Caution before buying a DSLR for video!

Since DSLRs or DSLTs are no longer made, I don't understand the logic of caution when buying one. Do you mean buying one as an antique for commemorative reasons? But in the late 2000s, video DLSRs (mainly the Canon 5D Mark II) were earth-shattering. When it was released in 2008, no professional video or cinema camera on earth used a full-frame sensor.

In 2008, the least expensive video alternative to the $2,699 5D Mark II was the $17,500 RED ONE, which used a Super 35 sensor. When the first demo video "Reverie" was released (BTW edited on Final Cut Studio), it was revolutionary. It was shot in one weekend on a pre-production 5D Mark II by a photographer, not a professional videographer.

The 5D Mark II recorded 1080p/24 or 1080p/30 MP4 at 38 megabit/sec, whereas the Sony SLT-A58 recorded 1080p/24 at 24 megabits/sec in AVCHD. Using MP4, the SLT-A58 could not record full HD but 1440 x 1080 at 12 megabit/sec. So there was likely a significant video quality difference between the SLT-A58 vs the 5D Mark II, especially if using MP4.

Whatever video image quality the 5D Mark II and 7D DSLRs produced, it was sufficient for the following feature films and TV episodes:

Feature films shot on the 5D Mark II or Canon 7D DSLRs, either mostly or partially:

• Black Swan (2010) – Used Canon EOS 5D Mark II and Canon 7D for select sequences.
• Iron Man 2 (2010) – Used Canon EOS 5D Mark II to capture parts of the Monaco Grand Prix race sequence.
• Captain America: The First Avenger (2011) – Used Canon 5D Mark II cameras for specific action sequences.
• Act of Valor (2012) – Canon 5D Mark II cameras were used almost exclusively. You can Google the 2012 Super Bowl commercial for this film. Cinematographer: Shane Hurlbut.

TV Episodes:

• House M.D. – “Help Me” (Season 6 Finale, 2010), shot entirely on the 5D Mark II.
• 24 (Season 8, 2010) – Fox’s action series 24 shot all the background plates (establishing city shots, vehicle interiors, etc.) using the 5D Mark II and 1D Mark IV DSLRs.
• Dexter (2010–2013) – Crime drama used DSLRs to capture shots that were otherwise difficult with its primary cameras.

Since that era, almost all digital cinema cameras use Super 35 or full-frame sensors, and full-frame mirrorless cameras shooting 4 K or higher resolution have replaced DSLRs in the lower-end market segment. The Sony FX6 cameras used by my documentary team have essentially the same sensor as the FX3 and A7SIII mirrorless cameras, and the video image quality is the same. The $80 million 2023 science fiction film The Creator was shot entirely on the Sony FX3.

I see that A(ccumulated) I(gnorance) left out that camera manufacturers deliberately soften pixel edges in order to avoid moiré patterning. (Camera manufacturers don't really want anyone to know this — doesn't make a great selling point.) Of course, the more pixels the less the obvious the effect and the better the overall image is, but still... softened slightly. Moiré is an issue that cannot be avoided in digital imaging.

I don't know why 1440x1080 appears to be higher quality than 1920x1080. The underlying codec for your camera is AVC (ie H.264) for both the MP4 container and AVCHD container.

AVCHD is a hassle to deal with in FCP because it should always be imported from the AVCHD container, which mandates "copy to library." Removing MTS files from the container and importing with "leave files in place" often causes major I/O performance problems.

"Critical sharpness" is not the only (or even primary) determinant of image quality. Otherwise, professional filmmakers would not use cinema lenses that produce a softer look.

This is video, not still photography. There are many things that affect the overall quality of a moving image, including codec and scene type. The quality cannot be accurately determined by looking at a still frame.

If you Google on "UVG Dataset" you can see and download some of the reference clips used by academic researchers. These are used because a given codec has strengths and weaknesses on certain scene types.

Judging video sequences by eye is also not always reliable. It's a good idea to also use image quality measurement tools. Apple has a free tool called AVQT (Advanced Video Quality Tool). It's designed to measure the perceptual image quality of one clip vs a reference standard. E.g, if you have a ProRes 422 clip and export that from FCP as H.264, how much is the image degraded?

You could possibly use AVQT to objectively measure the quality difference between the MP4 and AVCHD versions of your clips. Google "How to Use Apple's Advanced Video Quality Tool" to see how it's used and where to obtain it. You have to sign up for an Apple Developer account, but that is free.

You can’t judge video quality in this way without separating the effects of in-camera sharpening and general over-processing which is likely to be very significant with all of the cameras you mention. The only way to really judge would be to shoot raw video which is not possible.  

The most likely reason your HD video is less sharp than your 4K video is that it is lower resolution in the first place. If you shot SD video it would be even less sharp again, which would be clearly obvious when viewed at the same size on a larger monitor. There is nothing sinister happening. And there is nothing intrinsically wrong with DSLRs in terms of image sharpness, stills or video. I don’t know what you are talking about in relation to pixel softening.

The whole thing about DSLRs (as in the title) is pretty moot anyway as they have been replaced by mirrorless interchangeable lens cameras (MILCs) for most purposes, in particular for anyone shooting video. Nobody intending to shoot video is going to buy a DSLR in 2025 as MILCs are far more suited to video than DSLRs. For example, Nikon are still selling a few DSLR models but that market is exclusively for stills photographers. Even then, the major advances are all in the MILCs so most stills shooters are going that way anyway, moreover because they are backward compatible with the older lenses. 

Ian R. Brown wrote:

Of course the 1080p of my FZ300 is lower resolution than 4k and as such is not as good.

My point is that the 1080p resolution of the FZ300 is for some inexplicable reason noticeably inferior to the SD800 and Nikon P530.

This is something that has been noted for years by many users of FZ200 and FZ300 cameras. The 1080p they produce is not up to the usual standard compared with other cameras whilst the 4K is really good.

OK I'm not at all clear what point you are making. I have no experience of any of these cameras. I thought you were cautioning people about buying DSLRs for video based on your experience with a Sony DSLT camera that is not actually a DSLR anyway. My main point in any case is that nobody buys DSLRs for video now or for the last several years as they have been superseded by mirrorless cameras anyway. Time to move on I think.

I used DSLRs for many years, always got great video. They're not all made equal.

Ian R. Brown wrote:

My point was that you cannot assume that because a camera can take pin sharp photos it will produce high quality video.

I have been reminded of this because for the past couple of weeks I have dug out and been trying to squeeze something better out of my Sony . . . to no avail.

DSLR, SLT or mirrorless is unimportant. I haven't checked out mirrorless but I can't see any reason why their video capabilities would be greater than a DSLR.

I’ve not seen any evidence for this at all. The 8K raw video from my mirrorless Z8 is just as sharp as 8K raw stills, all else being equal. You are basing your entire argument on one very old camera and something regurgitated by an AI. If you don’t ask the right questions in the first place, you are unlikely to get the right answers anyway. 

The reason that the quality of video from my mirrorless 45 MP Z8 is vastly superior to that from my now-departed 45 MP D850 DSLR is that the technology has vastly improved in terms of video quality due to various factors including sensor design. In fact there has been a slight deterioration in the stills quality compared to the DSLR, as the newer camera is designed for speed which is necessary in order to read out 8K video (and higher frame rate stills). So if you don’t see any reason why this is the case, maybe it’s time you did some research and ask the right questions if you want to place your trust in an AI. Experience trumps hearsay.