JohnDizzo15 wrote
Personally, I generally agree with you on 12 vs 14 bit, as for 90% of users, they'll never see the difference.
aCuria wrote:
People often claim that bit depth is overrated and that 12-bit looks the same as 14-bit. This idea only holds true for cameras that cannot produce a clean enough signal to fill 14 bits in the first place. If a sensor delivers only 12 bits of usable information, the remaining bits simply record noise.
Cameras like Hasselblad's show the real value of greater bit depth. Their 16-bit files reveal smoother skin tones, cleaner shadows, and more natural transitions because the sensor actually produces more than 14 bits of meaningful data.
14 bits is not useless. Its just that not every sensor just can take advantage of it. The best sensors prove that higher bit depth does matter when the hardware is capable of delivering the signal quality to match....Show more →
Whether or not 14 bits is a useful benefit over 12 bits will depend on use cases for each photographer. Even sports/action photographers, for example, can benefit from clean shadow recovery in high contrast daylight scenes at low ISO. But is that a typical use case scenario? For my sports work, it's more often the exception than the norm, where instead I'm typically at ISO 800 and higher. By this point pretty much every current sensor has very similar DR performance. Therefore other factors are much more influential in my buying decisions.
I waited for the R6III's official release to confirm its feature set. For a lot of my work, it's actually very close to what I would want in a camera. 33MP is good. Deep buffer is good. Overall incremental improvements over the R6II are all welcome. The tipping point for me to go with the R5II instead was primarily its stacked sensor. Additional to this were a number of otherwise seemingly minor settings/options that remain excluded from the 6-series and that are more quality of life factors than real necessity. It also helped that holiday discounts reduced the price difference between the two cameras by more than half.
As an R62 and R8 owner, I was initially fairly excited about the prospects of the R6 Mark III.
I (wrongly) assumed Canon would bring significant upgrades to sensor readout speed and to the EVF. (I was looking forward to greater resolution and a blackout free mode.)
Neither of those things happened, unfortunately.
Not being a video shooter, the only significant improvement for me is in resolution. It's a welcome change, but just not enough to justify the upgrade from the R6 Mark II, IMO.
I owned all the 5-series DSLRs, but so far have resisted the call to spend the money on the R5 series models.
What I'd probably really like is an R3 Mark II with a BSI stacked 33+ MP sensor having the same size and handling of the current model, but I can't imagine how Canon could create meaningful daylight between this hypothetical camera and the R1 that could justify strong continuing sales of the latter. I think both cameras would just appeal to the same small market.
rscheffler wrote:
Whether or not 14 bits is a useful benefit over 12 bits will depend on use cases for each photographer. Even sports/action photographers, for example, can benefit from clean shadow recovery in high contrast daylight scenes at low ISO. But is that a typical use case scenario? For my sports work, it's more often the exception than the norm, where instead I'm typically at ISO 800 and higher. By this point pretty much every current sensor has very similar DR performance. Therefore other factors are much more influential in my buying decisions.
I waited for the R6III's official release to confirm its feature set. For a lot of my work, it's actually very close to what I would want in a camera. 33MP is good. Deep buffer is good. Overall incremental improvements over the R6II are all welcome. The tipping point for me to go with the R5II instead was primarily its stacked sensor. Additional to this were a number of otherwise seemingly minor settings/options that remain excluded from the 6-series and that are more quality of life factors than real necessity. It also helped that holiday discounts reduced the price difference between the two cameras by more than half.
Dynamic range is how much signal the sensor can capture.
Bit depth is how finely that signal is sliced when it’s encoded.
These two things are often correlated in practice, however, in that too little bit depth can truncate the signal available from the sensor. (This is a luxury we are afforded from modern sensor designs.)
Conversely, additional bit depth does nothing to improve the signal past the sensor limits.
Dynamic range is how much signal the sensor can capture.
Bit depth is how finely that signal is sliced when it’s encoded.
When the signal at higher ISOs is limited to X stops of DR, does 16, 14, or 12 bits make a real difference? Or are those extra slices just padding nonexistent image information?
rscheffler wrote:
When the signal at higher ISOs is limited to X stops of DR, does 16, 14, or 12 bits make a real difference? Or are those extra slices just padding nonexistent image information?
If the camera genuinely can’t resolve more than 12 bits, Canon is technically correct to cap the output at 12-bit rather than pretend there’s more data than there really is.
But when we say “damn, this camera doesn’t do 14-bit or 16-bit,” what we really mean is this: the sensor isn’t good enough. It’s not resolving fine tonal and color differences cleanly enough to deserve those extra bits.
In other words, the argument isn’t about missing bits... it’s about missing sensor performance.
rscheffler wrote:
Whether or not 14 bits is a useful benefit over 12 bits will depend on use cases for each photographer. Even sports/action photographers, for example, can benefit from clean shadow recovery in high contrast daylight scenes at low ISO. But is that a typical use case scenario? For my sports work, it's more often the exception than the norm, where instead I'm typically at ISO 800 and higher. By this point pretty much every current sensor has very similar DR performance. Therefore other factors are much more influential in my buying decisions.
I waited for the R6III's official release to confirm its feature set. For a lot of my work, it's actually very close to what I would want in a camera. 33MP is good. Deep buffer is good. Overall incremental improvements over the R6II are all welcome. The tipping point for me to go with the R5II instead was primarily its stacked sensor. Additional to this were a number of otherwise seemingly minor settings/options that remain excluded from the 6-series and that are more quality of life factors than real necessity. It also helped that holiday discounts reduced the price difference between the two cameras by more than half.
With AI software continuing to improve, is the difference between 12 and 14-bit going to eventually matter? Don’t get me wrong, at this point in time, I’ll take 14-bit every single time. However, look at what Adobe and DxO are doing with their algorithms. Imagine doing a 6-stop recovery and then NR software cleanly removes the noise and retains details like as if it was a clean shot to begin with. I think that day will come. It’s only a matter of when and not if.
There is a rumour that Canon may introduce open-gate 8K RAW video with a firmware update for the R5 II in February 2026. That would be awesome if true.
rscheffler wrote:
When the signal at higher ISOs is limited to X stops of DR, does 16, 14, or 12 bits make a real difference? Or are those extra slices just padding nonexistent image information?
With AI software continuing to improve, is the difference between 12 and 14-bit going to eventually matter? Don’t get me wrong, at this point in time, I’ll take 14-bit every single time. However, look at what Adobe and DxO are doing with their algorithms. Imagine doing a 6-stop recovery and then NR software cleanly removes the noise and retains details like as if it was a clean shot to begin with. I think that day will come. It’s only a matter of when and not if.
A high-quality original always matters. AI noise reduction does not retain details like as if it was a clean shot to begin with. It just adds fantasy detail to make the image look less noisy.
If that day comes, it will be the end of photography. You may as well leave the lens cap on, and leave all the drawing to the algorithms.
Toothwalker wrote:
A high-quality original always matters. AI noise reduction does not retain details like as if it was a clean shot to begin with. It just adds fantasy detail to make the image look less noisy.
If that day comes, it will be the end of photography. You may as well leave the lens cap on, and leave all the drawing to the algorithms.
Listen, I don’t necessarily disagree with the premise of your post. However, you can’t possibly tell me nothing amazing will not happen within the next decade as technology continues to improve. A decade ago, no one could have predicted what we have today.
lsquare wrote:
you can’t possibly tell me nothing amazing will not happen within the next decade as technology continues to improve.
There's still a whole lot of work to be done with respect to sensor readout speed - meaning rolling shutter will be eliminated, framerates will be something that needs to be set because they will be able to go way too high, and hopefully the readout noise involved that's sapping dynamic range in faster sensors will be a thing of the past in the process.
Then there's processing power - currently there's a limitation first of cost, faster processors cost more, then there's power usage, where heat limits recording times and causes other issues, and finally there's battery life, which sits adverse to processing power and is a problem for mirrorless cameras that are always 'live view'.
And then there's software. To be frank, any post-processing that can be articulated can be codified in an algorithm and eventually done in-camera. Now that we're getting higher-quality compression options (see JPEG-XL), it's entirely possible that much responsive photography (events, sports, whatever) could be done with high-quality results straight out of camera. Especially stuff that phones are doing now with multi-shot stuff, a camera can absolutely analyze a scene and suggest appropriate exposure settings and processing options to get the result that the photographer is after. This process could also be customizable, and could even 'learn' from the photographer such that it gets very good at what the photographer is interested in capturing.
lsquare wrote:
A decade ago, no one could have predicted what we have today.
I don't think that's really fair - we knew that mirrorless would be bringing super-accurate autofocus, and that having cameras that can do live previews during shooting as well as tracking were primarily a matter of sensor readout speed and processing speed, both being advanced with every major release even in DSLRs before mirrorless really hit its stride.
But I do get the sentiment - for people working with cameras regularly, advances seem to come at a snail's pace, and it's difficult to really see where things are going or how fast they can get there. And of course it's in the manufacturers' interest to not really be transparent about where they're taking their technology, as they want to sell the products that are already on the market!
aCuria wrote:
If the camera genuinely can’t resolve more than 12 bits, Canon is technically correct to cap the output at 12-bit rather than pretend there’s more data than there really is.
But when we say “damn, this camera doesn’t do 14-bit or 16-bit,” what we really mean is this: the sensor isn’t good enough. It’s not resolving fine tonal and color differences cleanly enough to deserve those extra bits.
In other words, the argument isn’t about missing bits... it’s about missing sensor performance.
Thanks for the link. I'll check it out when I'm able to. In the meantime, all I can gather from the two images/crops is that there are many more variables at play that would affect overall image quality than just 16 vs. 14 bits. When human vision is ~16M colours approximating 8 bits per RGB channel, in an image that is properly lit and requires minimal post work, is 16 instead of 14 bits actually visually perceptible? I appreciate that more bits allow better editing options, but at some point the diminishing returns won't outweigh other considerations that might instead sacrifice bit depth for other capabilities. As always, it really depends on what you prioritize for your work.
lsquare wrote:
With AI software continuing to improve, is the difference between 12 and 14-bit going to eventually matter? Don’t get me wrong, at this point in time, I’ll take 14-bit every single time. However, look at what Adobe and DxO are doing with their algorithms. Imagine doing a 6-stop recovery and then NR software cleanly removes the noise and retains details like as if it was a clean shot to begin with. I think that day will come. It’s only a matter of when and not if.
I appreciate where you're coming from and like you and others, yes, I'd prefer the best in-camera capture/information as possible.
I have daytime sports images shot at lower ISO in e-shutter (12-bits) where I've opened deep shadows on backlit players wearing black uniforms and indeed, the resulting shadow noise was objectionable. The usual/expected noisy Canon shadow quality. Then I ran those edited CRAW files (yes, lossy compressed RAW, the heresy!) through Denoise and the result was very nicely cleaned up deep shadows. Maybe there is some AI 'make believe' information in those areas but one would have to look really closely. The overall result though was really good and wouldn't have been possible in ~10 seconds with traditional NR techniques.
We're definitely moving in the direction where we're handing over more decisions about image content to automation, but this is the path photography has been on for ages as technology has progressed.
rscheffler wrote:
Thanks for the link. I'll check it out when I'm able to. In the meantime, all I can gather from the two images/crops is that there are many more variables at play that would affect overall image quality than just 16 vs. 14 bits. When human vision is ~16M colours approximating 8 bits per RGB channel, in an image that is properly lit and requires minimal post work, is 16 instead of 14 bits actually visually perceptible? I appreciate that more bits allow better editing options, but at some point the diminishing returns won't outweigh other considerations that might instead sacrifice bit depth for other capabilities. As always, it really depends on what you prioritize for your work.
I appreciate where you're coming from and like you and others, yes, I'd prefer the best in-camera capture/information as possible.
I have daytime sports images shot at lower ISO in e-shutter (12-bits) where I've opened deep shadows on backlit players wearing black uniforms and indeed, the resulting shadow noise was objectionable. The usual/expected noisy Canon shadow quality. Then I ran those edited CRAW files (yes, lossy compressed RAW, the heresy!) through Denoise and the result was very nicely cleaned up deep shadows. Maybe there is some AI 'make believe' information in those areas but one would have to look really closely. The overall result though was really good and wouldn't have been possible in ~10 seconds with traditional NR techniques.
We're definitely moving in the direction where we're handing over more decisions about image content to automation, but this is the path photography has been on for ages as technology has progressed....Show more →
Maybe I'm delusional, and I am aware of the drawbacks of the current AI algorithms, but a decade is a long time, especially in technology. It's not hyperbole to say a decade is an eternity! Computing power and algorithms will continue to improve. It would be premature to suggest that the "fake details/information" added can't be improved to be more natural. I've been shooting RAW for almost two decades. This is the reason why I do it, and that the software will continue to get better and bring new life to my images. If all of us are still alive a decade from now, let's revisit this thread and judge how well our comments have aged!
lsquare wrote:
With AI software continuing to improve, is the difference between 12 and 14-bit going to eventually matter? Don’t get me wrong, at this point in time, I’ll take 14-bit every single time. However, look at what Adobe and DxO are doing with their algorithms. Imagine doing a 6-stop recovery and then NR software cleanly removes the noise and retains details like as if it was a clean shot to begin with. I think that day will come. It’s only a matter of when and not if.
We are already there. Use a totally white image as the prior (meaning completely blown out beyond recovery) use the paintbrush tool to tell it where your subject is and the AI can hallucinate something for you.
At this point it ceases to be a photograph and you have to decide if that’s fine for you or not 😂
rscheffler wrote:
Thanks for the link. I'll check it out when I'm able to. In the meantime, all I can gather from the two images/crops is that there are many more variables at play that would affect overall image quality than just 16 vs. 14 bits. When human vision is ~16M colours approximating 8 bits per RGB channel, in an image that is properly lit and requires minimal post work, is 16 instead of 14 bits actually visually perceptible? I appreciate that more bits allow better editing options, but at some point the diminishing returns won't outweigh other considerations that might instead sacrifice bit depth for other capabilities. As always, it really depends on what you prioritize for your work.
I appreciate where you're coming from and like you and others, yes, I'd prefer the best in-camera capture/information as possible.
I have daytime sports images shot at lower ISO in e-shutter (12-bits) where I've opened deep shadows on backlit players wearing black uniforms and indeed, the resulting shadow noise was objectionable. The usual/expected noisy Canon shadow quality. Then I ran those edited CRAW files (yes, lossy compressed RAW, the heresy!) through Denoise and the result was very nicely cleaned up deep shadows. Maybe there is some AI 'make believe' information in those areas but one would have to look really closely. The overall result though was really good and wouldn't have been possible in ~10 seconds with traditional NR techniques.
We're definitely moving in the direction where we're handing over more decisions about image content to automation, but this is the path photography has been on for ages as technology has progressed....Show more →
Both images were shot in studio conditions and the 16 bit image is clearly better
