The Sony FX9: The Camera That Refuses to Die

So for my first rant that I’m finally posting to this site, let me rant about a camera.

If you read reports about the cameras that are being used in big budget cinema and television, you’ll see some similar themes: Arri is still king, film cameras are cool again, and the Sony Venice is gaining a growing foothold. And to a large extent, that’s all true. Many of the narrative and commercial sets I’ve worked on have used either the Alexa Mini or the newer Alexa 35. Even low-budget indies will often break the bank to rent an Alexa Mini, because the DP insists on shooting on one.

But much of my work lately here in the DC area has been in unscripted TV and other corporate shoots. And in that space, the Sony FX9 still, to this day, reigns supreme.

The kinds of shoots we’re talking about here are ones where takes are recorded for anywhere from several minutes to hours at a time. We’re either shooting interviews with two cameras at locked angles, or going run-and-gun with operators running cameras on their shoulders. Given the extended length of these takes, production companies and their internal post houses put a strong priority on keeping file sizes in check, so shooting on highly compressed codecs, such as Sony’s XAVC or other AVC/HEVC-based codecs, is usually preferred. And somehow, in this year of 2026, many of these shows are still being mostly shot and distributed in 1080p HD.

So it’s no surprise that after its 2014 release, the Sony FS7 (later followed by its Mk. II revision in late 2016) took this sector by storm. It had a good Super35-sized 4K sensor, integrated XLR audio inputs (usually to embed a mix out from a separate audio mixer), SDI video outputs, an included eyepiece for the side-mounted monitor, integrated ND filters, and a positive-locking E mount that allowed the use of cheaper photography lenses made for Sony’s mirrorless cameras, which were also a new product category at the time. It also used the relatively new XQD cards, which were faster than SD cards while still being relatively affordable, compact, and usable with USB card readers (mostly produced by Sony themselves). And it also helped that the camera its self, at roughly $10k in the USD, came in at a price that freelance operators could justify buying the camera for themselves and charging their own rental rates to the jobs they bring the camera on. Even to this day, the FS7 is getting quite a bit of use on smaller-scale unscripted TV shows (think your typical TLC or HGTV fare).

In the fall of 2019, Sony followed up the FS7 with the FX9, which followed much of the same formula, but moved to a full-frame 6K sensor (though you can still only record in 4K, downscaled from the native 6K), upgraded the SDI outputs to 12G (up from 3G), a dual 800/4000 native ISO, a new Cine EI exposure setting, an upgraded electronic ND filter, internal Wi-Fi for remote control, and other smaller improvements. Though in terms of form factor, control layout, price point, and using those same XQD cards (and oh boy, I’ll get to this later), it still follows the same general formula that the FS7 did. It didn’t take long for the FX9 to become the new standard in the same market that the FS7 once held (though its adoption was certainly slowed a bit by the COVID-19 pandemic), and owner-operators across the board adopted the new camera. It also later gained the ability to shoot in ProRes RAW, but this required the use of both an external Atomos recording monitor, as well as a rather massive extension unit for the camera that cost more than double the price of the recorder.

Operators that I’ve worked with absolutely love the FX9. Even though smaller and lighter cameras have come out that are arguably more capable (including, to some extent, Sony’s own FX6), they still prefer the weight distribution of the FX9 body, and they love the fact that the included monitor eyepiece (despite being somewhat fiddly) and control grip means that, aside from a left hand grip and a bottom plate, it comes ready to go out of the box for shoulder-mounted operation.

Sony later expanded the FX series in 2020 with the FX6, which took the full-frame 4K sensor from the a7S III, and packed it into a body that was essentially a smaller version of the FX9; features missing from the FX6 include the second SDI output, the genlock input, the positive-locking E mount (a standard E mount from the mirrorless cameras is provided instead) and the integrated XLR inputs (XLR audio can still be used with the included top handle attached). The XQD card slots are also swapped out for CFExpress Type A, which I would consider to be an upgrade over the FX9. The FX6, with its more affordable price point, proved to be an even bigger sales success for Sony, though none of the operators I know ever bought one.

Then 2021 brought the FX3, which was essentially a variant of the a7S III in a body better suited to cinema workflows (though still in a mirrorless-style form factor), and effectively became a successor to the a7S line as a whole. Some of the FX9 owners I know bought FX3s as B-cams, especially for use on gimbals. The FX30 (like the FX3, but with the a6700’s smaller sensor) and FX2 (based on the A7 IV, with an added EVF and mechanical shutter for photos) would follow in the few years after, though these never saw as much professional use.

In early 2025 though, Sony discontinued the FX9, now over four years old, with no immediate successor. There’s been the occasional rumor about an FX9 II, but as of writing this in September 2026, no such camera has materialized. Sony announced the Burano at IBC 2023 with a feature set clearly targeting much of the same use case as the FX9, but at two and a half times the price of the FX9, it’s in another league entirely.

Even approaching its sixth birthday, there’s a lot to like about the FX9, and it remains a highly capable camera. For one, its native ISO values at 800 and 4000 are more useful in most scenarios than the FX3 and FX6, which have a massive gap between the native ISOs at 800 and 12800. Its 6K sensor also allows for a slightly cleaner downsampled 4K output, and that extra resolution means that you can crop down into Super 35 mode and still get a native 4K output. Given the widespread use of S35 cinema zooms, many of which offer focal ranges unmatched by full frame lenses, this is incredibly useful, even in the unscripted television space where vintage and more characterful glass is usually shied away from. And yes, it has to be pointed out how the ENG-style form factor is very friendly to run-and-gun operators. A good 1st AC will know how to build a camera package to fit any operator’s preferences, but this fits the bill for these kinds of shooting scenarios nearly right out of the gate.

But after years of working with the FX9, boy do I have a few bones to pick with it. I’ll start with the least annoying, but go up from there.

The smallest gripe is probably the lack of support for Cooke /i support when using a PL mount adapter (all of which are third-party, as Sony doesn’t offer their own). Granted, that’s probably less essential for the kinds of shoots this camera will be used on, but it’s still nice to have the camera be able to receive metadata from the lens. I bring this up because Canon, who also targets similar use cases with their cinema cameras, offers their own RF to PL adapter for their full current cinema lineup (even the mirrorless-style C50) that has /i support, complete with screw mounts to secure the adapter to the camera. Red also offers similar /i-supporting PL mount adapters on their Komodo and V-Raptor cameras, attaching to both RF and Z mounts on the camera body. To be fair, this may not be possible on the E-mount, as shown by the fact that Sony’s Burano and Venice 1 & 2 cameras use an extra set of pins outside of the E-mount its self to support lens data passthrough on both the included PL adapter, as well as third party LPL adapters.

I’d also prefer if the video outputs, when showing status info on the display out, would put that info text in black margins around the image, rather directly atop the image. This would make that info text a bit easier to read, and also make it easier to extract a full, clean image out of the feed for purposes such as live grading and using scopes. This is something that the new FX5 has fixed, as it mimics the behavior of Sony’s higher-end CineAlta cameras (such as the Venice and Burano) rather than its FX-series predecessors.

The next slightly more annoying issue is that the camera has no way to apply volume names to media cards. When offloading an XQD card from an FX9, it simply shows up as an unnamed volume This means that when several cards from multiple FX9s show up on my DIT cart on a multicam shoot, this means that I have to insert each one, one at a time, then rename the card to something unique, eject the card, and reinsert it before repeating the process on the next card. This gets to be incredibly tedious, especially when Sony’s own Burano and even the FX5 have options to apply volume names based on the reel number, and most other cinema cameras do this by default.

Now for the biggest gripe of all. We need to talk about the XQD cards.

XQD was still a relatively new format when it appeared way back on the FS7. Nikon was the first to support it on some of their higher-end DSLRs, but then Sony got in on it with the FS7 and a few other camcorders. It uses one lane of PCIe 2.0 (the standard was later updated to PCIe 3.0, but I’m not aware of any card that ever reached into those speeds), with Sony’s cards supporting up to 440 MB/s of transfer speeds. This remains faster than even the fastest SD UHS-II cards (though the new, seldom-used SD Express is faster), but slower than the fastest speeds that CFast could offer, which XQD was under the same umbrella with at the CompactFlash Association. But the potential was there, with PCIe offering much higher theoretical headroom for performance compared to the older SATA-based CFast cards. The card design also incorporated a partially metal casing to help with heat dissipation.

But by 2017, the CFA took that potential along with the NVMe standard finding its way into desktop computers, and published the new CFExpress standard. The initial card format (later redefined as Type B) was physically identical to XQD, but now supported two lanes of PCIe 3.0, pushing data rates to upwards of 1800+ MB/s on the fastest cards (which has since doubled thanks to the introduction of CFExpress 4.0, which uses PCIe 4.0). The standard reached critical mass in terms of widespread adoption when version 2 of the standard was published, which defined smaller, 1-lane Type A cards and larger, 4-lane Type C cards (as of 2026 though, no one has adopted Type C). Nikon even published firmware updates for most of their XQD-supporting cameras to allow them to use the new CFExpress cards. Sony, however, never released any such updates for their cameras, despite the fact that the CFExpress 2.0 standard had been published several months before the FX9 was released. Sony would even go on to adopt the CFExpress Type A format in their photo cameras, as well as the rest of the FX series starting in 2020 (so far being the only camera manufacturer to adopt Type A), along with Type B in the Burano in 2023.

The first issue, of course, is with the transfer speeds. Sure, the write speeds are more than sufficient when it comes to the demands that the camera puts on it, but the offload speeds are a mere fraction of what CFExpress Type B, or even Type A can support. It feels borderline insulting when a microSD Express card (thank Nintendo for that being an actually available standard) that I can put in an action cam offloads at least 60% faster than the XQD card, or when the CFExpress Type A cards out of the FX3 offload two to four times as fast. That adds up considerably at the end of a shooting day, especially when multiple cameras are involved.

The second issue is with the readers. CFExpress card readers are available from every manufacturer in the flash storage card business, prices range from $45 for cheaper USB3 readers, up to $100-$150 for readers that support USB4, and therefore have enough headroom to offload Type B 4.0 cards at full speed. When it comes to XQD though, there’s only one option for a USB reader, directly from Sony. It costs $100, and also supports CFExpress Type B cards in the same slot. There’s also two third party readers using Thunderbolt 3; one from Blackjet costing $60, and a dual-slot reader from Sonnet costing a whopping $350 (and I own one, for some reason). Since these readers though, due to using the PCIe tunneling found in Thunderbolt, have the card make a direct connection to CPU, on top of the fact that it uses a proprietary protocol that predates NVMe, you have to install an obscure driver for either Windows (x86 only) or Mac OS (x86 and ARM). However, since newer USB4 CFExpress Type B readers are also using PCIe tunneling, they also work with XQD cards when running in USB4 mode (at least it does with the Nextorage reader that I’ve tested) provided that the driver is installed, which those readers’ manufacturers were completely unaware of when I mentioned it to them.

Lastly, and most importantly, there’s the issue of available capacities of the cards themselves, and just availability in general. XQD cards were never produced in capacities larger than 256GB, and Sony was the only manufacturer to continue producing XQD cards after CFExpress first hit the market. The final generation of XQD cards from Sony came in 120GB and 240GB capacities. Today, only the 120GB card remains in production and available for purchase. When recording XAVC-I in 1080p at 23.98 FPS, this lasts for about 2 hours and 18 minutes. In 4K UHD, it lasts just 48 minutes. On a recent show I worked on that ran up to five of these cameras for upwards of 8 hours a day, all with only one AC on most days, this made us dependent on the fact that the camera had two card slots, and could instantly switch slots while rolling if the active card filled up. And that’s all just from shooting in HD; had this show been targeting 4K acquisition (or if any future season calls for it), the card capacity issue alone would disqualify the FX9 for our purposes. XQD as a media format is on its last gasps of life support, and the cameras that use it will become increasingly difficult to continue using once it decides it’s time to condemn that production line to the same fate as it did for physical game discs for the PS5.

Relying on dual-slot recording like this also opened tons of room for error. For one, operators would often accidentally press the slot select button, causing clips to be non-sequentially scrambled across two cards, which our 1st AC had to address by taping a cardboard flap over the button. Also, this opened the possibility of cards being pulled from cameras without immediately knowing if both cards had been written to (we lost a card this way at one point). Ideally, only one primary recording card should ever be in the camera, and when that card is changed, that’s something that can be logged in the camera report so everyone knows what footage is where.

So what’ll it be that finally knocks the FX9 off its perch? For some uses, the FX6 may be enough and at least solve the XQD problem, but it still has the aforementioned drawbacks compared to the FX9 that could be a deal breaker.

In terms of outright capability, the Burano solves just about every issue I have with the FX9, but at $25k, the price will be a tough sell for owner-operators and production companies.

The Canon C400 also got my attention when it was first unveiled. I haven’t had the chance to work with one yet (I’ve only seen it at a trade show), but at least on paper, it seems to beat out the FX9 in just about every metric. It has three base ISOs, records 6K RAW (or 4K RAW at Super 35) internally in 3:2 open gate, adds an ethernet port, and crucially records to CFExpress Type B. However, given that it’s Canon, those used to Sony workflows and neck deep in E-mount lenses could find that to be an insurmountable hurdle. Production companies also tend to be used to asking for cameras by the brand name rather than what it can actually do for them. It also lacks any sort of viewfinder or eyepiece, but a third-party external viewfinder, such as Kinefinity’s Eagle SDI, can mostly solve that issue.

The camera that seems most likely to replace the FX9 is one that doesn’t exist yet, but one that feels inevitable that Sony will make after the recent release of the FX5. That camera succeeds the FX3 with some long awaited features such as internal X-OCN recording (with new, smaller C1 and C2 variants to boot), 3:2 open gate recording, more sensical info overlays, three base ISOs with a dual gain option at ISO 800, and much more. Plus, given that it’s a 5K sensor, the 4K scan mode will still fit within the image circle of many S35-rated lenses, or there’s a 3.2K S35 mode that could be reasonably upscaled the rest of the way. The FX5 sits in nearly the same mirrorless-style body as the FX3, but one would certainly think that Sony will take sensor and processing hardware, and put it in a larger, better equipped body such as that of the FX6, or possibly even bridging the gap between that and the FX9. But this is Sony we’re talking about here, so just because something seems so obvious for them to do, it doesn’t mean that they’ll actually do it.

Anyways, that’s enough ranting about a camera. It still has some life left in it, and it’s an investment that has paid off for my friends who own them, being so lucrative that I know at least one person who bought two. But its dominance will eventually have to come to an end, and as someone who has to deal with the video feeds and footage that comes out of it, that time can’t come soon enough.