Is MLV all that it's cracked up to be?

Started by Midphase, March 06, 2014, 08:42:48 AM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

budafilms

Have someone two 5D Mark III to put one in RAW and the other MLV at same time, equal shoot, same lens, etc?

In my case, if .RAW have sound, it's the same.

I did a Full Movie - about 4 month - in RAW without sound reference and it's a pain...

but it's true .RAW it's more stable when you shoot 8 hours a day.

I'ts my experience!

Audionut

Quote from: Midphase on March 07, 2014, 06:08:03 AM
I wish they'd

Ah, wishful thinking.  I wish I had $100,000,000.
Wishing, and wanting, and expecting, doesn't make things happen.

I can assure you, that adding MLV support to existing software, is magnitudes easier then developing an entire NLE.

Have you started writing letters to Adobe et all, yet?

andy kh

@budafilms yeah its a pain in the ass without sound reference so i never shoot raw and was never interested. for making films i prefer mlv since i can sync up audio which i recorded with my external recorder and with the reference audio i can do ADR.
5D Mark III - 70D

Markus

I also find the question in this thread relevant. The following argument reminds me of a reaction that is all too familiar for one who has been in a lot of feedback situations making/creating films and videos. It's not easy to be open when you worked your ass off on something in a particular direction. But with time you learn you have a lot to gain if you can be open and listen to the curses in the church. 

By the way, the mlv module seem to be getting really good lately but the robustness, speed, instant start, and simplicity of the standard raw format is a set of powerful strengths!

Midphase

Quote from: Audionut on March 07, 2014, 06:51:43 AM
Ah, wishful thinking.  I wish I had $100,000,000.
Wishing, and wanting, and expecting, doesn't make things happen.

I can assure you, that adding MLV support to existing software, is magnitudes easier then developing an entire NLE.

Have you started writing letters to Adobe et all, yet?

Why the attitude? Is it so wrong to communicate what I think are shortcomings of ML raw or some of the apps? Why shouldn't you be interested, even happy to get useful feedback from guys who are trying to use these tools in real-world situations?

Perhaps something got lost in translation, but I never asked for ML to develop its own NLE. I merely asked for ML to develop a more user friendly way to get the data from the camera into an NLE. Without a solid back-end, all of the wonderful work that is being done on the in-camera raw video is kinda moot.

P.S.

Yes, I have repeatedly put in requests to both Adobe and Resolve for .raw and .mlv support.

Audionut

Quote from: Midphase on March 07, 2014, 08:06:43 AM
even happy to get useful feedback from guys who are trying to use these tools in real-world situations?

I must have missed the useful feedback.  Care to provide a link?


Quote from: Midphase on March 07, 2014, 08:06:43 AM
Without a solid back-end, all of the wonderful work that is being done on the in-camera raw video is kinda moot.

In your own opinion, or the opinion shared by few.  Just because Midphase has a problem with CMD lines, doesn't make the current implementation, moot!

Dispense with the over exaggerations please.

Midphase

I take back what I said about you and I sharing a drink and getting along. You sound like a really angry person with a very thin tolerance for anyone who does not agree with you.

Audionut

So, no link then?

No need to get personal.  Discuss the matter in a manner that doesn't rely on over exaggerations.

Midphase

If you have nothing useful to contribute to this thread, could you please stop wasting your time posting on it? I'm sure there are tons more things out there that annoy the crap out of you that you could be tending to.

Regarding feedback:

I mentioned that Thomas' RAWMagic nailed a good solution. Unfortunately no such app exists with support for .mlv. Command-line conversion tools are simply not a real-world way to manage this data. Particularly not on a set. If command-line was such a great way to get things done, then we'd all be still using MS-DOS.

I mentioned that MLVMystic is on the right track, but I also outlined its limitations and what should be addressed to make it a robust option. I also mentioned that none of the current apps for the Mac output CDNG from .mlv, I think this is an omission that should be corrected as Adobe is more likely to support CDNG than DNG in the future.

If none of this sounds like useful feedback, please refer to the first paragraph of this post and kindly oblige with my request.

a1ex

Post-processing:

These post-processing tools are not developed by ML, other than some reference implementations (raw2dng/mlv_dump), which are command-line tools to keep them simple.

The programs you are talking about are developed by other authors (not by ML devs), so their authors will most likely welcome your feedback.

Recording modules

Judging from user feedback, my impression is that raw_rec is pretty much stable (except for some small quirks), but minimalist (no sound, no metadata), and mlv_rec is still work in progress, has the big advantage of audio recording, but write speeds are a bit inconsistent, and being quite a bit more complex, there are more things likely to break. However, it's designed to be extensible (friendly for future development), which can't be said about the RAW format.

An example regarding extensibility: suppose you may want to do some FPN correction, and for that you need to record optical black data (either complete or summarized somehow) or some sort of bias or dark frames. With the RAW format, you have no other choice but creating external sidecar files. With MLV, you can simply add this optical black data to MLV spec and implement it. The converters that don't know about this new feature will simply ignore it gracefully (this is allowed by the design of MLV format); the ones that do will use this correction data and output better images.

Once mlv_rec will be better in all aspects, raw_rec will most likely go away. For now, my impression is that it isn't there yet, so I still keep the old raw_rec in the nightly builds.

The cause of pink frames is not yet fully understood, and recent versions of mlv_rec were reported to have fewer pink frames at least on 550D. On other cameras, raw_rec may be better, I don't know.

If you report things like one module seems to be more solid during long usage, a useful feedback should include some statistics. For example, say you've recorded a bunch of footage at settings XYZ, and got 15 pink frames out of 12345. Then, say you've changed setting Y and got 30 pink frames out of 1234. This feedback is much better than vague reports about intermittent failures, but providing a way to reproduce a bug is even better.

I use the camera primarily for photos, so I didn't actually use the raw recorders in the last few months.

Quote from: Markus on March 07, 2014, 06:43:54 AM
Is a working sound module for standard raw rec impossible or just not something anyone is interested in putting time in developing?

Probably quite easy, but since the MLV file format is extensible, I don't expect further development on raw_rec (other than maybe some basic maintenance), but anybody is free to give it a try.

I don't exclude a lite version of mlv_rec at some point in the future, rewritten with performance and robustness in mind, rather than feature-completeness. But I'm confident that MLV as file format is here to stay.

Markus

Quote from: a1ex on March 07, 2014, 02:04:57 PM
Post-processing:

Probably quite easy, but since the MLV file format is extensible, I don't expect further development on raw_rec (other than maybe some basic maintenance), but anybody is free to give it a try.

I don't exclude a lite version of mlv_rec at some point in the future, rewritten with performance and robustness in mind, rather than feature-completeness. But I'm confident that MLV as file format is here to stay.

Thanks for a good and extensive answer. Interesting as you describe future capabilities of mlv and I also think it is here to stay. Actually I don't really think anyone here doubts that. However a snd module for raw-rec would be a great tool in the toolbox. Would be greatfull to anyone who would pick that up. It would give us a complete stable solution whilst mlv development continues.

1%

raw_rec with sound would need a bunch of changes, I don't think its as simple as adding the hooks from mlv_snd... mlv marks buffers specifically for sound and records a different way. if you add the mlv writing/buffering methods to raw_rec then you'll have mlv_rec and might lose the speed advantage.


mixmastermike

Quote from: a1ex on March 07, 2014, 02:04:57 PM
Post-processing:

These post-processing tools are not developed by ML, other than some reference implementations (raw2dng/mlv_dump), which are command-line tools to keep them simple.

The programs you are talking about are developed by other authors (not by ML devs), so their authors will most likely welcome your feedback.

Recording modules

Judging from user feedback, my impression is that raw_rec is pretty much stable (except for some small quirks), but minimalist (no sound, no metadata), and mlv_rec is still work in progress, has the big advantage of audio recording, but write speeds are a bit inconsistent, and being quite a bit more complex, there are more things likely to break. However, it's designed to be extensible (friendly for future development), which can't be said about the RAW format.

An example regarding extensibility: suppose you may want to do some FPN correction, and for that you need to record optical black data (either complete or summarized somehow) or some sort of bias or dark frames. With the RAW format, you have no other choice but creating external sidecar files. With MLV, you can simply add this optical black data to MLV spec and implement it. The converters that don't know about this new feature will simply ignore it gracefully (this is allowed by the design of MLV format); the ones that do will use this correction data and output better images.

Once mlv_rec will be better in all aspects, raw_rec will most likely go away. For now, my impression is that it isn't there yet, so I still keep the old raw_rec in the nightly builds.

The cause of pink frames is not yet fully understood, and recent versions of mlv_rec were reported to have fewer pink frames at least on 550D. On other cameras, raw_rec may be better, I don't know.

If you report things like one module seems to be more solid during long usage, a useful feedback should include some statistics. For example, say you've recorded a bunch of footage at settings XYZ, and got 15 pink frames out of 12345. Then, say you've changed setting Y and got 30 pink frames out of 1234. This feedback is much better than vague reports about intermittent failures, but providing a way to reproduce a bug is even better.

I use the camera primarily for photos, so I didn't actually use the raw recorders in the last few months.

Probably quite easy, but since the MLV file format is extensible, I don't expect further development on raw_rec (other than maybe some basic maintenance), but anybody is free to give it a try.

I don't exclude a lite version of mlv_rec at some point in the future, rewritten with performance and robustness in mind, rather than feature-completeness. But I'm confident that MLV as file format is here to stay.

First of all thanks for Devs getting involved in this thread , there are answers that I would like to pose.

RAW_REC to me seems pretty stable too, however for the application I need I need sound as I only have 2 people involved in the project I would love to record in RAW.

MLV_REC is conceptually great, however the recording speeds/buffer performance is very up and down. Im working on a cooking show for web distribution 24p is not the look I need at all. the motion blur is too filmic, 30P would be ideal. If card spanning can return to full HD 30p with MLV and MLV_SND I will be a very happy man. I will also feel that ML has achieved a a fully fictional raw video camera with sound, a real video camera so to speak, a complete functional unit to rival video cameras. In post this means I can use Pluraleyes to auto sync external professional sound, reducing post processing time.

Is it too early to say if 30P full HD will be possible?

Is there any way you could explain what is going on with the write speeds to a laymen, Im big on solving problems but just unfortunately not coding ones. I built a shoulder rig to film the show and be able to start and stop recording, and audio recording all while the camera is on my shoulder. I built the rig with the fact a 5D mkiii might not acutaly be the camera to do this project, so the baseplate can be changed and as such the camera. However the canon sensor is amazing and produces superb skin tones and colors in my opinion much better that say a Gh3 or Black Magic (Subjective I know, lets not start another fight here)

I would just love to see this project done on MLV_REC and MLV_SND at 30P does anyone think that is possible in the future?




Midphase

Thank you a1ex for your very informative reply. Information from posts like yours and 1%'s goes so much further in clearing up some confusion that (at least for me) has crept up about the usefulness of switching  to .mlv



Pierluca72

My first post here... First of all thank you guy for the amazing work. Actually I love .mlv on Mac. Workflow isn't difficult at all. I created an automator workflow using the dump and command.. one click and I have my dng and wav. No preview on the mac? yes there is one, is Davinci lite (free). It can play in real time (almost) 22fps the dng files. However there is one thing that I'm struggling with and I would really appreciate some hints here. The audio track is slightly longer than the video track. Am I doing something wrong? I use the same fps settings in davinci as the clip. Thank you in advance

Pierluca72

Quote from: Pierluca72 on March 08, 2014, 12:13:24 AM
No preview on the mac? yes there is one, is Davinci lite (free).

I just realized that previewing a dng file is not the same as previewing a .mlv. I still have to convert to dng first so.. I take that back.. But still please a hint for the sync

Frank7D

Quote from: Pierluca72 on March 08, 2014, 12:13:24 AM
The audio track is slightly longer than the video track. Am I doing something wrong? I use the same fps settings in davinci as the clip. Thank you in advance

All mine are that way too, yet they are in sync nonetheless.

daisermac

MLV works perfectly for me, using nightly build Febr. 15th on 5Dm3. No card spanning, good old K-Bay 64GB 1000x, recording 1080p25. Not a single corrupted frame since then (had a lot with the first builds from december). I am even using HDMI, which is said to have bad effects on write speed. One in ten recordings stops because of slow speed, but I can live with that. Camera hangs if I switch of HDMI before waiting for camera shutdown - I can live with that too. Sound is a bless. MetaData (e.g. for automatic distortion correction) is a bless. MLRawViewer is a great software. Conversion with the shell script works without problems.

Pierluca72

Quote from: Frank7D on March 08, 2014, 12:43:33 AM
All mine are that way too, yet they are in sync nonetheless.
OK thanks. They are in sync actually. I just have to trim off the audio track so that it matches to the video. Has anyone found a quick way of doing this when you work with many clips?

ted ramasola

In resolve, once you have dragged all the dngs and wav files into the media pool, right click the root and right click auto sync based on time code. in edit choose new timeline unselect create empty timeline. all dngs will be linked and trimmed. Though in cases of longer clips, the wav seemed to show no waveform halfway, I manually linked those problematic clips only.
5DmkII  / 7D
www.ramasolaproductions.com
Texas

andy kh

Quote from: Pierluca72 on March 11, 2014, 03:56:08 PM
OK thanks. They are in sync actually. I just have to trim off the audio track so that it matches to the video. Has anyone found a quick way of doing this when you work with many clips?

i will tell you the best way to deal with so many clips if you are using windows. simply convert to cdng using raw2cdng and import all those clips in premiere pro and wala..all your audio and video clips are in sync. u neednt do any manual sync
5D Mark III - 70D

budafilms

Quote from: ted ramasola on March 11, 2014, 04:22:41 PM
In resolve, once you have dragged all the dngs and wav files into the media pool, right click the root and right click auto sync based on time code. in edit choose new timeline unselect create empty timeline. all dngs will be linked and trimmed. Though in cases of longer clips, the wav seemed to show no waveform halfway, I manually linked those problematic clips only.

Hi Ted, I was trying that, and no result!
Have I put an specific module for time cod? or activate something?

ted ramasola

It must be how the dngs were extracted. I'm using tonybeccar's windows based extractor. http://www.magiclantern.fm/forum/index.php?topic=10198.msg98334;topicseen#msg98334

Current one is buggy, someone posted a link to an older 1.6 version  that is working for me. :http://www.samuel-maurin.com/MLV-Converter-1.6.rar

What dng converter/extraction are you using?
5DmkII  / 7D
www.ramasolaproductions.com
Texas

budafilms

Quote from: ted ramasola on March 12, 2014, 03:55:20 AM
It must be how the dngs were extracted. I'm using tonybeccar's windows based extractor. http://www.magiclantern.fm/forum/index.php?topic=10198.msg98334;topicseen#msg98334

Current one is buggy, someone posted a link to an older 1.6 version  that is working for me. :http://www.samuel-maurin.com/MLV-Converter-1.6.rar

What dng converter/extraction are you using?

I use Apple!
I open RAW file with RawMagic or MLV with Mystic.
After that I import into DaVinci.
And the edition with Final X.

Sganzerla

I'm one guy who have done 8 videos using RAW video with my MKIII and I have the same concerns as Midphase.

Still don't know if I should stick with RAW for now, 75% of me say yes.