Canon EOS R5 reviewPart 2: Videography

Part 1: Photography | Part 2: Videography

This review is based on R5 firmware 1.8.1

When it came out, most of the headlines about R5 talked about its amazing video capabilities which were at the time pretty unique, like 8K30, 4K120, oversampled 4K from 8K, internal 4:2:2 10-bit recording etc.

After waiting Nikon for years to produce something similar (not counting the heavy and expensive Z9), I got tired of waiting and got the R5.

Since I record wedding videos as a single shooter, which is slightly stressful to put it mildly, I use a lot of assists the camera provides. This is important for the remainder of this review, so I’ll spend a minute to elaborate my process.

On my Nikon Z6, I mostly shoot in aperture-priority and use exposure compensation to guide the camera metering, so I can respond to quickly changing lighting conditions while not missing a shot, especially when I’m moving during single-takes.

I also use auto-WB for everything but the dinner reception; again, same rationale as with the Av mode.

Finally, I use continuous AF on-demand (meaning it focuses constantly while I keep the shutter button half-pressed). You would think this is self-evident, but apparently not to everyone, as I’ll explain shortly.

The goals of my upgrade

My intentions were to jump from the 1080p to 4K workflow. Since I use various amounts of slow-motion in my editing (50 and 100 fps), upgrading to 4K while retaining this style was quite demanding for the camera, which is why I waited this long for an upgrade.

I also wanted to improve my color grading look & feel, so being able to internally record 10-bit log 4:2:2 footage on R5 was a huge gain for me. Not only that, but this footage is more forgiving of mistakes in exposure or white balance. This means more time spent thinking about the artistic aspects while shooting and less time dwelling on fractions of a stop of exposure compensation, or a few hundred Kelvins here and there.

Another big wish of mine was getting access to Canon’s amazing face detection in video, so I can stop maniacally tugging on the joystick, trying to keep it over someone’s face while recording. Again, reliable face detection would go a great length to freeing up more of my mental resources for actual art.

Furthermore, Z6 has only basic stabilization, at least with adapted F-mount lenses I prefer, so a lot of additional stabilization was done later electronically, while editing. This was painfully slow and also degraded the image quality by slightly zooming into the image, so 1080p footage wasn’t even 1080 anymore.

The R5 promised huge leap in stabilization with native lenses, with sensor-shift IBIS working in tandem with lens IS, so I was hoping to send post-hoc stabilization to history.

The STM motor in the RF lenses was also supposed to be very quiet. I was looking forward to recording clean soundtracks without the endless AF motor stutter from my beloved Nikon 28mm f/1.4 (lens I used for all of my videography ever since I got it).

I couldn’t however find reliable reviews describing the AF sound on the RF 24mm, so I was only cautiously optimistic.

Finally, Z6 has some UI quirks which are driving me crazy, like having to choose between sound levels display or live histogram, even though there’s plenty of screen real estate to fit both at the same time. My hope was that R5 would display both, based on video reviews where I saw both displayed in video standby screen.

On the other hand, Z6 has an amazing way to adjust exposure compensation while recording which results in a smooth transition (using easily accessible soft buttons which wouldn’t visibly shake the camera when pressed). I needed to figure out an alternative for this, preferably avoiding using the dials on the body because they are loud, clicky and often shake the camera as you move your hand to operate them.

The plan was to use the RF control ring on the lens for direct exposure compensation.

I was also wary of R5’s propensity for overheating during high quality video recording, something made worse by the fact I’m often shooting in peak summer season, in direct sunlight at temperatures easily reaching 40C (104F). This was something I’d keep an eye on, keeping a cool backup camera near me at all times.

What I actually got

Okay, now that you know the expectations I had going into this, you should know that I did a lot of research to try and confirm I will be getting all these things. I read 940 advanced user’s manual, a heap of detailed reviews and even more Youtube reviews.

The trouble was that most of the people doing the reviews had specific use-cases in mind with needs very different from mine (mine being a classic set of run ’n’ gun, candid, documentary, reportage wedding requirements). So, with a bit of apprehension and a good measure of “surely Canon can do this, it’s just a bit of software”, I clicked the order button.

Face detection

The biggest unknown I couldn’t completely verify in advance was the exact face detection implementation and its accessibility during recording. Even reading the complete manual wasn’t too informative in this regard. Naturally, this was one of the first things I checked after unboxing the camera.

Unfortunately, I was pretty disappointed. With the caveat that I may have missed something despite all the research, I could not find an easy way to toggle face detection on/off during video recording.

Ideally, I would have a key that would engage face AF while depressed during video recording (like in R5’s photo mode). I would have settled for a way to easily turn it on and off before starting the recording. But alas, neither were available.

The quickest ways I found to toggle face detection were:

1. Turning off all modes between face detect and your favorite AF mode; then activating the Q menu and leaving the selection highlight on AF modes. Then, when needed, using the Q key and control dial to jump between those modes in a single click (because all other modes in-between were turned off). Okay, got all that? Not exactly a single-key millisecond adjustment!
2. Creating two custom shooting modes, one with face detection on and another with it off; then switching between them as necessary using the MODE key and the control dial. Again, not that quick or practical.

Remember, the goal of all of this was to use advantages of face detection with a quick and easy override if it messed up, which it inevitably would at some point, no matter how good it was.

In addition, I couldn’t get point-and-then-track on a face using camera’s joystick to work as it does in photo mode. Maybe it’s the high demands of reading and processing 4K video that makes the camera more limited in this regard?

So as much as I loved face detection in photo mode, it was a pretty big letdown in video mode.

Focusing modes while recording

I never use single-shot AF in video because things are always moving; even if they’re not, I’m moving. This is why I use continuous AF on-demand, which means AF is working while I keep the shutter button half-pressed. When I let go of it, focus remains fixed where it was. Perfectly predictable.

However, to my surprise I found that R5 has a very different philosophy.

This was so surprising to me that I’m still thinking I did something wrong and missed an important detail. (Do email me if I did!)

However, I went through every camera setting multiple times and the only thing I figured out is what I’ll describe below.

In video mode, R5 defaults to single shot AF, regardless of what kind of focus box you picked. Apparently, there isn’t a continuous AF on-demand setting you can activate!

There is only something Canon calls “servo AF” (in video mode), which is continuous AF working constantly! It doesn’t require a half-press of the shutter button nor cares about anything else. It just keeps focusing for as long as the camera is on. It doesn’t even have to be recording!

While the use case for this persistent focus is beyond me (okay, some fringe cases come to mind, but…), it gets worse.

To enable or disable it, you MUST open the camera menu. Yes, really.

You have the option of pausing servo AF, either via touchscreen or a custom button (I used the DOF preview button for this, somewhat similarly to what I use in photo mode). Keep in mind though, this will only pause servo AF for a certain amount of time – like until you switch shooting modes, turn camera off or until it goes into standby.

All of this means is that you can hardly ever be certain about the status of servo AF after some time has passed, and this is just recipe for disaster as it makes the AF less predictable.

For example, when I’m filming speeches, exchange of vows or rings and similar, I will keep a face in my focus box and occasionally activate AF to make sure the person is still in focus if any one of us moved.

I won’t do it constantly, because if someone bumps into me and the frame moves, focus can jump to background – making a small bump more noticeable and irreparable. I can also have people cutting the line of sight momentarily (like the officiant moving at an unfortunate moment, waiter crossing the frame during speeches, someone sticking his phone out in front to get a photo etc.).

If AF isn’t active during those distractions, focus will remain where it should be despite the interruption. Servo AF would make this a pain and I would regularly have to check its status and toggle it constantly when I wanted to refocus.

Even worse, if I wanted to do focus tracking, like doing a walking shot along the couple, I would basically have to pick whether I want constant focusing (thereby having to be certain my focus box won’t ever leave the subject’s face), or whether I’d prefer single focus attempts as we move, with all the potential for disaster that entails.

I may be a bit dramatic here, but each wedding has those few moments where you’re running backwards just to get a good shot and you really need your AF system to make it easier for you.

It’s beyond my comprehension why Canon implemented it this way in video mode. In photo mode it’s perfectly sensible – choice between single and continuous, and a shortcut to switch when you please.

In video it’s a mess. The camera doesn’t have the most basic continuous focus function! I find it unfathomable that it is even possible to have such two different AF implementations in a single product.

Video focus tracking and STM motor noise

Okay, onwards. The AF performance of RF 24mm f/1.8 IS STM in video is pretty good. It was quick and precise, but focus tracking of objects at close focusing distance was jumpy. I expected a video lens like this to keep an object in smooth focus when I was getting closer/farther away from the object, but instead it periodically jumped to correct focus instead of continuously adjusting it. It was serious enough that I wouldn’t be comfortable using that footage.

(Sidenote: I will mention that in one particularly detailed Youtube review the filmmaker went through all combinations of tracking sensitivity and stickyness to see if this issue can be helped. It did help somewhat, but it still wasn’t perfect. However, with all the settings tweaked like that, the question remains on how the tracking of more regular and important things will work, like people at a distance. It’s all compromises, making one thing better and the other one worse. This is why I decided not to fiddle with these settings.)

The be fair, Z6 with adapted F-glass has a similarish problem, but it’s only occasionally present. And besides, the lens motors of F-lenses were not really designed for video use.

This was somewhat disappointing. I didn’t really test with other RF glass so it may be due to this particular lens, but given the video-leaning STM motor I have to wonder whether other lenses would perform differently.

Let’s now talk about AF noise during sound recording.

I spent a lot of effort trying to figure out in advance how loud this lens was in video AF when using internal or external mic, but I hardly found anything useful. So here are results of my testing.

When recording using an external mic, the lens is reasonably quiet. You will hear a bit of AF sounds in quiet environment when using a fairly directional shotgun mic, but in normal circumstances of a wedding with all the background noise – it’s inaudible, for all intents and purposes.

If you’re using a less directional external stereo mic, the noise will get slightly more audible, but it’s still okay. And for internal sound recording, you will definitely notice it in the soundtrack, being noticeably louder than with external mics (which is to be expected).

I know this is all pretty abstract, so here’s some quick and dirty measurements to help illustrate:

(Details: I set R5 manual audio gain to around half. Background noise floor was at -50 dB, and with these settings a normal tone of voice at 1m produced sound level of roughly -15 dB.)

Again, these are for RF 24mm/1.8, but I’m guessing most RF lenses with stepping motors (STM) will produce comparable results.

Log recording, noise

Okay, moving on to the actual footage. I can keep this short: the 4K 10-bit Clog3 footage with 4:2:2 sampling looks amazing!

It’s a pleasure to color correct and grade (given your computer can handle it) and you can get stunning colors out of it. It’s forgiving, it’s manipulable and looks great with a good creative LUT applied.

You can also push it much more than 8-bit footage, +2 EV easily, so if you miss your exposure it’s much easier to fix later. Pulling is more conservative of course, depending on your exposure and whether you really burnt something to a crisp, but you can still get away with a much larger correction than with 8-bit footage.

I only shot Clog3 in testing, for more than a few reasons. First is that it pretty perfectly fits the 12 EV of dynamic range sensor on R5 offers in video mode, and the second is that it doesn’t use the aggressive shadow clipping of the original Clog (which offers 1 EV less of DR compared to Clog3). So yes, you get more shadow noise, but you can easily clip it yourself later – if you want.

Here’s a great article going into much more details of Clog vs Clog3 which I highly recommend if you’re interested in the nitty-gritty of log recording.

I was also curious to see what footage shot at ISO 12800 looks like and whether it’s better to shoot low-light in log or standard (Rec.709) gamma.

I know what the theory says: if you’re shooting a scene with narrow dynamic range at high ISO, don’t use log as it will waste the 10-bit space and end up looking worse… but after testing at dinner reception/party in a modestly lit venue, I much preferred the log footage!

After color grading it looked much better, with normal contrast and reasonable amount of noise. The 8-bit Rec.709 footage was too contrasty and saturated even in the flat video profile and even before any grading.

I will mention here that for all video recording I used the in-camera video noise reduction set to level 2.

In quick tests I did at high ISO in log, noise reduction set at level 0 was just too noisy. Level 1 was beginning to help a bit, but not enough. NR level 2 did a great job, leaving some visible noise but preserving most of the details. NR level 3 was starting to eat into the details and the low-frequency noise was still visible, so it was actually looking worse than at level 2.

Of course, some of you may prefer doing the noise reduction in post, with something like NeatVideo, which produces excellent results. However, it’s expensive, and I don’t mean just in money – I mean in the processing time.

Editing a wedding and using it on multiple clips would make editing much slower and more painful, so I decided to let the Canon’s DSP do its job and reduce the noise in-camera.

Resolution, oversampling, details…

EOS R5 can shoot 8K and downsample this to 4K before writing it to the memory card. Canon is calling this high-quality mode (also known as ‘oversampled 4K’).

The benefit is amazing level of detail in the 4K footage which looks jaw-dropping. Once you see the difference between the oversampled 4K and regular 4K (line-skipped) it’s very, very difficult to go back.

Even though you realize most of your clients will watch these videos on their phones and will hardly be able to tell the difference between 1080p and 4K, there’s something really seductive in highly detailed footage.

Of course, it’s also practical – if you need to crop it, stabilize it or do any extensive post-processing on it, the extra details already present in it will proof it against heavier degradation and softness.

So, it’s all good, right? Weeeeell, about that…

There is only one mode that can capture 4K HQ footage, and that’s 4K at 25/30 fps. If you want to shoot 50/60 fps or higher, forget about it. You get the standard mushy 4K, which looks average at best. And I may be generous here.

Furthermore, you may have noticed that 4K HQ is downsampled from 8K capture. So yes, that means that your sensor (and also battery and heat dissipation management) are also shooting 8K, which many cameras are still struggling with, let alone a camera from 2020.

To put it clearly, you will overheat your camera much more easily and expend more batteries.

Now that we’re talking about video recording modes, let’s mention a hugely important feature – R5 is able to record (most) video to both card slots simultaneously! So, we finally get photographers’ twin card redundancy with all the peace of mind that implies, all without an external recorder.

Of course, there are some caveats: you need very fast SD cards (v30 or v60). Even then you can’t record all video modes to an SD card, which also means you can’t dual-record them to both cards. Demanding modes like 8K and 4K100/120 are CFExpress only. (There’s a table in the user’s manual detailing all the combinations.)

This means you can only dual-record some of the footage (which is still better than nothing).

The problem arises when you set dual card recording for redundancy in a supported mode (like 4K25) and then switch to something more demanding (like 4K100 for slow-mo) for an occasional clip.

What I would expect (and hope) would happen is perhaps a warning that camera switched to single-slot recording, allowing me to dismiss it quickly and get back to recording on a single card. Ideally, and now I’m just dreaming, when I switched back to a supported video mode, it would automatically resume dual-card recording.

Fat chance.

This is not at all what happens. Camera gives you a warning that you can’t dual-record, but then stops you from shooting anything! You’re forced either to revert to a supported mode, or go menu-digging to turn off dual-card recording.

I thought I could work around this by creating a custom shooting mode with 4K100 going to single card, and other modes for dual-card recording, but no dice. It turns out that for all the settings custom modes save, they do NOT save card recording strategy. It’s the one setting affecting all the modes on the camera!

Let me reiterate: each time you want to do ultra slow-mo, you need to turn off dual-card recording and then remember to turn it back on when done.

Better than nothing? I guess, but only slightly.

And now that we’re talking about ultra slow-mo, I will mention that for some reason Canon insists on saving this footage as actually slowed down video (e.g. 25 fps file for 100 fps recording) and doesn’t record any sound!

I’m not sure what the idea behind this was.

On Nikon Z6 I can shoot 100fps with normal sound and later decide whether I want to slow it down, how much and what to do with the sound. If my couple said something funny or romantic, I could decide to use the clip at normal speed and have this nice soundbite. I could also afford to sometimes forget I’m still shooting 100fps and yet get usable footage with sound.

Canon forces me to have the prescience of what will be happening in the future. I guess I could live with this, but I can see this biting me in the %$# few times a year.

It seems I’m not alone in wondering about the rationale of this approach, as Nikon made a subtle stab at Canon during their Z8 announcement, saying Z8 can record “4K120 with sound”. If I hadn’t just tested the R5, I’d probably never notice this reference.

Light-metering, base ISO with Clog

Exposing log footage is a skill and it’s the only way to get clean recordings using the full potential of Clog.

The basic idea is to force the camera to use the whole dynamic range available in log-profile by letting less light hit the sensor, allowing the log-gamma curve save extra information in the highlights which would otherwise get clipped.

The camera helps with this: it tells you that by activating the log-profile, your base ISO (lowest hardware ISO available) is now higher. In regular mode, base ISO is 100. With Clog it becomes 400, and in Clog3 it’s 800.

While this does nothing to the actual signal from the sensor, it’s telling the light-meter to confuse the user into saving more highlight information by underexposing by two stops (as if the ISO amplification was actually higher).

However, setting the exposure by what the evaluative metering now told me about the scene would consistently result in about 2 stop underexposure in Clog3 compared to optimal exposure for later color grading, which I don’t think should be happening.

In any case, I found myself constantly filming at +2 EV exposure compensation, and in some cases I went up to +3. This was worrying because R5 only allows -/+ 3EV compensation, so if for whatever reason I needed even more light (like shooting in snow, in a room with white walls or details of a white wedding dress), I would need to stop recording, switch from Av to M and then dial in the manual settings.

Hardly ideal for quick shooting style.

Okay, now we come to the ‘fun’ stuff. The first moment when I realized that I will have serious problems shooting weddings with R5.

The R5 does not have a live histogram during video recording. Let me repeat that. It shows a histogram in photo mode, it shows live histogram in video-standby mode… but press record and you’re on your own!

This was so shocking to me that I need to talk about my feelings here for a moment and vent a bit.

Live histogram is one of those essential features that anyone who isn’t working within a film crew needs all the time! (Or waveforms, ideally. But let’s not fantasize too much.)

In constantly changing lighting conditions, it’s essential to keep an eye on the histogram and continuously guide your exposure. If you add log-gamma into the mix, things get even more complicated and it’s even harder to judge proper exposure just by looking at the screen, especially in direct sunlight. View assist (a display-only LUT offered on R5) only adds another variable.

However, live histogram has become a normal, everyday thing. My iPhone has it, even my ancient Z6 has it. And in all the videos of people reviewing R5, talking about proper log exposure and discussing the histogram I never actually noticed that none of them were recording while displaying it!

So, when I started using R5, I spent hours poring over documentation, forums and googling aimlessly. I even asked ChatGPT which, true to its form, hallucinated a response that ‘of course it’s possible’.

Well, it’s not.

Why? My only guess is that it’s deliberate and Canon just wanted to protect its C-range of cinema cameras, starting with the R5C… but c’mon Canon, no live histogram in 2020? It’s not like we’re asking for waveforms or vectorscope!

After this shocking realization, it was of little comfort to learn that at least I had zebras easily accessible. You can map them to one of the customizable buttons (in my case I used the focus point selection key) and as long as you’re not recording you can toggle them on or off. Again, it would be more helpful if I could toggle them while recording to check exposure and then turn them off to actually see what I’m recording, but I guess it’s better than nothing.

The good thing is that the camera has dual zebras, so you can set zebra 1 to reveal 50% mid-grey tones and zebra 2 to show top 5% hightlights. The percentages are customizable so you can go wild. (I also need to mention that in some cases zebra 2 wasn’t showing at all, even though they were turned on and I most definitely had burnt out highlights. They later showed up and I never determined the cause.)

As mentioned in the photography part of the review, I set lens control ring as direct exposure compensation and it worked okay. The ring is fairly narrow (and close to focus ring) on RF 24mm, so it would take some practice to be able to grab it quickly and turn it reliably.

The ring is clicky, but if you turn it slowly, the noise it produces is very quiet. It would get picked up by an external mic during recording only in very quiet environments. Basically, nothing I’d worry about during a wedding day.

Stabilization

R5 boasts up to 6 stops of hybrid stabilization, which means using the sensor-shift IBIS in synergy with RF lens IS (if it’s available). This provides varying amounts of stabilization, depending on the lens in question. (Different lenses have differing imaging circles, which in turn limits how effective sensor-shift IBIS can be. This is why RF lenses have advantage over EF lenses – they are designed for a larger mount a give a bigger imaging circle.)

One of the reasons why I wanted to try out a native RF lens for video was to see what’s the best stabilization I can get in a Canon camera.

Long story short – it works great! The difference comparing to Z6 with adapted glass is huge. Shooting run ’n’ gun style and getting reasonably steady footage that doesn’t need additional stabilization later on finally looked achievable!

Of course, your mileage may vary depending on exact lens used and your shooting style, but I’d say we’re definitely getting there.

Various ergonomic observations

Let’s now do a round of quick impressions.

When shooting in direct sunlight, LCDs are difficult to see and recording video via EVF usually produces less than stellar results, mostly due to the way you hold your camera.

It helps a lot if you bump the LCD brightness to maximum, but on most cameras this requires a lot of menu digging. When going from interior to exterior and back frequently, I don’t bother with this. I would, but don’t have the time to do it. Maximum brightness inside will drain the battery quickly and it’s more difficult to expose properly when the LCD is exploding with brightness, even with a histogram.

Canon, however, has a simple but clever solution. They allow you to map a custom key to temporarily toggle maximum brightness. It’s easy, quick and helps a great deal, both in seeing the LCD in sunlight and saving battery when you don’t need it. The best part is that you won’t forget to dial down the brightness because this is just a temporary bump in brightness.

I mapped it to the ‘light’ button which is easily accessible during recording and I just loved this function!

On the topic of battery life, Canon has several modes of battery saving in regards to both sensor readout and the LCD display.

In photo mode you can opt for lower refresh rate to conserve power. In video mode there is also something akin to low res readout during standby.

These are all menu settings, but there is one setting called ‘eco mode’ which you can activate directly via a custom key (I chose the ‘set’ button). When activated, camera darkens the LCD very quickly (after 2 seconds) and turns it off soon after. It also enters standby sooner than usual.

It’s great for moments when you have pauses between shoots because not much is happening, making it easy to jump back to shooting by just touching the shutter button. You can also go back to full power mode with a single press of the shortcut key, without any menu digging.

Eco mode somewhat helps with the fact the power switch for the camera is on the opposite side and needs a second hand to turn on/off, which makes it laborious to toggle power frequently to save the battery or keep it cooler (which is easy on cameras which have power switch near or around the shutter button).

Final grievance I have is with the fully articulating screen. Yes, again.

I talked about this in photography part of this review, but now I have to touch upon it again for a different reason.

As you realized by now, I shoot run ’n’ gun with wide open apertures. I don’t use gimbals, steadycams or anything similar as it would be hard to focus at f/1.4, so I’m shooting handheld, but with the goal of producing pretty stable footage (swaying instead of shaking).

I try to achieve this by keeping the camera steady using the camera strap adjusted just-so and stretched taut between my neck and semi-extended arms. This greatly helps with keeping it steadier and floating, instead of shaking.

For more comfortable holding position during a long day, I keep the camera slightly below my eyeline, with the screen flipped out so I can look at it from above.

With Nikon’s single-axis articulating screen this works great as it just moves a bit out of place, but stays on the camera’s central axis. It’s not sticking out so much it would easily get damaged.

However, Canon’s fully-articulating screen cannot do this. To provide the fully-articulating functionality it needs to be flipped out to the camera’s side, and then you can turn it any way you like. Obviously, this means that it sticks out to one side, increasing the footprint of the camera, and even worse, the stretched camera strap goes directly across the middle of the LCD so I can’t properly see it!

Furthermore, using the camera strap with LCD extended like this is a recipe for disaster – getting entangled and breaking it off sounds probable at some point.

Overall conclusion

It was a really long review, so I’ll keep this short.

I loved the quick and precise AF on RF lenses along with many ways to customize the controls, shooting modes and camera menus. Canon offered some really clever quick shortcuts like direct access to eco mode, maximum screen brightness and zebras toggle. I also instantly fell in love with 10-bit Clog3 footage in 4K HQ mode; with its colors, details and wide latitude it offers for color grading.

But despite all of that, the R5 had a few things that I just couldn’t get over for professional use.

No continuous AF on-demand and no way to quickly toggle face detection while recording, no live histogram during recording, the mess with dual-card recording in non-supported modes, the screen that must be flipped out sideways to use it at an angle, heavy underexposure in Clog and no audio in high FPS recording – all of these amount to too many compromises for my particular shooting style.

Any one of these would probably be okay on its own, but taken together they turned out to be a dealbreaker for me. Again, most of these are software issues fixable with a firmware, but if they weren’t fixed in last 3 years, I doubt they ever will be.

Hence, I’m moving on to Nikon Z8 and looking forward to creating a detailed comparison of R5 vs Z8 later in 2023.

Stay tuned!

Note: all images in this article were captured on R5 with RF 24mm f/1.8 in raw mode and processed in Adobe Lightroom.

Back to Part 1: Photography review of R5