Canon Interview CP+2025: Action Priority AF, Lenses, and a HUGE software move by Canon!
posted Wednesday, April 9, 2025 at 2:30 PM EDT
I sat down once again with Canon, Inc's top executives at the CP+ 2025 photo trade show in Yokohama Japan earlier this year, and it was one of the most compelling interviews I've had to date. It feels like Canon is entering a whole new era, based not only on their technical expertise and how well they've leveraged AI technology, but also thanks to an entirely new strategic initiative led by their younger staff members vs senior executives.
AI's integration into cameras and the incredible optical advancements in recent years make this one of the most exciting times ever for the industry as a whole, but Canon is not only leveraging their core technologies but truly adapting to the new realities of the market – more than I could have imagined only a few months ago. Canon's new Action Priority autofocus modes are the most advanced use of AI technology we've seen in a camera to date while at the same time their new and totally unexpected Live Switcher Mobile iPhone app could do more to engage the youthful "Creator" market than anything I've seen yet. Read on for all the details!
I met with the following people, many thanks for their kindness in sharing an hour of their tightly-booked time with me:
- Go Tokura, Head of Imaging Group, Chief Executive, Image Communications Business Operations
- Manabu Kato, Unit Executive, IMG Business Unit 1, Imaging Business Operations
- Yasuhiko Shiomi, Advisory Director, Unit Executive, IMG Development Unit, Imaging Business Operations
- Tetsushi Hibi, Advisory Director, IMG Optical Business Unit, Imaging Business Operations
The executives for this interview, from left to right:
Yasuhiko Shiomi, Leader for body/main system development
Go Tokura, overall executive for Canon's imaging business
Manabu Kato, manager of business unit for photo and video
Tetsushi Hibi: Leader for optical design
Could Cross-type AF come to lower-end models?
DE: The Cross-type AF is a major new feature in the EOS R1. It seems to me that it might not increase production cost that much, as I'd think they would only need the same number of mask layers during manufacturing. Since your CMOS sensor technology allows per-pixel readout anyway, and leaving out the R1's Action Priority AF modes, perhaps all that Cross-type AF needs is just a slightly faster processor to handle both AF directions. Is that the case? Does this mean that Cross-type AF could show up in lower-end cameras over time?
Shiomi-san: The manufacturing process for cross-type AF requires a very specific sensor and dedicated manufacturing equipment, making it more costly and complex than you might think. However, customers have given us excellent feedback about its performance, so we are considering expanding it to lower-tier models in the future.
How do the AI-based Action Priority AF modes work?
RDE: The R1 and R5 Mark II's AI-based Action Priority modes for sports shooting seem like a true breakthrough, in that the camera "understands" so much about what's going on compared to any AF system I've seen before. At the most basic level, it seems like it's mainly selecting the player who has possession of the ball and then using eye priority for the specific point of focus. Is that a fair statement?
Shiomi-san: That's not quite the complete picture, and we can't reveal too many technical details. What I can say is that the system analyzes the spatial relationship between players and the ball, with particular attention to the player's joint positions and their distance from the ball. The camera uses this information to detect when a player is performing an action that merits focus priority.
Tokura-san: This is actually a fairly common approach in computer vision -- looking at human skeletal forms to recognize postures and actions. This is a standard way to interpret human movement in visual systems.
RDE: The Action Priority modes seem to understand certain specific plays, such as the set/spike in volleyball or an "alley oop" in basketball. Are these plays specifically trained for, or is it just that the camera is very good at following the ball as it's handed off between players? How does it behave differently in the case of (for example) a goal save vs a sliding tackle in soccer?
Shiomi-san: Taking basketball as an example, we understand the different scenarios in the game -- how goals are scored, how players move, how they dribble, and so on. Our training approach involves teaching the system to recognize various ball movements and player postures separately, then integrating the two types of information together.
Regarding your question about soccer, when two different actions occur simultaneously -- like a sliding tackle and a goalkeeper making a save -- the system prioritizes the player closest to the ball. This prioritization framework is a key part of the system's training.
Action Priority autofocus understands the games?
RDE: So the camera is really watching the game and kind of understanding what's going on? In between exposures, it's watching the movement of the ball and the action of the players. It's a continuous process.
Shiomi-san: Yes, exactly. The system is continuously analyzing the live view feed, constantly looking for motion patterns that match the action categories it has been trained to recognize.
How does the DIGIC Accelerator coordinate conventional and AI AF?
RDE: These new Action Priority modes are thanks to the new DIGIC Accelerator. I know you can't describe how this functions in too much detail, but can you tell our readers anything about it? It seems like it's most likely doing sophisticated, multi-frame AI inference based on what it sees in the stream of images from the sensor, is that the case?
Shiomi-san: We've integrated the AI capabilities with our traditional tracking methods. Since not every scene contains recognizable action, when the system doesn't detect specific actions, it reverts to conventional subject tracking -- looking for heads and eyes. Both approaches work together seamlessly.
RDE: Is it an either/or situation, in that either the DIGIC Accelerator is running AF operation or the conventional AF processor is but never both, or do the two work cooperatively?
Shiomi-san: Let me clarify -- the DIGIC Accelerator ultimately makes all autofocus decisions. Within the accelerator we have processing for advanced action recognition and what we previously called “conventional tracking”, but it’s a unified system, not two separate systems working in parallel.
RDE: So there's not a separate conventional AF system, there’s just the one DIGIC Accelerator chip.
Shiomi: That's correct. There's just one AF sensor, and the DIGIC Accelerator processes all the information from it. While there are two processors in the camera, the AF system itself is unified under the DIGIC Accelerator's control.
How much of their business is APS-C?
RDE: You now have a very full APS-C lineup. Can you say anything about how active that market is for you? (I know you can't give specific numbers, but just in terms of the order of magnitude of the number of APS-C bodies vs FF bodies, is it 10:1? 5:1? 100:1? A lot or a little?)
Kato-san: The APS-C market is currently very dynamic, primarily because these cameras offer an excellent balance of affordability and performance. Our R50 and R100 models introduced in 2023 have become global bestsellers. We intend to continue expanding and diversifying this lineup to meet the varied needs of our customers.
Regarding the comparison between full-frame and APS-C markets, while I can't share specific figures, I can tell you that the ratio is more favorable to full-frame than any of the ratios you mentioned (10:1, 5:1, or 100:1). While the APS-C market is significant, the balance between APS-C and full-frame is more tilted toward full-frame than the ratios you suggested. [Ed. note: Wow, I had no idea the balance was that lopsided towards full-frame!]
Interestingly, the proportion of full-frame cameras has actually increased in the mirrorless era compared to the DSLR era. This is because many first-time camera buyers are now choosing entry-level full-frame models like the R8 and RP, which have become much more affordable.
RDE: So the low end of full size is competing with APS-C, and many go with full size from the beginning?
Kato-san: One of the main factors driving this trend is that entry-level full-frame cameras have become much more affordable. It's now possible to purchase an EOS RP for less than a thousand dollars, which makes full-frame much more accessible to new users.
RDE: I was surprised this time, in that every year in the past I would ask you about the possibility of a really affordable interchangeable lens camera. And now Canon has actually done that, with the lowest priced APS-C of anyone. I was surprised to see this happen after so many years, I was surprised to see that happen and I think it's a good thing for the industry.
Kato: Thank you. We've been listening carefully to customer feedback. We recognize the importance of offering more affordable options, and we're committed to continuing to address this market need.
What breakthroughs were needed for the new 70-200mm f/2.8 L IS USM Z lens?
RDE: The "Holy Trinity" of lenses was completed back in 2019 with the RF70-200mm F2.8 L IS USM and now you’ve come out with the RF70-200mm F2.8 L IS USM Z with improved specs. What was the specification or combination of specifications you were driving for that made it hard to bring this new version to fruition? What was the key breakthrough that let you bring it to market?
Hibi-san: One of our primary goals for this lens was to enhance its compatibility with video shooting. This presented several technical challenges we needed to overcome.
First, we needed to implement power zoom functionality to enable smooth, controlled automatic zooming.
Second, we had to minimize focus breathing -- the change in field of view that can occur when adjusting focus.
Third, we had to solve the problem of focus shift during zooming. With this wide focal range of 70-200mm, maintaining precise focus throughout the zoom range was particularly challenging. [Ed. Note: This is the definition of a true zoom lens rather than just a variable focal length lens which shifts focus when zooming.]
These were the key technical hurdles we needed to overcome to bring this lens to market.
Is there really a market for stereo ("3D") lenses?
RDE: 2024 was a very busy year for lens announcements! I think you now have more than 50 lenses for the RF mount, including the cine models. Do you feel that you have the basics covered now, and so will be focusing on more special-purpose lenses? (Two examples are the 3.9mm fisheye and 7.8mm "Dual" (stereo pair) lenses for your APS-C models. I'm very curious about those; is there a significant market for stereo-pair image capture for VR?)
Kato-san: We intend to continue expanding our lens lineup. While I can't share specifics, we still see gaps in our long focal length offerings. Plus, it's been about seven years since we launched our first RF lenses in 2018, so we think it’s time to upgrade the technology with second generation models.
Regarding specialized lenses like the 3D/stereo options -- we're committed to this category despite the market still being in early stages. We believe the 3D space will grow significantly in the future, and we want to be at the forefront of this development.
I should also emphasize that plan to expand our range of affordable lenses as well.
The future for third-party lenses?
RDE: I (and many of your users) greatly appreciate you allowing third-party AF lenses on the RF platform! The selection still seems limited though. It's currently APS-C only; are there any prospects for third-party full-frame AF lenses?
Tokura-san: As in previous years, I'm afraid we can't disclose details about our agreements with third-party manufacturers. However, as you've likely noticed, the range of third-party lenses continues to expand each year, though we don't closely monitor which specific models are being released.
I would suggest keeping an eye on what third-party manufacturers are introducing to the market. And I suspect we'll be having this same conversation again next year.
[Ed. Note: Tokura-san has been very patient with me asking this question every year :-) The teaser about keeping an eye on third-party announcements this year is encouraging though!]
Canon goes iPhone? A totally mobile-based live-broadcasting app!
RDE: I was very surprised to see you release the Live Switcher Mobile app last month for controlling multi-camera live streams. It's currently iDevice (iPhone and iPad) focused, and doesn't involve any Canon hardware in its use. What was the motivation to develop it, and what does it show about Canon's strategy going forward?
Kato-san: I had wanted to ask you if you were surprised at seeing this! We are now working with the mindset that we should focus not only on hardware but also on software. With that in mind, we are fully committed to providing solutions beyond just apps and are moving forward with a strong determination to do so.
As you know, the demand for live streaming capabilities is enormous, particularly among younger users. We developed this app for two main reasons: first, to meet the market need for accessible multi-camera livestreaming, and second, to introduce Canon as a brand to users who might later become interested in our cameras and lenses.
Tokura-san: It’s important to note that this project might never have happened if it had been left to those of us around this table (that is, the senior managers). It actually originated from our younger employees, who proposed the concept and championed it through development. This represents a new approach to product development for us.
RDE: Yes, I can see it as being a kind of brand ambassador product to all the streamers who know only smartphones and have little awareness of Canon or even dedicated cameras in general. That hadn’t occurred to me, but it makes a lot of sense now that I hear it.
[Ed. note: This is particularly significant for Canon. They've always struck me as a very traditional Japanese company, with all that implies about conservative decision-making and a strongly heirarchical culture. Live Switcher Mobile entirely breaks that mold though; it's a very strong move into software vs hardware, is far more open than I've ever seen a Canon product be (not even requiring a Canon camera to use it), and jit umps squarely into the middle of modern, mobile-only content creators' workflows. The fact that senior management is paying so much attention to and so strongly supporting the ideas of younger staff bodes very well for Canon's future.]
Will people actually pay for the full version of Live Switcher Mobile?
RDE: Live Switcher Mobile seems very capable, making it easy to do things that previously would have required a computer and video switcher to accomplish. Unlocking its full capabilities does cost $17.99/month in the US though, while most live-streamers use the free OBS software. (Which does require the use of a computer though, and isn't nearly as intuitive to use.) What's been the initial uptake, and in a very rough percentage, how many are using the free version vs paid?
Kato-san: While I can't share specific user numbers, the initial response has been very positive. As to free vs paid, it’s hard to know where to draw the line, but for now we want as many people as possible to try it, so we decided to offer all the basic functions for free. Strategically, we hope the app will increase awareness of Canon among smartphone users and that will lead to higher revenue from our cameras and lenses. Of course it would be good if the software made a profit, but the main goal is to get people interested in Canon and our camera lenses.
[Ed. note: Note the emphasis on lenses here; I think they rightly recognize that the main reason many streamers might want to use dedicated cameras is for creative control via lens choices, rather than basic image quality or advanced video capabilities.]
Will Canon cameras be integrated into Live Switcher Mobile at some point?
RDE: Are there plans to incorporate Canon cameras into Live Switcher Mobile as possible video sources? Or is this an entirely separate direction for Canon going forward, as a subscription-software/SAAS vs a hardware provider?
Kato-san: Yes, we are definitely considering integration with Canon cameras.
RDE: How? I'm trying to think how that could happen because the cameras are not sort of set up to feed video to a smartphone.
Shiomi-san: We would need to update our camera firmware, but both our cameras and smartphones already have WiFi capabilities, so the technical foundation is there. The key would be implementing the proper protocol to enable direct streaming from camera to smartphone. The RTMP (Real-Time Messaging Protocol) standard is already well-established for streaming video content.l
[Ed Note #1: Duh. Of course the cameras already have fully capable WiFi transmitters, all that’s required is a firmware change. - And of course there’s an established protocol for it, apparently going all the way back to a patent filed by Adobe in 2003 and granted in 2007.
[Ed. Note #2: Canon actually beat me to it by a couple of weeks, launching their new EOS R50 V on March 26. Aimed at video creators, this updated version of their budget-friendly EOS R50 drops the original's EVF in favor of a fully-articulated LCD screen as you'd expect to find on a vlogging-oriented camera. They also tweaked focus modes, added log-video recording capability and mic/headphone jacks. Most significantly, though, the new EOS R50 V can act as a source camera for Live Switcher Mobile. This provides a fully-Canon, easy to use live-streaming platform for creators who want to move beyond smartphone and tablet cameras in their productions. This strikes me as a remarkably progressive and intelligent move by Canon, erasing the boundaries between "real" cameras and smartphones for the live-streaming and video-creator crowd. Two thumbs way up!]
Summary
Out of all the many times I've talked with Canon executives about new developments, this interview feels by far the most significant; even the initial introduction of the RF mount pales in comparision. The reason is that Canon has met their users where they are in two incredibly important areas.
Sports shooters have always been core customers for Canon's excellent autofocus, with many evolutions and even revolutions over the years. The new Action Priority autofocus modes mark an entirely new era in camera intelligence. AI-based autofocus was a revolution in and of itself, moving from simple pattern recognition with face- and eye-based AF that's been with us for decades now to being able to recognize a wide range of subjects and the critical focus target for each.
Canon's Action Priority goes far beyond all that though, by tracking and understanding what's going on in a scene. For subjects like soccer and the other ball-related sports Action Priority supports, the camera is watching as the play evolves, aware of not only the ball position but the detailed actions of the players. It understands arm, body and leg positions relative to the ball and knows when a key kick, block or spike is in the process of happening. For the first time, the camera itself understands the "decisive moment" of Henri Cartier-Bresson, at least in the world of sports. This is just the beginning; it will surely expand to cover other sports as well, and I think in the future it will even extend to things like human expressions and actions. In the meantime though, even if you're "only" interested in sports shooting, Canon's Action Priority modes for autofocus are a true technological breakthrough.
The second huge development is Canon's move into a purely software product aimed at the generation of video creators who work entirely on their mobile platforms. It's hard to overstate how big a shift this is for a company like Canon, whose stock in trade is the hardware side of photography and videography. Rather than sitting at a distance and hoping that creators will want to use a "real" camera enough to put up with the pain of integrating one into their workflow, Canon has met users where they are. providing an easy to use solution for multi-camera video streaming. Even the free version is capable and worth using, and the paid solution brings valuable additional features at an affordable price while still maintaining the ease of use that will be key for so many mobile-only video streamers.
While the initial release of the new Live Switcher Mobile app only supported smartphones and tablets as input sources, Canon has now also released the EOS R50 V, a full-capability mirrorless camera uniquely tailored to the needs of video creators, and able to act as a video source for the Live Switcher Mobile platform.
This is an incredibly significant move for an old-line camera company, stepping directly into a world they've been an awkward fit for in the past. (As has been the case for all the other mainstream camera makers as well.) It was interesting to note Tokura-san's comment that the Live Switcher Mobile would never have existed if it had been left up to the senior executives to choose a path. It was a younger generation of Canon employees who saw the market and pushed to address it, and huge kudos to the senior management who listened to what they were proposing.
It's always interesting to peel back the curtain a little and see what's going on in the minds of photo/video industry leaders, but this was one of my most surprising and enlightening interviews ever.
(What do you think? Let me know in the comments below; I'm genuinely interested to hear your thoughts, so I'll watch the comments for a while after publishing this.)