Tragic Lantern for EOS M

[a1ex]

Can the nightly builds also be compiled and zipped like Jerry was doing at bot fly before since I have no dev skills for compiling.

You were not aware of this page? http://builds.magiclantern.fm/#/

[gary2013]

no, i was always on this M thread and bot fly from Jerry. I will bookmark that page and watch it. thanx Alex and Jordan.

[funkysound]

The word was chosen by 1% more than one year ago.

The good news is that @jordancolburn already updated ML to 2.0.2 ( https://bitbucket.org/hudson/magic-lantern/pull-request/303/updated-eosm-to-run-on-firmware-v-202/diff ), and if all goes well, we may have official nightly builds for the EOS M soon.

Yeah ... a1ex - thanks for giving us users a hopeful sign after this long time of suffering. And another thanks goes to @jordancolburn - really hope all goes well!

[jordancolburn]

And another thanks goes to @jordancolburn - really hope all goes well!

Thanks, I got the basics for 2.02 working based on looking through 1%'s code.  He documented everything pretty well with his commit messages, so it wasn't too bad.  Once EOSM gets back into the nightlies, we should start a new thread since it will be missing any bug fixes 1% did and any features specific to TL.  I'd appreciate anyone's help who might want to point out missing features or bugs, so I can try to pull the most necessary stuff over first.

[funkysound]

Thanks, I got the basics for 2.02 working based on looking through 1%'s code.  He documented everything pretty well with his commit messages, so it wasn't too bad.  Once EOSM gets back into the nightlies, we should start a new thread since it will be missing any bug fixes 1% did and any features specific to TL.  I'd appreciate anyone's help who might want to point out missing features or bugs, so I can try to pull the most necessary stuff over first.

Sounds good to me. I think for video it´s most importend to have first the normal daily used stuff like zebra, focus peaking, 3x zoom, record restart, ... and to get rid of this f... shutter bug. With other words a basic version which could be more or less called stable before working on raw a.s.o. ...

[jordancolburn]

All the basics do seem to work for me.  Zebras focus peaking(the old style i think), 3x zoom (fantastic with the 22mm lens) and all that seem good. I think the stuff to iron out will be little bugs like with AF on, it seems to track faces through the ML menu and 3x zoom setting persists after turning the camera off, so switching to a non ML card, you get stuck in crop mode for video.

[cpreston]

I am so so glad to hear that somebody is trying to get a stable version of ML EOS-M up.  I am one of those people who chose Canon's because of Magic Lantern and things like magic zoom, peaking, false color, and zebras.  These are the things that are built into normal video cameras and help to make a Canon DSLR a viable choice for videography.  Thank you.

[gurayvarol]

Hi guys,
Thanx for all your hard efforts. I think my problem seems to be unique so I decided to have it written here. I have EOS M with 1.06 firmware and ML installed on a 64GB card. I am trying to get a constant timelapse but the camera becomes unresponsive after 2 or 3 days. The camera mentioned a shoot malloc when I was trying to format the card (keeping the ML)

Is it the same problem with the shutterbug that is being discussed in various topics or is it something new?
would installing 2.02 fix that?
Would changing the card to 32GB fix that?

I am running a test with the same features but on a 32GB card I would let you know the results in a few days but if there is anyone who has already tried it can save my time and would be appreciated

[meditant]

I hope you continue the development for EOS-M, i have magic lantern on my 5DIII's and it's very very cool !

I hope i can use magic or tragic lantern on my EOS-M !

Best regards and 1000 thanks for you good job

[gary2013]

...I'd appreciate anyone's help who might want to point out missing features or bugs, so I can try to pull the most necessary stuff over first.

I know, I have said it over and over again, but since you asked, I would like to see the audio headphone out to work on the M. 1% spoke a few times as if he was just about to get to working on it and that he felt it was doable. Having the ability to hear the audio before , during and after an interview is crucial along with seeing the video playback on the lcd of the M. The whole purpose, at least for me, was buying the M to use in making documentaries while being discreet and having light weight without having lots of extra gear to carry and setup. The  sound on camera is just fine for me. So, please try and get the audio out for headphones on the AV port as well as getting the audio meters to be visible before recording and during recording.

Most of us have agreed that we will never be able to get 1080p30 or 24p raw on the M considering its hardware limitations. I can use H264 but I would like to see ways to improve the codec to get the absolute best quality video at 1080p30. CBR? VBR? Setings? New controls?? Is there such a thing as getting ProRes thru software on this camera?

Are there any ways to eliminate or reduce aliasing and moire thru ML software? I know about Mosiacs hardware filter, but that is costly and sometimes not good for wide angle shots. A software  solution would be fantastic. I am wondering about not doing line skipping which I hear causes the problems. A new algorithm from raw to the encoder, maybe? Just guessing about all of this since I am not a dev.

How about some new way to open ML menus besides the current 2 finger swipe. outdoors on a bright day using a loupe or sunshade makes it a PIA to get to the touchscreen. I think Alex suggested awhile back a long press and hold the SET button? Some sort of single button press can maybe work.

[jordancolburn]

I know, I have said it over and over again, but since you asked, I would like to see the audio headphone out to work on the M. 1% spoke a few times as if he was just about to get to working on it and that he felt it was doable. Having the ability to hear the audio before , during and after an interview is crucial along with seeing the video playback on the lcd of the M. The whole purpose, at least for me, was buying the M to use in making documentaries while being discreet and having light weight without having lots of extra gear to carry and setup. The  sound on camera is just fine for me. So, please try and get the audio out for headphones on the AV port as well as getting the audio meters to be visible before recording and during recording.

Lightweight docs is why I got the M too, and audio will be one of the big things I'll look at.  I think audio meters before recording probably can't happen since the chip is only active once you start.  You can test levels in the Canon manual setup and see meters before recording, or just record a test clip to test the levels (this is what I do on the t3i). I am a little more concerned with trying to pick up all the bugs and stability stuff first, and 1% is sure to be better and faster for new feature stuff like that, but I'll give it a try to.

The EOSM seems perfect for grabbing quick interviews for Docs, a simple clip on lav with 1/8" connector can fit in a jacket pocket with the EOSM, 22mm lens and a backup battery and card, and you are always ready to grab a quick interview or story (on some recent trips, we always find that when people are most at ease and willing to share their stories, is when we're least prepared shoot and out to eat or something).

[R4dev]

Hi there! I'm new to the forum, and want to thank you all for your hard work.

I have a problem that I hope you can help with. I recently bought an EOS-M and couldn't wait to put ML on it. It downloaded fine and everything seems to work fine, except when I try to  shoot RAW video. It seems to start shooting, then after a second the camera restarts, without saving any files. It happens in crop mode as well. I'm using Transcent 32GB SDHC class 10 card and I tested with 1408x362 resolution at 2.35:1 ratio. Sorry if this problem has been resolved before, the topic is 89 pages long now and my search found nothing.

[funkysound]

The EOSM seems perfect for grabbing quick interviews for Docs, a simple clip on lav with 1/8" connector can fit in a jacket pocket with the EOSM, 22mm lens and a backup battery and card, and you are always ready to grab a quick interview or story (on some recent trips, we always find that when people are most at ease and willing to share their stories, is when we're least prepared shoot and out to eat or something).

That´s exactly how it is! As less official as possible gives the best results - that´s why we don´t use big cams very often for theses cases.

[jordancolburn]

Hi there! I'm new to the forum, and want to thank you all for your hard work.

I have a problem that I hope you can help with. I recently bought an EOS-M and couldn't wait to put ML on it. It downloaded fine and everything seems to work fine, except when I try to  shoot RAW video. It seems to start shooting, then after a second the camera restarts, without saving any files. It happens in crop mode as well. I'm using Transcent 32GB SDHC class 10 card and I tested with 1408x362 resolution at 2.35:1 ratio. Sorry if this problem has been resolved before, the topic is 89 pages long now and my search found nothing.

While I never experienced the camera restarting, I had issues with a class 10 transend card.  Try turning on FPS override to something like 8 fps just to test the stability and post workflow, and look for something like a sandisk 45 MBs card once you're ready to take the next step.

[AnthonyEngelken]

I can use H264 but I would like to see ways to improve the codec to get the absolute best quality video at 1080p30. CBR? VBR? Setings? New controls?? Is there such a thing as getting ProRes thru software on this camera?

This is what the bitrate controls are for; 3.0x CBR has always worked quite nicely for me and reduces a ton of compression artifacts in the shadows. I've even shot some stuff with that value cranked up to 5.0x without much trouble.

Are there any ways to eliminate or reduce aliasing and moire thru ML software?

Definitely, it's movie crop mode (2.7x zoom) which captures from the middle 37% of the sensor without skipping lines. That certainly narrows your field of view, but it's a 5K sensor. To take advantage of the whole sensor without line skipping, the camera would be shooting at a resolution the hardware couldn't keep up with, and that the transcode processor doesn't support.

These questions coming from you Gary are puzzling; you've done a ton of experimenting, and I thought you knew all of this already.

Just to make note of it here for Jordan, the ML features I use all the time are Bit Rate, FPS Override, Movie Crop Mode, Crop Marks, Histogram, Intervalometer, and everything from the exposure tab.

[gary2013]

This is what the bitrate controls are for; 3.0x CBR has always worked quite nicely for me and reduces a ton of compression artifacts in the shadows. I've even shot some stuff with that value cranked up to 5.0x without much trouble.

Definitely, it's movie crop mode (2.7x zoom) which captures from the middle 37% of the sensor without skipping lines. That certainly narrows your field of view, but it's a 5K sensor. To take advantage of the whole sensor without line skipping, the camera would be shooting at a resolution the hardware couldn't keep up with, and that the transcode processor doesn't support.

These questions coming from you Gary are puzzling; you've done a ton of experimenting, and I thought you knew all of this already.

Just to make note of it here for Jordan, the ML features I use all the time are Bit Rate, FPS Override, Movie Crop Mode, Crop Marks, Histogram, Intervalometer, and everything from the exposure tab.

Yes, I knew most of it. I wanted to see if i could push for some new info or maybe the devs could maybe expand on what we have. I should have said "besides the crop mode" for less/no aliasing so we can keep our wide angles. I have heard somewhere that possibly new algorithms would help eliminate aliasing and moire. We have to remember that some things a year ago or more were not possible and then the devs came up with new good things for us. I am now at 2.5x with CBR. I have tried higher, like up to 10x,  but it was not always stable. I see I get an average of 77 mbs using 2.5x CBR. Also on VBR.

[RavingRover]

While I never experienced the camera restarting, I had issues with a class 10 transend card.  Try turning on FPS override to something like 8 fps just to test the stability and post workflow, and look for something like a sandisk 45 MBs card once you're ready to take the next step.

I also do not get more than the buffer's worth of raw (*.mlv files)!  I use a 60w/80r MB/s Sandisk Extreme. I am attempting RAW at the default rez of 1408x792 @ 24fps.

I get as high as 139mbps compressed crop video via CBR 3.1x setting - high ISO.

HELP: I do not see how to change either the resolution or the fps, thus have not tried any other.
PLZ tell me how to change these someone.

Also, why can I not get at least a bit more RAW video?

Thanks for any help!!

[krisc]

Hi there! I'm new to the forum, and want to thank you all for your hard work.

I have a problem that I hope you can help with. I recently bought an EOS-M and couldn't wait to put ML on it. It ................ is 89 pages long now and my search found nothing.

People that have ML running on EOS M ..... Brand new camera, after initial tests wanted to get ML. Finaly I can get to "green screen" stage - so it boots and loads FIR.
On subsequent power up it stays there flashing green led for ever.
SDHC card is 32GB Transcend class 10 ( very fast, excellent on on Sony video cams and perfect on EOS M as well). Have also SanDisk SDHC 32GB class 10  but did not try it

Aby help will be appreciated .....    Kris

[AnthonyEngelken]

Yes, I knew most of it. I wanted to see if i could push for some new info or maybe the devs could maybe expand on what we have. I should have said "besides the crop mode" for less/no aliasing so we can keep our wide angles. I have heard somewhere that possibly new algorithms would help eliminate aliasing and moire. We have to remember that some things a year ago or more were not possible and then the devs came up with new good things for us. I am now at 2.5x with CBR. I have tried higher, like up to 10x,  but it was not always stable. I see I get an average of 77 mbs using 2.5x CBR. Also on VBR.

An alternate solution that may be plausible would be to capture in 5K and have the transcode chip downscale before storage. That being said, I understand that the black box is an enigma, and the M's buffer wouldn't likely put up with a 5K throughput. I don't even expect that the sensor is fast enough to deal with 5K capture at movie frame rates without ridiculous jello/roll. Even if all of this were possible, the transcoder's scaling algorithms might be shit and make your footage look terrible anyway.

That's a lot of ifs, and not even an ideal solution. What I would actually LOVE however is a high quality EF mount behind-lens wide angle converter to compensate for a 4.3x crop. Alas, nothing like that exists to my knowledge.

I wonder what a shorter EF mount adapter would do on the M beside screw with focus?

[RavingRover]

How do you change the RAW resolution and fps rates diff from defaults ?  (ie - do you need to change files or is it via ML menu?)

Thx!

[RavingRover]

An alternate solution that may be plausible would be to capture in 5K and have the transcode chip downscale before storage. That being said, I understand that the black box is an enigma, and the M's buffer wouldn't likely put up with a 5K throughput. I don't even expect that the sensor is fast enough to deal with 5K capture at movie frame rates without ridiculous jello/roll. Even if all of this were possible, the transcoder's scaling algorithms might be shit and make your footage look terrible anyway.

That's a lot of ifs, and not even an ideal solution. What I would actually LOVE however is a high quality EF mount behind-lens wide angle converter to compensate for a 4.3x crop. Alas, nothing like that exists to my knowledge.

I wonder what a shorter EF mount adapter would do on the M beside screw with focus?

Why would CDAF be bothered by a shorter EF mount?  CDAF is sensor based, and as long as it can communicate to the lens and change its focus mechanism, it keeps trying until the 'optimum' contrast occurs? Correct?  Same for PDAF, just a feedback system of AF.

The shorter EF mount would only effect the light 'cone' from lens, and in the case of EF to M, I would think it would improve but marginally 'crop' the cone a bit, no effect to AF, me thinks!

But a 5k video, not likely, I think the canon h/w & encoder h/w is not capable, so it all would be done via SD card memory - or post processing of RAW video.

Or are you suggesting still image processing at fps of 24 or something?

What is the highest bit rate for compressed video ppl have gotten from the M ? 

I have seen peaks of 140mbps (CBR 3.2x), but my card is I believe not the fastest around.

[AnthonyEngelken]

Why would CDAF be bothered by a shorter EF mount?  CDAF is sensor based, and as long as it can communicate to the lens and change its focus mechanism, it keeps trying until the 'optimum' contrast occurs? Correct?  Same for PDAF, just a feedback system of AF.

Auto-focus aside, wouldn't changing the flange focal distance affect focus in general? Beside focus distance not matching the lens' markings, you might also prevent a lens from focusing as near or as far as it previously could. Would be a cool experiment though; might buy one of the cheap after-market adapters on eBay and try modifying it. The light circle that hits the sensor only has to have a diameter of 9.48mm in movie crop mode, while with the adapter, EF lenses are producing a 43.23mm circle I think. Anyone that's a lens expert wanna do the math on this?

[RavingRover]

Auto-focus aside, wouldn't changing the flange focal distance affect focus in general? Beside focus distance not matching the lens' markings, you might also prevent a lens from focusing as near or as far as it previously could. Would be a cool experiment though; might buy one of the cheap after-market adapters on eBay and try modifying it. The light circle that hits the sensor only has to have a diameter of 9.93mm in movie crop mode, while with the adapter, EF lenses are producing a 43.23mm circle I think. Anyone that's a lens expert wanna do the math on this?

Sure there might be situations that 'infinity' or 'macro' may be affected but look at how on sensor CDAF/PDAF works, it has NOTHING to do with the lens beyond being able to control the focus mechanism. Correct?

This is a deficiency of the focus mechanism - ie it just can not be moved far enough NOT due the CDAF/PDAF control system not being able to tell it what to do - beyond of course the AF attempt 'time out' which already occurs!! AF is a open loop system simply due to the limited AF mechanism.

Also this problem of near and far will be dependent on each Len's capabilities, some lens will have the issue others will not, but if the distance of coupler/mechanism is too large ALL lens will be effected of course.

I believe CDAF has always been more accurate than PDAF just simply prior to last few years, where Panasonic and others of the 4/3 group got CDAF quite fast with little to NO hunting, it was way too slow compared to the 30yr old fast PDAF tech!

Part of the reason PDAF is not as accurate is that the AF sensor is dislocated and not on/at the sensor, but with Hybrid AF I/II, Canon put both CDAF & PDAF sensors on the pic sensor, but CDAF is how YOU do manual focusing NOT PDAF ! We see focus as highest contrast via sharp edges - CDAF.

Canon is now close but NOT leading in the CDAF focus speed, it will get there hopefully but it is a combination of CDAF controlling algorithms AND lens focus mechanism speed ! 

The 22mm M lens AF is far slower than the 18-55mm! The 22mm lens focus moves the complete group of lens elements, the 18-55/11-22mm move a smaller/lighter sub-group of elements! lens design!

Panasonic and 4/3 group had to optimize their lens for AF speed, and some are still pretty slow, as slow as the 22mm M lens or maybe worse!!

And yes it would be a fun test!
I believe overall PDAF is faster on more lens than is CDAF, especially for pro level great lens - sports/BIF/etc must have very fast focus, not yet for CDAF on more lens.

[AnthonyEngelken]

Sure there might be situations that 'infinity' or 'macro' may be affected but look at how on sensor CDAF/PDAF works, it has NOTHING to do with the lens beyond being able to control the focus mechanism. Correct?

Hmm... I think we're having two different discussions. I don't really care about AF; never use it with video. I'm just wondering about the complications in mounting manual lenses closer to the sensor.

[RavingRover]

No AF focus NO manual focus, I suspect you do care!  AF can move the focus mech. as far as it can go, understand?