct
Depends on what we are talking about. 5D3 reaches ~130 MB/s when writing on CF and SD at the same time. For the 6D it also exists a SD overclocking, but for the exact numbers please read the corresponding threads.

Camera-specific discussion / Re: Canon EOS M
« on: July 28, 2020, 10:31:00 PM »
I now recorded a 70GB clip (33752 frames) - the cam aborted recording there, more was not possible. I can open the file without any problem in MLVApp and process it. MLVApp needs around 700MB RAM and ffmpeg 800MB RAM. So no need at all for 8GB+ RAM. I don't think there is a difference for even bigger clips, because the clips are processed frame by frame. The used RAM is constant for the entire processing. The limit currently is at 2^64 frames... I think that is enough.
I strongly recommend to enable MAPP files in MLVApp for such big files!

Pages: [1] 2 3 ... 66