Menu

Show posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Show posts Menu

Messages - KarateBrot

#1
how did you get such smooth shots without any camera shake?
#2
General Help Q&A / Re: HDR Video
November 06, 2012, 07:45:56 PM
would it be possible to write a frame to the footage after it's half exposed and then the same frame after it's fully exposed?

With that there still would be just one stream but because the exposure of both frames starts simultaneously and the ghosting only occurs in the motion blur one shouldn't notice it (for high shutter speeds there wouldn't be any ghosting). and one can work with a constant ISO value (a lot less noise).
i had this idea in mind for a while and if it's possible i think there is a lot of potential for further HDR video development.
#3
General Help Q&A / Re: The hobbit framerate achivable?
October 27, 2012, 07:59:49 PM
yeah it was done for 3d. as far as people say it is looking very good
#4
Quote from: okaro on October 22, 2012, 02:59:27 PM
I tried to use visioncolor on the 550D as picturestyle preset but it won´t function in movie-mode, although set in M (photo mode).
any idea why?
Connect it to the computer once again and go into movie mode. Then check for the picture styles with the EOS Utility. If they're not installed, well, obviously install them. That should do it.
#5
yeah that's extreme. but the downside is the poor video quality. i'd rather have a better sensor and better colors than a huge resolution
#6
General Chat / Re: Magic Lantern for 700D?
October 17, 2012, 03:58:45 PM
Oh I see... too bad I didn't notice :o
#7
General Chat / Re: Magic Lantern for 700D?
October 17, 2012, 12:20:45 PM
The photoshopping at the side view is so bad I don't know what to say :D
#8
QuoteThe core problem was the fact that my previewing of my camera's .MOV files on my PC desktop using WMP and VLC, resulted in me believing that contrast/levels were different to what they really were...and then doubting the NLE software...

A great tool (especially for compositing work) is the "RV" viewer from Tweak Software.
#9
I recently bought a picture style for my 550D. It's called "VisionColor" and is modelled after 35mm analog film stock. I did some testing and IMO it's absolutely amazing!

They also made a new picture style called "VisionTech". It's VisionColor but with slightly raised shadows.
VisionLog is still in the beta phase but it's looking promising.

Visit their website and take a look for yourself:
http://visioncolor.wordpress.com/

#10
oh great! i was about to ask if there was a new compiled version around :)
#11
General Help Q&A / Re: 1/47 shutter?
October 12, 2012, 01:56:08 AM
At 24 fps you want to use a shutter speed of 1/48s (180° rule).

|1/48s - 1/60s| = 0,0042s
|1/48s - 1/47s| = 0,00044s

1/47s is 10 times closer to 1/48s than 1/60s would be (and btw. 1/47s is two times as close as 1/50s would be). 1/47s and 1/48s are practically the same shutter speeds. So I wouldn't bother. Just use the 180° rule as close as possible and you'll get your desired look. In this case 1/47s is the best and most precise setting you can choose.
#12
you can let your computer rename those files automatically. on windows it's possible and i guess on OSX too
#13
My point with the (at least) 16 bit was just concerning the case when you need to pass the sequence along to another program after you applied a LUT (or grading). 32 bit is of course always better like you said.
You're propably right as I shouldn't bother about these things with ugly compressed h.264 footage. If there is not much of a difference I just forget about that. Thanks :)
#14
Yeah sure, I know about 32 bit linear workspace. My point just was that maybe IF the HDR processing is dependent from applying a LUT BEFORE processing you would better deal with footage higher than 8 bit. Just so that the pixels have got the correct brightness value to interpolate the two frames in a realistic/physical way. But I don't know enough to say if there really is a difference if you apply the LUT before or after merging. (Because even if you're working in 32 bit linear you can alter the pixel's in a non-linear way (baking the gamma) and it's still linear workspace and if you interpolate "wrong" values in a non-linear way (for example gamma 2.2) you will get a totally wrong result. That's why I think it may be important to deal with the exact time to apply the LUT to get the correct color values)
Or do I get your wires crossed?
#15
But 16 bit would be essential if you apply a LUT before the conversion so that you have more color values in the dark spots.
Since I don't know if there is a difference if you apply the LUT before or after merging to EXR's just using trial and error to see what's better seems to be the best option, indeed. I wish I would know mathematics a lot better because then I propably had an answer to that.
#16
If you convert them to EXR's you definitely should apply the LUT before conversion and save it in 16 bit (otherwise shooting cinestyle and applying the LUT wouldn't make any sense because in 8 bit you would loose all detail in the blacks again). then convert it to EXR's.
BUT if your program wants to get fed with the sRGB color space (for example) you need to apply a sRGB conversion after your LUT and then let your program make EXR's.
the reason is: to let the program guess the right exposure of each pixel they need to respond to light in a linear way or like I said in sRGB (or whatever the program wants it to be)

Edit:
I don't know if the LUTs from the different picture styles cancel out the native sRGB from the mov files as well or if they just linearize the picture style but preserve the sRGB curve. so I can't be sure what color space it is. if you don't know better: just apply the lut in 16 bit and then convert it to exr's and don't mind about sRGB.

can anyone help me out on this? I guess if you apply a LUT (for example for CineStyle) you are still in sRGB. Right or wrong?
#17
Feature Requests / Re: film style "flash frame" effect
October 03, 2012, 02:07:45 AM
sounds interesting.

(btw.: according to your explanation the exposure is INVERSELY-proportional to the fps)
#18
Native ISO on Canon's sensors could be 125. It has got far less noise compared to 160. But I also prefer to use multiples of 160
#19
For those who don't already know Ginger HDR can produce 32bit floating point sequences of your ML HDR footage. And since 32bit images have got a linear colorspace I got the idea to convert it to a logarithmic colorspace in Nuke. This way you can easily achieve great and more natural looking gradients between dark and bright areas without clipping color channels.

Please tell me what you think about it.


Here's how I did it:


First I filmed the HDR footage with ML (ISO: 80/1250 [4EV];Neutral [0,-4,-2,0])
like you would propably normally do.

This is how it looked like with ISO 80/1250:


I merged the raw file with Ginger HDR to get a 32bit .exr-sequence.
After that I imported this sequence into Nuke and applied a "log2lin"
node in lin2log-mode and a "ColorCorrect" node.


This is what I did:
- Log2lin: I turned down the gamma (0.45) until I thought I had enough contrast in the shadow areas.
- ColorCorrect: I turned down the gamma (0.38) to get rid of the flat look of the image.
- ColorCorrect: I turned up the gain (2.1) to brighten it up again.
- ColorCorrect: I turned up the Saturation (1.14) to make it a bit more appealing.
- Log2lin: Because now some colors might clip I turned down the whitepoint just a bit (660).
Voilà!

Note: I also added a "F_DeNoise" node right after the source clip to denoise the footage since noise can be a problem with ML HDR.


- - - - -


This is the final result: (click to enlarge)


Example 2 (blurry): (click to enlarge)

Note: Notice the gradient in the sky. From the sky to the nearly edge of the sun no color channel is clipping.

Unfortunately there were no clouds when I filmed otherwise the results could have been way more interesting. Anyway, I'm pleased with the result.
If you think it doesn't look appealing, yet you can tweak the colors very well since we are dealing with 32bit footage ;)

- - - - -

Example 3: (click to enlarge)


Example 4: (click to enlarge)


Example 5: (click to enlarge)

Note: Color bleeding because of red sheet of fabric below

Example 6: (click to enlarge)

Note: There's a lot of banding because of the bokeh at different ISOs
#20
That's what i meant. let's explain it with film. the film frame gets exposed by light. half way through the exposition process the frame gets doubled and put to stream A. after the whole exposition process the original frame gets put to stream B. this way you get 2 frames which started to get exposed at the same time.
in theory you can get an infinite spectrum of exposed frames this way even though you only exposed 1/48th of a second for example.
that's the cool thing about an electronic sensor. you can record at several shutter speeds at the exact same time with only one sensor.

edit: i wonder if some day our cool ML crew has got so much control over the camera so that this is technically possible. but it's already capable of more than i have ever dreamt of :)
#21
who said that shutter speed does not change? as far as i know HDRx works with two video streams recorded simultaneously but with different shutter speeds. since it's an electronic camera it's not a problem to record with different shutter speeds simultaneously. the sensor just needs to be exposed by light. the electronic signals can be buffered. once the exposure for the slow shutter speed is finished the buffered data gets passed along to stream A but without deleting the buffer. now the sensor gets exposed until the long exposure is finished. now this data goes to stream B. now the buffer can be cleared to record a new frame. this way you end up with two video streams. and as far as i know these two streams get merged to HDR on the computer with the red software.
#22
I don't know what should be wrong with it. With ML I get perfect pictures and videos <3 And I can't say that about the Canon firmware itself.

btw.
If you set the ISO to for example 160 in ML make sure you lower the standard ISO of 200 by -0.3EV to get 160, not by raising ISO 100 or something like that.
#23
Shoot a video with your lens cap on with ISO 80 and 160. Then in post crank up contrast and probably down gamma. Then compare the two shots to really see if 80 has got more noise or not.
#24
Feature Requests / Re: Auto WB setting
August 31, 2012, 07:13:23 AM
I support this little tweak
#25
General Chat / Re: Ultrathin lens gives perfect image
August 29, 2012, 03:33:17 AM
oh my god this is blowing my mind (I'm studying physics)