Canon EOS 7D RAW Video?

[Marsu42]

I wish they'd pool the resources, and concentrate on the 5D3, and get it 100% ready for "prime time" and bring that one at least to Beta

I wish they'd concentrate only on the 60d, because that's what I have :-p ... remember ml started out and is still based on exploring, hacking and bringing more features and better usability to the cripped xxxd and xxd camera bodies - most ml enthusiasts and contributors only have the opportunity to see a camera of the 5d3 price range thorough a store window.

[cayenne]

5d3's is pretty solid as is, the interface could be better but i think ML as is is kind of fiddly so Im not really sure how much more you can expect out of it. EOSHD's guide to getting the firmware working is all you really need, the only major searching might be now that the link to the fw in his guide could be out of date.
All in all im not too worried about the flags or bricking it, I would expect its a minor fix for canon if something were to go wrong(as its just software), unlike getting moisture into it or damage from being dropped.

Well, the parts about installing ML itself are pretty straight forward, but getting to the parts where you have to find the nightly builds, get those parts separate, put them in xyz folder, do the hokey pokey ....is where I get lost.

And also, it the workflow for processing isn't that well laid out IMHO...rather than having it in a forum post that can get miles long with updates, they should have one page, that is updated to always reflect the latest instruction set with proper links, etc.....not only for installing but for post processing.

I'd like to see all files, including the RAW video parts be put into one firmware update rather than having to get the RAW video parts separate....and have a better spelled out post process.

Having to read a bunch of parts all separated out is the sign of true alpha and I'd like to see a more finished product which fromwhat I have read, should be possible for the 5D3....

C

[cayenne]

I wish they'd pool the resources, and concentrate on the 5D3, and get it 100% ready for "prime time" and bring that one at least to Beta

I wish they'd concentrate only on the 60d, because that's what I have :-p ... remember ml started out and is still based on exploring, hacking and bringing more features and better usability to the cripped xxxd and xxd camera bodies - most ml enthusiasts and contributors only have the opportunity to see a camera of the 5d3 price range thorough a store window.

Of course, part of my wants is that I own a 5D3. However, my reasoning was that, currently the 5D3 is the best supported, has the best video and apparently most stuff is working on it, giving the impression that if they concentrated on it, for maybe 1-2 weeks, they could call it a version ready build and put that out, and then, concentrate on the next best supported camera and put that out...rather than try to have all of them working at once.

I get the impression at this point, that the lowest common denominator is holding up the cameras that are closer to working. I'd say get those out the door in a more beta or release ready form, and then go back and work on the camera platforms that are really needing much more work.

C

[Otara]

Autoboot was solved on the 7D a couple of weeks ago and is now very simple.

I guess we have pretty different ideas about what 'very simple' means.  That post is from yesterday, a more detailed and helpful one is here:  http://www.magiclantern.fm/forum/index.php?topic=7464.0

Its prefaced by "this is a guide only for if you want the VERY EXPERIMENTAL and POTENTIALLY DANGEROUS autoboot version of the beta 2".  Partly bum covering of course, but not complete hyperbole.

Things are progressing quickly and theres some very promising stuff happening, but also it means levels of testing in the wild are going to be much lower.   

Otara

[thelebaron]

The version I have includes the raw module in it, there wasnt any extra hunting for it on the forums. But I agree, their forums are pretty terrible and the way they lay out things is rubbish over there

[crazyrunner33]

Of course, part of my wants is that I own a 5D3. However, my reasoning was that, currently the 5D3 is the best supported, has the best video and apparently most stuff is working on it, giving the impression that if they concentrated on it, for maybe 1-2 weeks, they could call it a version ready build and put that out, and then, concentrate on the next best supported camera and put that out...rather than try to have all of them working at once.

I get the impression at this point, that the lowest common denominator is holding up the cameras that are closer to working. I'd say get those out the door in a more beta or release ready form, and then go back and work on the camera platforms that are really needing much more work.

C

I respectfully disagree. While I'm not a developer, I feel that the other cameras are not what's holding the 5D back in development and that 1-2 weeks being focused on the 5D alone will not bring a RAW beta or release for the 5D Mark III. I think the 5D is not being slowed down, or by much by the other cameras because a lot of the developers that are working on porting the other cameras are doing so because they own them. This isn't stopping the developers who are focusing on the Magic Lantern source.

It's also hard to make a dedicated beta or full RAW release just for the 5D by focusing everyone on it for one to two weeks because the TAW video is seeing improvements and changes every day. I personally do not want to see a RAW release until the developers and community are happy with the new MLV video format that they're currently developing. If you want to try out Magic Lantern's nightly then you'll have to follow the instructions or purchase EOSHD's guide, DVXUsers also has a guide that's free.

As for post production, it's easy. I find that it's best to use Davinci Resolve since it was designed for video and allows for a speedy workflow. Currently you shoot your RAW video, copy over the RAW files and run them through raw2dng which creates DNG folders. These DNG folders are all you need, Resolve will recognize these folders as video clips and you'll be good to go.

[DFM]

The first builds were complicated as we had to compile the source from dailies scattered about on various sites, now everything is handled by EOSCard - if you can update your firmware with the camera menu and have a CF card reader, you can install ML (yes, OS X users have a couple more hoops to go through b/c EOSCard is Windows-only).

The biggest risk is if you interrupt the firmware update process, that's the cause of most "brick" reports but even that is usually fixable. You have to be really pushing things to do any permanent damage (in theory some of the super-high-bitrate video modes could overheat the chips, and there's a theoretical issue with dual-ISO affecting the buffers but nobody's seen it happen). Is it as stable as Canon's firmware? No, hence the 'alpha' tag - but realistically the bits that work are working fine, and the bits that aren't (such as continuous RAW) are not enabled in the builds that most users will be downloading. A few modules come and go as they're fine-tuned (ETTR, etc), but the core stuff like zebras and histograms are as reliable on the 7D now as they are on the other bodies. If you only want to shoot raw video it's not in any way ready for you yet and I wouldn't bother installing it. If you're shooting stills or H.264, even something like peaking or the Magic Zoom window could be a game-changer.

If you don't want to risk setting the autoboot flag, you can still use the original alpha 2 .FIR file, and 'update' your firmware each time you turn the camera on. That version does not make any permanent changes to the camera whatsoever.


I've been using ML-7D on commercial shoots for months, never had a problem. If ML itself errors out, remove the battery. Even with the bootflag set, if you want to go back to vanilla Canon firmware at any time just use a non-bootable CF card without the ML files on.

Autoboot was solved on the 7D a couple of weeks ago and is now very simple.

I guess we have pretty different ideas about what 'very simple' means.  That post is from yesterday, a more detailed and helpful one is here:  http://www.magiclantern.fm/forum/index.php?topic=7464.0

Its prefaced by "this is a guide only for if you want the VERY EXPERIMENTAL and POTENTIALLY DANGEROUS autoboot version of the beta 2".  Partly bum covering of course, but not complete hyperbole.

Things are progressing quickly and theres some very promising stuff happening, but also it means levels of testing in the wild are going to be much lower.   

Otara

[DFM]

ML now has continuous RAW video on the 7D 

Max resolution in non-crop mode is 1152x622 at 24p or 30p, and 1152x464 at 60p*. It's aliased as you'd expect from line-skipping, but it's not too bad. You need a 1000x card to get the maximums, but people are getting 1152x576 continuous with a 400x Transcend and on a Sandisk Extreme 60mb/s mine's barely breaking a sweat.

*Note that the 7D builds don't have frame rate override, so 24p = 23.976 etc.

Using 5x crop mode (press the LiveView zoom button) you can remove the aliasing, at the expense of a very narrow FoV (a 50mm lens gives you about 100mm width at 1 metre). 10x crop mode doesn't work at all.

All this is still well below the max possible write speed of the 7D so the limitation at this point is in memory-wrangling. Full-HD is not an option for non-crop mode as the sensor data just isn't there, but in 5x crop mode you can in theory hit 2520x1200 (nobody's got that to work continuously - yet!)

Get the raw_rec enabled ML files from http://www.magiclantern.fm/forum/index.php?topic=3974.msg65694#msg65694