But there might be even more in the pipeline. Maybe even some I could not see coming, just as it was the case of the Fujifilm GFX Eterna.
My best guess so far is that it could be the Fujifilm GFX100RF fixed lens camera or the Fujifilm GFX Eterna. But that’s just a guess. And in fact it does not really matter which codename which camera has. There will always be a codename because there will always be a camera coming.
Thierry Gibralta shared a follow up autofocus test video which you can see above. I will provide a quick summary, but seeing with your own eyes is always the best option.
VIDEO AUTOFOCUS
previously constantly pulsing and hunting. Now much less pulsing. Overall for static talking had the X-H2S video AF can be used now
walking away and towards the camera, the AF overall sticks to the person. When walking away, it can lose focus for a second or two, but that’s not relevant for his work, because he does not need to film people who walk away from the camera. Focus is not really smooth, there are some steps in the focus, but it is still for sure usable footage. A client would not really notice those AF steps.
when running towards the camera, the AF steps are more visible
when moving around with frequent turns, with eye AF, as long as the distance does not change too much it is good. If distance changes a lot and if you move out of the frame and then back, the focus can have a hard time. But with face detection he did not have this issue, so if you shoot people dancing, better use face detection
in the past, when people passed by in the background, the AF would jump to the person passing by. With new firmware, eye/face detection did not get distracted by people passing by. He tested it with people passing behind and in front of the camera, and it would occasionally lose him, especially if he placed himself on the very side of the frame. If he stays in the center, the camera usually picked him up. But if you are on the side of the frame, it can jump to a person more in the center of the frame
camera on the table with objects on the table and camera moved to focus on one object or the other. Multi Mode without subject detection. This gives the biggest problems. The camera did not detect the objects close to it on the table. Multi without subject detection is the worst performing
STILLS AUTOFOCUS
tracking is working very well and most pictures are in focus (people walking around and on bycicles)
success rate is pretty good
not perfect. he had instances like tracking a person on a bicycle but then the AF jumped to 2 kids walking by nearby
CONCLUSION
Autofocus is drastically improved over previous firmware
It is not perfect
for his work (wedding etc) and what he does, now he can use it for his work
he’d like smoother focus transitions and eliminate the times it jumps to another person. Also Multi mode with no subject detection should be improved
I have been testing the latest firmware on the XH 2 wide open on my bordie.
My dog is a good and difficult subject to test focus performance.In the past quite some miss focus on the nostrils , black-white high contrast transition zones.
Even when eyes almost closed and the eyes being surrounded by black fur, the AF performance is now very good! Using the XF 50/1 wide open here with good results, is quite an achievement!
My Nikon Z8 cannot manage this!!
Andrea Cimini
Andrea Cimini was very critical about the mess with Fujifilm’s AF. Now he tested the new version, especially for stills photography tracking mountain-bikers jumping and racing around, side by side with an X-H2S with older firmware.
old firmware 93% hit rate (perfectly super sharp images)
new firwmare 98% hit rate (perfectly super sharp images)
with subject detection
old firmware 95% hit rate (the ones not in focus are totally out of focus, NOTHING was in focus at all)
new firwmare 99% to 100% hit rate (the totally out of focus does not happen anymore. It can focus on something else)
with car/bike detection
old firmware: 87% in focus
new firmware: 93% in focus
NOTE: the new firwmare started tracking the subject from more far away compared to the new firmware
using wide tracking (not subject detection, but you put the focus box on anything and it will keep tracking it)
condition: biker suddenly appearing in the frame
old firmware: 85% in focus
new firmware: 96% in focus
Conclusions so far
it’s not at level of competition, but the autofocus is usable again
now he feels a bit safer and he can go shooting an assignement without fear
electronic shutter gives better results. With mechanical shutter the percentage drops by 10% (at 20fps)
To make it clear: while there was an improvement, for his type of (very challenging) sports shooting conditions (bikers racing inside forests), he thinks Fujifilm must improve even further. But it is workable again.
Petapixel reviewed the Fujifilm X-M5. Regarding Autofocus they say:
comparing to Canon EOS R10, because that’s one of their favorite cameras as far as autofocus on APS-C goes at the same price
not quite the same hit rate of the Canon, but actually surprisingly close
in this price range there is nothing out there that gives you 10 bit, Log recording with 6.2K open gate
pre-production X-M5 had issue with focusing on the background. Production sample is much better but it can still be an issue
overheating: impressing performance for such a small camera body. And you can even use the Fan accessory for even better performance. Extremely impressive
probably the most capable video camera at that price
More reviews below.
AF Better than What YouTube is Talking About?
The first of Pav SZ looks actually very good in terms of Autofocus.
You can see the X-M5 keeps tracking the subject even when other people enter the frame. It had a few occasions in which the person turned around and it focused on the shoulder instead of the back of the head.
Godwin Isaac (video below), was a bit worried about the X-M5 autofocus, given all the autofocus sh*tstorm YouTube is throwing on Fujifilm autofocus. But he says “autofocus is way better than what YouTube is talking about“.
So why is that? Why are so many people happy with the autofocus?
Pav SZ says that if you use it in real life shooting conditions, it will work just fine. But if you set it up for failure in rather tricky and unrealistic test conditions, it might struggle more than in real life use. So that’s his theory.
I Remain Utterly Disappointed ;)
My position remains unchanged: even in the most ridiculous never to happen absurd and unrealistic test conditions, I want Fujifilm cameras to never fail once. I will always be utterly disappointed if in a nonsense test Fujifilm cameras hunt for a fraction of a second. ;)
Of course I am exaggerating. But the truth is I believe there is still room for improvement, and Fujifilm has to work on it. And I am not kidding here: it’s paramount that they keep working on it. Clean up your code, get the algorithm right. And we have actually shared an article which shows in which areas Fujifilm needs to improve (and can improve, since they got it right with other cameras already):
As we told you yesterday, an anonymous source contacted us with the details about the upcoming fixed lens Fujifilm GFX camera.
First off: THANKS for the help!!! And second: it’s all very accurate according to my trusted sources, except for one thing (and at this point you know what). If you want, we can chat 100% anonymously via Signal to talk about it.
So these are the first specs of the Fujifilm GFX RF.
Name: Fujifilm GFX100RF
100 Megapixel
No IBIS
GF35mmF4 Lens
Coming March 2025 (it’s unclear if this is the date for the announcement of for the shipping… or maybe even both? I will let you know if I figure that out).
No IBIS makes sense, considering that it will be very compact for a medium format camera. In fact, we told you it will be “about the size of the Fujifilm X-Pro3“.
Also 35mm was to expect, as we always told you it will be a 28mm equivalent lens (hence 35mm on GFX). We also told you it will be an f/3.2 lens, but we were not sure if it was f/3.2 on GFX or f/3.2 Full Frame equivalent. Well, we can now say it will be f/3.2 FF DOF equivalent (hence f/4 of GFX)
