14 BIT vs 10 BIT RAW VIDEO | CANON EOSM | IS THERE A DIFFERENCE?
Вставка
- Опубліковано 3 жов 2024
- In this video I compare the difference in image quality between the 14 bit and 10 bit raw video that comes from the canon eosm using magic lantern or crop mood hacked firmware.
Check out my Ko-Fi page for luts or to donate to my channel!
Ko-fi.com/lumingatevisuals
Also the thumbnail was AI Generated, look closely its not an EOSM! Hahaha
Music from #Uppbeat (free for Creators!):
uppbeat.io/t/s...
License code: W92GNTXQ4NVHS0RU
Thanks for the video! At 10 bit if you raise the shadows in MLV app you clearly see a quantization on the sensor noise that makes it very ugly. On the other hand, between 12 and 14 bit there is no significant difference, meaning that the sensor noise is actually larger than the quantization step of 14 bit AD conversion.
So my suggestion is: in a controlled environment, like an interview, 10 bit is fine, but if you want to play safe and might need to raise the shadows or increase the exposure in post, go for 12 bit.
Great advice thanks!
@@lumingate Very welcome! Thanks for your videos, actually they made me jump into this whole Magic Lantern journey (even if I am using the 5D3 nowadays).
@@parallelvisuals Awesome to hear!
@@lumingate You should try the 5D3 with ML if you can get your hands on one!
10 bit is probably perfect for most cases. The only time 12 or 14 bit could help is with very soft widespread gradients, like doing a heavy color grade on a sky at sunrise or sunset scene. There it could reduce banding....
exactly
Problem is 10bit linear RAW doesn't capture the full dynamic range of this camera. 10bit linear captures 10 stops.
This camera is capable of 12.4 when shooting RAW images. To capture 12.4 stops full tonal range in all stops you need 14-bit Linear. That's because this camera captures light values linearly. Each stop when doubling the light value. 10-bit linear is not the same as 10-bit Log. Log is much much more efficient. 10-bit slog would be enough to encode all the dynamic range this camera has to offer. Problem is it encoded light linearly.
Most scenes do not require the full 12.4 stops. As a matter of fact most camera's can use 6-8 stops and capture enough.
Only during high contrast scenes do you need the full capability of this camera and that's when you should shoot 14-bit. But then you'd have to sacrifice resolution.
Its ultimately your choice
Extremely big difference in low light, dark and shadows...
@@vd9795 In theory yes, but the EOS-M is very noisy in darker areas, which negate the 14 bit capability of Magic Lantern... 12 bit might be a good compromise...
@@TheCruisinCrewFind a great denoiser and ETTR
From what I've seen 14 bit video has less chroma noise in the shadows and certainly has advantages that some experienced colorist could take advantage of. As an ordinary user, without much knowledge, I will probably be satisfied with a 10-bit recording.
totally agree. Try to lift the shadows on darker scenes. 14bit has MUCH less banding and noise is much more natural less distracting. Thou in MOST situations (when exposed to the right and correctly) there is almost no noticeable difference for most situations. For darker scenes and HDR scenes or just night or low light I use only 12 or 14 bit. Also underexposed 10bit with lifted shadows has much more banding (green lines in my testing).
Keep in mind that on the EOS-M (in 3K cropmood for instance) there are heavy patterns visible in the noise at 100ISO or
Get a 5d mark III
You will thank me later
And use Dual Iso and if you want it sharp and high rez use the UHDW 4k at 12 bit(or 10 bit), as long as its exposed well its continuous recording.
I use this camera alot recently and I modified a EFS 18-55 lens so I can use 18mm at UHDW 4k cause the crop on the sensor is 1.5x on 4k.
If you love 1080p then I highly suggest recording 40/30/24fps 14bit(lossless OR native) on a 5D3 and then run it through a AI upscaler because we might not see all the data, but using 14bit gives the ai upscaler more data to work with in the highs and lows (colors or otherwise) to produce the best upscaled image
Also, how can you not see at 3:30 the massive difference in the images, the 10 bits greens are muddy and brown while the 14 bit green has highlight depths so you see more color nuance in the green foliage. 10 bit is just an image of decent video while the 14 bit is appetizing and pleasant to the eyes , similar to how people love Fujifilm colors
this.
I bought an used 5DMK3 just a while ago and the difference in quality between this and its APSC brethens is very much noticable.
The neat part is, that you don't even have to buy bulky, expensive EF glass.
I almost exclusively use 10-18, 18-135 and 55-250mm EFs after modding the mount and couldn't be any happier with the results.
Shooting 12 bit is extremely stable in my experience. No reason to drop down to 10 but I need further testing to absolutely sure.
I mostly use 10bit, it's enough for 90% of subjects; colors are pretty much identical and I cannot spot any differences. However, Gabriel angel from ML forums made a very technical comparison, and it is clear that 14bit make you gain at least one and a half extra stop in shadow information over 10bit. For shots where there's strong light and shadows, using 14bit can definitely make a noticeable difference.
If you're using an 8 bit display to look at it, you won't see any difference in color rendering between 10 and 14 bit color. What you *could* potentially see is 14 bit will have a bit more usable dynamic range and potentially cleaner noise floor than 10 bit, but outside of that, if you can control the lighting, 10 bit should be sufficient for most uses.
You got that whole question wrong... 14 bit can resolve all 12 stops with this camera and 10 bit only around 9 stops of dynamic range with this camera.
Also at 4:10 Its not extreme color grading it just extremely bad color grading example😅you decreased gradations with it. So you showed even less difference than it was already.
Bit depth on canon RAW works completely different from other video formats.
Good use of 14 bit is starts to show in shadows because it uses linear brightness gradations instead of gamma corrected on rec709 videos. It literally loose up to 1.5 stops in shadows in 10 bits because it doesn't record shadow gradations that far anymore, so you will get high contrast noise instead of details in darker arias. So that is the whole point of test. Brighter you go less difference you will get. also MLVapp lose its ability to restore highlights accurately with 10 bits. So you also loose additional 1.2 stops in highlights. So best reason to shoot in 14 is to preserve and gain maximum dynamic range. It does not even give any difference with 8bit in bright aria because its RAW. it compensate its own lack of gradations with noise. More bits you have more details in shadows preserved. That is the whole point.
Thank you very much for this comparison, very useful to help choosing between the 10 or 14 bit. I can not remember where but I have seen a comparaison of the noise in the shadows or the underexposed part of the video that shows more noise in the 10 bit than the 14 bit. So if the ligh is well controled 10 bit all the way. Have a good day.
I do see a difference with 14 bit being more color depth. I actually prefer 14bit for darker scenes. 😉👍
I've learned something about cameras and especially RAW formats which might suprise you
TLDR: 10-bit ML RAW leads to only 10 stops of dynamic range which with the first 2-3 stops in the shadows being compressed. Shoot in 14-bit ML RAW to unlock the full 11-12 stops of Dynamic Range.
So cameras encode the light values from the sensor in two ways. Logarithmic or Linear.
ML Raw stores images Linear light values. For each stop we go up in exposure we need twice as many code values. 10-Bit in ML can only record 10 stops.
If stop 1 uses 1 code value, stop 2 will use 2, stop 3 will use 4, stop 4 will use 8 and so on.
10-bit = 1024 values = 10stops
12-bit = 4096 = 12 stops
14 = 16,384 = 14 stops
Also known as
2 power of 10 = 1024
2 power of 12 = 4096
2 power of 14 = 16 384
So as you can see from the above, we only have 10-bit data and as a result 1024 code values to play with.
We can only record an absolute maximum of 10 stops of dynamic range using linear data.
This is why most 14 stop RAW cameras use 16-bit data for Linear or use 12-bit LOG. 12-bit LOG is
So to capture the 11-12 stops this camera is capable of, you need to shoot in 14-bit.
This is the reason why 14-bit RAW images from this camera seem to have much more dynamic range than the 10-bit ML RAW. It's become the camera is incapable of storing values above 10-bit when shooting in that mode, leading to you only capturing 10 stops of dynamic range with the first 2-3 stops in the shadows lacking tonal range and looking compressed.
Try lowering the resolution to get 14-bit recording and be amazed at the dynamic range this camera is actually capable of shooting in ML RAW.
Wow
In English please
How about go for a 12bit and lost 0.4 stop of detail. Is it enough to notice the difference?
I use my M for b cam on interviews. I've come to terms with what it needs, in terms of light, and I am quite happy to use it now. It routinely looks better than my M6 Mkii 4K
Lower bit rate starts breaking up in the shadows, it's all down to other settings too. 10bit iso100/200 shadows have more muddy degraded noise but crank iso up to 400/800 and you'll notice less noise but more of the grain type noise which can be eliminated with noise reduction. Finally if you export prores 422hq from mlvapp it becomes 10bit. So 14bit is nice but necessary? Definitely not.
I agree in some part, but be aware that 10/12/14 bit n camera is linear, while once you exrport it, it is not, so you cannot compare the bitdepths directly.
For example, 10 bit log can have much larger dynamic range than 14 bit linear encoding.
Great and useful video. This is confirmation that I Should just continue using 10bit😁
Nice comparison! Great info in this 👏
Thank you, Sir! I wanted to know the difference.
Great content! I would like to see a noise test with this!!
Yes and a sky will show some banding probably?
You won't notice the difference with the subject examples you were shooting - You need to be looking at stuff like skies and areas where there is more gradation colours. These were bad examples unfortunately. You should do this test again.
This is useful. Thank you.
This was great, thanks for the comparison.
Maybe 12 bit is the answer?
My thoughts exactly. I shoot 12 bit with no issues.
@@TheFilmmakingChanneldoubt there is a difference
Easiest way to see it is looking at the sky, which is many of slightly different shades of blue. In 10 bit you can see obvious lines where multiple shades are merged into one, but with 14 bit you will see all of the different shades of blue.
Id recommend You to get some decent C mount zoom for M, then you can shoot almost 3k, true 3k resolution on 100 euro cam and you get shallow dof of about 16mm film and even some more, leneses for 2/3 sensors will work too. Also there is some difference where each area of colour has its own depth and own "cloud" of its color covering darker areas, hard to explain but 10bit looks more sterile and 14bit looks more organic
Sorry if I’m misunderstanding, but because this was exported for UA-cam, what we’re looking at is 8bit anyway, right? Can’t really tell the difference unless grading. Again I don’t really know.
How can we tell? We're watching on UA-cam, the great limiting factor
Would love to see your Eos M workflow from start to end.
Picked up an EOS M thanks to you :)
After watching your videos, I couldn’t help it ($50 on Karrot Market)
On a side note, was the main video shot on an M50?
Awesome! That's a steal!
No main can is xt4!
Devices like the Atomos Ninja V produces 10 bit color as is considered a high end image so why the 10 bit image should be considered as less quality?
First, I never would have known about the thumbnail had you not mentioned it! Second, what settings are your shooting in? Third and last, what card are you using? Thanks!
Shooting in the uhd mode and using the 64gb 170mbs san disk cards
I don't believe there are any displays capable of anything higher than 10bit if you go into the display settings on your monitor or tv you will see,its crazy to think that these old cameras are capturing more information than can be displayed currently
hi great review and samples thank you. not an eos m user but considering to dabble into it. quick question - would ML firmware overwrite the normal canon interface and hinder its original photo taking functionality? thinking of getting one with the 22mm you recommended as a everyday travel kit mainly to shoot stills but also with the option of taking great videos with ML
If you hold the Set button on startup, ML won't start so you can take pictures with the normal Canon interface
I've also noticed that in the most recent Crop Mood build by Bilal, you can just use the Info button to get rid of the ML interface and use the touch controls but in Danne's most recent build, this method is a lot more messed up in my opinion unless I'm missing out on something
👍
Thanks for your testing 😊 I think it's not only shooting 10 or 14 bit also depend on monitor you watch. By the way my monitor is 10bit supported, and I don't see any difference.
Yay for eos m video
hello, any tripod recomendation for this camera? nice video
Which lense is used in the clips?
are you living in Korea? I've noticed it when I saw the first cut of your video!
Yeah i am ATM!
Is he holding an invisible mic? 😮
great stuff!
I think differene could emerge in over and underexposing
im planning on buy a second hand canon eos m as my first camera {mainly cuz most of the magic lantern videos are based around that camera model} could you please help me to know what all i should check in the camera before making the purchase, i will be meeting the seller and checking the camera in person
Just make sure all the buttons work and the sensor looks clean
@@lumingate THANK YOUUUU !!!
just get the bmpcc og; best value.
No difference to my eyes.
I use 12-bit most of the time.
I recently made a short video about my first hive inspection after a harsh winter.
Most of the shots were taken with Canon EOS-M Crop Mood 5.2K, 12-bit, Canon EF-S 18-135mm, 55-250mm.
ua-cam.com/video/9KqsKjLaDak/v-deo.html
i will put the number here so you can understand why you wont need 14 bit. 1bit= 2^1
8 bit give you 256 shades of blue, 256 red 256 green. Combined 256*256*256=16 777 216 shades of color.
10 bit give RGB :1024*1024*1024 = 1 073 741 824 color (we are from 16 million of color to 1 billion)
12 bit give: RGB : 4096*4096*4096 = 68 719 476 736 color
14 bit give RGB16 384*16384*16384 = 4 398 046 511 104 (14bit is 4 trillions color). who need 4 trillion color? And what in kind of display can you watch that?