MLV App 1.14 - All in one MLV Video Post Processing App [Windows, Mac and Linux]

Started by ilia3101, July 08, 2017, 10:19:19 PM

Previous topic - Next topic

0 Members and 5 Guests are viewing this topic.

escho

Servus masc

1.
My mlvs are quite big. 4 GByte and more... The last one had 7,6 GByte.

2.
hankYou. I love doing this. I wrote this guide for me. And than I thougt: Why not sharing it?

3.
This is difficult. I tried using qt-creator in debug-mode, but no crash here. All is fine! Using qt-creator normal: crash. And compiling outside qt-creator: crash. I´m a bit lost at the moment. I will search the commit, the crash comes in, later.
Would be interesting, whether the crash exists on Ubuntu too. So waiting for boucyball :)

https://sternenkarten.com/
600D, 6D, openSUSE Tumbleweed

masc

1. Ahhh... you mean when importing just one big file?! Yes... that is still a problem. But the GUI gets no information about the importing process and progress. The only I can do is a statusbar which file from how many files is imported.

3. Did you try make clean, and then again qmake and make? Sounds strange...
5D3.113 | EOSM.202

escho

1.
Yes. It´s always one big file, I´m importing. So, if the gui gives no progrssinfo, let´s forget it.

3.
I always compile a clean project. Sometimes I´m lazy. So I wrote a nice little bash-script for me to do this (delete all, clone git, make clean, qmake, make and some more...). And when I do this stuff manual, mlvapp crash too.

Since I had a bigger hardisk-crash this week, I´m working on a brand new opensuse now. But with my old opensuse ,before the hd died, I had the same crash.
https://sternenkarten.com/
600D, 6D, openSUSE Tumbleweed

escho

I went back to this commit:
"fixed compilation error saying type 'uint16_t' is not defined"
Same crash

Shoudn´t we go to github-issues with this problem?

edit:
Since my script write a compile-log: Maybe this implicit declaration is interesting:

g++ -o filter.o ../../src/processing/filter/filter.c
../../src/processing/filter/filter.c: In function 'initFilterObject':
../../src/processing/filter/filter.c:41:24: warning: implicit declaration of function 'fmemopen'; did you mean 'freopen'? [-Wimplicit-function-declaration]


nochmal edit:
created an issue on github: https://github.com/ilia3101/MLV-App/issues/74


https://sternenkarten.com/
600D, 6D, openSUSE Tumbleweed

Justus

I have some difficulties with the App. I came back to Magic Lantern after a few month of not using it. Before I used raw2cdng, which didnt work for me with lossless 10bit recording now.
This app did work perfectly but I am missing the sound in Davinci. I have sound in the app and can playback the clips with sound. Also I have sound when I preview clips in the Media Storage in Davinci. As soon I pull the clip from Media Storage into the Media Pool I lose the sound. It still shows a sound clip in the Edit Tab but there is nothing to hear and no waveform.
Maybe you guys know whats going on.

masc

How did you export? CDNG? There was a bug with timecode, which is fixed now. You will get it with the coming v0.14 (hopefully) in the next days. Or you compile yourself and you'll have it faster ;)
Workaround for now without compiling anything: add the wav manually to Resolve.
5D3.113 | EOSM.202

Justus

Oh okay, thank you. Yeah, CNDG Lossless. I will propably wait then. Changing the wav in a few hundred clips is nothing I am interested in^^

bouncyball



masc

Yay... we finally got it. Thank you all for reporting bugs and wishes.
New in v0.14 alpha:
- Faster playback and export (the more cores, the more you'll feel the difference)
- Added processing filters for film emulation powered by neural networks :ghost:
- New actions and shortcuts for better user experience
- MLV export (Fast Pass, Compressed, Averaged Frame (for darkframe creation), Extract Internal Darkframe)
- MLV darkframe subtraction (external / internal)
- Autodetection for focus pixels and vertical stretching
- Stretching CDNG export frames
- Vector scope
- Added markers for over-/underexposed in histogram
- Timecode label / Duration (edited clip) label
- Audio only batch export
- Bugfixes
Have fun!

5D3.113 | EOSM.202

ilia3101

Best release yet!

AI (neural network) powered film emulation filters!!! <-- Maybe a first in open source image processing software! (gonna overtake adobe at this rate)
Much better performance in export and playback (which is often realtime now)
All of the other problems complained about should be solved.
+ lots of other stuff <3

How I feel about this release.

andy kh

did a quick test on my windows pc. it works wonderful. love it
5D Mark III - 70D

bouncyball

@Ilia3101

Hey why you dumped the cocoa version from THIS page? It had 200+ downloads anyway! :)

Danne

This version is bad ass! Briefly checked it but had no darkframes available. Tell me since I am lazy over here, will the darkframe subtracion function work with multiple darkframes with various sizes or is one at the time and applies to all corresponding MLV files?


Filter function, awesome:

masc

You can have one darkframe per clip. You can use a MLV for that, or load the darkframe into your clip.
5D3.113 | EOSM.202

bouncyball

Yes one DF per clip.

If you export MLV with dark frame loaded, exported MLV will include the dark frame internally and loading this MLV after that will activate the 'Int' button.

It is possible to have one internal and one external dark frame simultaneously and switching between them (the effect of subtraction is seen straight away)

Also you can cut MLV including audio and it will be in sync and time code in resolve also will be correctly shown (after DNG export).

MLV export has 4 options (no raw correction is done):
1. export MLV untouched (fast pass)
2. export compressed MLV (only if original was uncompressed)
3. export averaged frame MLV (dark frame)
4. extract existing internal dark frame from MLV and save as a separate MLV.

With full compatibility of mlv_dump's averaging/subtraction.

As for filter option, there is an endless possibility to add more :)

mario1000

Hello dear developers,

congratulation and a very big THANK YOU for this new version. Everything works very nice exept one thnig: when i try to import a MLV file for dark frame substraction the program craches reproducible. I am using windows 8.1.  The only error message I get is that windows is looking for a solution and I am asked to close the program. Any ideas? Thanks in advance

bouncyball

Which version of MLV App - 32bit or 64bit?

Upload that MLV somewhere and I'll take a look :)

mario1000

Quote from: bouncyball on February 15, 2018, 05:44:10 PM
Which version of MLV App - 32bit or 64bit?

Upload that MLV somewhere and I'll take a look :)

I tested it with both versions (32 and 64 bit) and get the same result.
Here you can download my test files (video file and  the darkframe file):

https://ufile.io/3pa2v
https://ufile.io/3bauu

Best regards
M.

bouncyball

You recorded but did not average all frames of m15-1742.mlv.

How to average it:
1. just load m15-1742.mlv to MLV app (not as dark frame)
2. press 'Ctrl-,' or the export settings button
3. select codec MLV and the option below - Averaged Frame, then close export settings dialog
4. export MLV (name it the way to recognize that this is a dark frame, e.g. m15-1742_Dark_Frame.mlv)
5. delete m15-1742.mlv from session
6. load m15-1741.mlv into session and load m15-1742_Dark_Frame.mlv as a dark frame
7. play with Off/Ext, you should see the difference immediately
8. optionally, you can now export m15-1741.mlv from MLV App and it will include your dark frame internally, after loading exported MLV the Int button will be activated and you can play with Off/Int to see difference
9. Phew... the end :)

Anyway thank you for the feedback, the app crashes because it can not import compressed MLV as dark frame (my bad, I should've added check for it) otherwise as far as MLV, loaded as dark frame, is not compressed you can subtract any MLV from any MLV, again as far as video resolutions match :)

regards
bb

mario1000

Thanks for the help, now it works! Indeed it makes a very big difference using the darkframe.

regards
M.

Kharak

I just did some initial testing as i was interested in this neural network film look. What is the difference between this, dare I say, A.I.? And a LUT?

And should i be worried that it will become aware of itself and send robots in to the past?

Joke aside, where are these film looks/emulations from? It reminds me of filmconvert's looks, where in it turns everything brownish.

Also i could not see lossless mlv's (from one of the first versions) 59'976 fps in the import window, but when i dragged and dropped them from folder there was no problem.

Win 64 version.
once you go raw you never go back

ilia3101

There shouldn't be much of a difference between this and a LUT in terms of what it does(just does it a lot differently and slower :D), and it was even trained on some samples that someone I know filmconverted(shhhhh) WHOOps I mean real film

LUT support might be added at some point too, they seem quite simple actually.

The "AI" only has 3 inputs and 3 outputs... I would imagine its difficult to do any damage with just three floating point variables. Don't worry :D

Kharak

You should talk to hyjalinim or how you spell it, he has some really nice examples of Ektar 100 film. I find his Ektar 100 LUT requires a very specific exposure to yield nice results, perhaps that is something the A.I can learn to do for us? Just a suggsstion.

I could chime in some examples myself, if need be. Have some pictures that nailed the exposure for that Ektar 100 LUT.  But I suppose its better to talk to the man who made the film emulation. And i dont know how the A.I learns exactly.

Anyways, keep up the good work. Its very nice that there is so much work done on the post-processing side of ML.
once you go raw you never go back

bouncyball

Quote from: Kharak on February 17, 2018, 01:02:21 PM
Also i could not see lossless mlv's (from one of the first versions) 59'976 fps in the import window
Could you upload the sample? I'm curious about it :)