[DONE] HDR with more than 9 pictures?

[MrFurious]

I know the new auto does unlimited number of exposures, but I'm a visual effect supervisor and I require more than 9 images and I need to shoot the same number of exposures for each angle when capturing a spherical HDR.

Will the HDR templates on the to do list be able to accomidate more than 9 exposures?

if not can you just increase the number?  is adding another pad level for x to xx a pain?

thanks in advance.

Andy

[Marsu42]

if not can you just increase the number?  is adding another pad level for x to xx a pain?

Personally(!), since this is requested again and again I really don't see a reason not to do it - how many frames max. do make sense to you? Like 1....9 and then 10/15/20/25? Btw you can also just change this inside the ml source yourself and compile, it's not *that* diffictult to do.

[a1ex]

I don't see any reason for doing it. You can always get the intermediate exposures by applying exposure compensation in post; no need to use 0.5 EV between brackets just because you like to kill the shutter.

[3pointedit]

hmmm, for vfx I guess the end goal is a large EXR with low noise and good rendition across all ranges? Thus you would use a narrower component (sweet spot) of each exposure, and allow you to acurately expose for extremes like the sun or light fixture elements (without clipping).

[Audionut]

1EV steps will give you nearly 20EV of dynamic range. 

I've mentioned this previously.  Is this dupe request month!

[a1ex]

Can you show the difference between an EXR obtained from say a 9x2 bracket and one obtained from say 33x0.5?

Or maybe 9x4 vs 65x0.5?

[Marsu42]

I don't see any reason for doing it. You can always get the intermediate exposures by applying exposure compensation in post; no need to use 0.5 EV between brackets just because you like to kill the shutter.

I personally use 1.5 ev brackets and +-ev them in post, but as the op stated he's a " visual effect supervisor" which sounds mighty professional I'd really like to hear if he can come up with a reason why he'd need more than 9 for whatever he's doing...

[MrFurious]

Hey guys,

Thanks for the conversation and the very good points.  Here are my answers and I'll preface it with that I have over 16 years in "the business"  and sometimes we need to compromise our onset data gathering because of massive time constraints.

Generally, I create a spherical HDR and is use to replicate in CG the onset lighting using a min of 12 stops in 1 EV steps. Our software takes advantage of this massive amount of range and makes a huge difference when the data is correct.

As mentioned having the right amount of steps I can really find that sweet spot with as little interpolation as possible. We often have to re-expose and color correct the final spherical .hdr images to match our footage.  This critical step is where an image with less underlying data could fall apart and effect the final product. (in my opinion)

For our purposes we don't really need to over expose for more than 1 or two stops over the "perfect exposure".  So I maximize my steps within a limited range of exposure.  Secondly as I mentioned before time is a huge commodity when on set and we're often holding up the entire crew from breaking down a set or moving on to the next shot.  Therefore I generally need to capture my 3 angles (using a sigma 8mm) often times in under 1.5 minutes.  That gets me the data I need and the crew is happy and often stunned when i'm done so quickly.  So using an EV step of 1.5 will get me close but depending on the lighting conditions it may add another 20-30 to capture my full set of images.  On a film/TV/Commercial set that is an amazing amount of time.  Normally I'm about halfway done when someone is "kindly" asking me to clear the set.

Usually I use a promote but my main camera's usb port shuffled off the mortal coil...and I don't want to shell out $400-$600 for a repair when ML does ALMOST exactly what I need.

So my long winded answer is a bit of what has been mentioned and for time efficiency so I can do my job and get out as soon as possible.

To the dev team, thanks again for your amazing efforts and thanks for listening to my humble request.

Also I apologize if this is a duplicate request I did my best to search the forum before my post.  I guess semantics got the best of me.

Andy

[MrFurious]

And to elaborate a bit more on my .hdr process for this specific application.  I will go from completely black or rendering the light sources to be only as big as a pixel or two to 1 or 2 stops overexposed image.

Our software will actually use this massively underexposed data to place spatially correct direct light sources. (both color and intensity with a little bit of processing)

So depending on the artificial or natural lighting conditions (day or night) my goal is to render an image that captures the perceived amount of light from the location.  There's not arty stuff at this point, i just need to accurately capture whats there.

Having the added flexibility of any EV step and lets say on the top end 15-20 exposures I can get what I need in the minimum amount of time possible.  I really don't care how many times my shutter fires, if i don't get the data correctly captured I'm going to catch hell back at the office. and some poor digital artist will pay the price in extra time to complete his task.

I hope this paints a picture of my request, and is asked for out of need not just to do it because we can. (but why not right?)



Andy

p.s.

And the whole spirit of ML is to not assume what people are using the hardware full but give its full capability and let the user decide what is to much or not needed.  (my .02)

[Marsu42]

As I stated above, I ack that a lot of people might be mislead by the higher count into thinking that should +-1ev 1/3 and kill their shutter, but then again ml relies on people thinking a bit for themselves and higher shutter rarings mean more Canon camera sales, so they might like ml more :->

But if the ml devs won't implement this anyone can just change it in the source code and recompile, it's really simple: change line 4398 and the .max number in shoot.c - here https://bitbucket.org/hudson/magic-lantern/src/764a78057fe87fc3a52ffbde2d2b48c5a38e2b82/src/shoot.c?at=unified

[nanomad]

No need to edit the source code, just change the config file manually and modify the hdr.frames parameter. Then don't touch the Advanced bracket menu in-camera

[MrFurious]

ahhh cool.  thanks i'll check out the config file!  that will do!

thanks for the answer!

Andy

[MrFurious]

actually just tried this and it the number doesn't stick.  i'll see if i can have a buddy look at the source code.  if you can can make this simple change for me that would be awesome.

all i can do is ask.

anyway off to a much need vacation! i'll check the list when i get back!

Andy

[dmilligan]

what about just using the intervalometer with manual exposure ramp?

start at your brightest exposure, set the manual exposure ramp to -1.0EV, and set the interval to "like crazy"

[MrFurious]

Like I said before,  I've been happy using my promote and nothing else really gives me the flexibility to shoot how I want. (again usb port busted, so my promote is useless).

The expansion of this little feature I would have thought to be trivial, again for me its about ease of use/setup/speed/adjustability.  Which I would get from ML.  And believe me I have explored all of the suggested options in the past, they just don't work for my specific needs.  So thank you for your suggestions.

Again this being an open source project I don't except you guys (the devs) to jump on every little feature requested you would get nothing done, but I hoped to make a compelling case.

Lastly to the Devs, if you find it in your hearts to grant me this one favor I'd be happy to contribute $50 to the cause.  I'm guessing that it would take less than an hour to implement and the return on the effort would be high dollar return (at least a nice dinner or a bunch of drinks).  Is there an official donation link or something?  Feel free to contact me off of this thread if you would like to take me up on my offer.

Thanks again,

Andy

[Marsu42]

Lastly to the Devs, if you find it in your hearts to grant me this one favor I'd be happy to contribute $50 to the cause.  I'm guessing that it would take less than an hour to implement

It would take less than than one minute to implement, but it's about the general question what bounds ml should have for the general user - do features or values need adopting that only a very small minority uses/needs while it might confuse others?

And here, it's really up to the most active devs to decide (I'm not one of them) where to steer the project because as you said since it's oss modifying such a trivial thing yourself is not a big issue.

[MrFurious]

Confusing feature that cater to a few.. I agree.  Arbitrary limits to existing features...I just don't get sorry.  How will up to 18 pics versus 9 pics confuse someone?

And also that's where my $50 comes in... take my money...do the change and pay for the server space for a while... or go take the team for some beers???

Anyway I guess i'm going to have to learn how to do it myself if the dev's don't agree. As stated above the change this code doesn't appear to be do hard.  If indeed it's only tied to that one line.  I've just never compiled anything before.

Cheers,

Andy

[Marsu42]

Confusing feature that cater to a few.. I agree.  Arbitrary limits to existing features...I just don't get sorry.  How will up to 18 pics versus 9 pics confuse someone?

It's not my personal opinion, mind you, but it might make most users shoot a lot of frames @+-0.5ev while you could accomplish the very same thing with 1/3rd of the brackets and +-1.5-2ev and postprocessing - so the limit is a shutter saver.

And also that's where my $50 comes in... take my money...do the change and pay for the server space for a while... or go take the team for some beers???

I'm not a dev, but I guess the guesture is appreciated and some $$$ for sponsored camera bodies are nice. But ultimately, ml isn't about getting paid for programing, even if I think some form of compensation for the most active devs makes sense given their talents and the value they create by enhancing €3000 cameras that most contributors here are never able to afford themselves.

Happy compiling, really, just getting the pre-build vm, hg pull, change 1 line and typing make isn't that hard to do :-)

[Greg]

This work on 500D :

Code Select Expand

{
    .name = "Frames",
    .priv       = &hdr_steps,
    .min = 1,
    .max = 15,
    .icon_type = IT_PERCENT,
    .choices = CHOICES("Autodetect", "2", "3", "4", "5", "6", "7", "8", "9", "10", "11", "12", "13", "14", "15"),
    .help = "Number of bracketed shots. Can be computed automatically.",
},

[MrFurious]

Greg thats awesome!  I think i hurt my brain when i looked at all the links for getting the compiler working on OSX.  care like sharing that with me?  I have a 5DmkII.

if not i'll continue to go through the docs and hopefully get the compiler up and running.

Andy

[nanomad]

I've got an idea on how to fix it, should be in the next nightly build. Just drop me a PM with your camera model if you want to test it earlier..

[dmilligan]

try this:

https://dl.dropboxusercontent.com/u/74060/autoexec.bin

(use the latest nightly and replace with this autoexec, I only tried it with a 60D but should work fine for you)

[nanomad]

That depends if it is a unified build or not though

edit:

Feature done, waiting to be merged after I test it at home

https://bitbucket.org/hudson/magic-lantern/pull-request/217/

[MrFurious]

rock on!  I tested the link you sent and works great!

thanks again,

Andy

[boogotti84]

Yes this is very useful for creating Hdri,s for vfx artists, any builds for 6D? These extra stops are needed for recreating cgi lights correctly