Fuji X-T3 released

[scorsesefan]

Can anyone who has the x-t3 tell me if there is a discernible difference quality between, say, the 200mbps recording and the 400mbps?

 

[puredrifting]

Can anyone who has the x-t3 tell me if there is a discernible difference quality between, say, the 200mbps recording and the 400mbps?

I can't see a difference. But I am looking at it on a 27" Retina iMac screen, maybe on a 65" OLED, you might? If you did, it would be very slight. The way I look at it, all of my 200Mbps comes into FCP X and is transcoded to Prores 422 at a data rate of 471 Mbps avg. Both the 200 and 400 Mbps are 10-bit, although 4:2:0. I only have two 128GB V60 ProGrade cards so why eat up all of my cards for a data rate that I cannot see any difference with? I find a much larger difference in shooting F-Log than an in-camera simulation, although I am having fun with those for quickie turnaround stuff or just playing with looks. If I was shooting green screen, I would buy a Ninja V for that to get the desired 4:2:2 but I have my C200 to shoot green screen. I think 200Mbps works great and gives me 160 minutes of UHD H265 Long GOP at 23.98 fps with the two 128GB V60 cards. When I can afford to buy a pair or a few pairs of V90 256GB cards, shooting 400 Mbps might make more sense. But the 200 in 10-bit looks amazing, not sure how much better you need? None of my clients are that picky, so far, they've all requested and been happy with H265 Long GOP at 200 Mbps and one of them is using it in a documentary that will eventually be in theaters.

The V90 256GB cards are still kind of pricey, although cheaper than CFast 2.0 cards for my C200.

 

[puredrifting]

Thanks. The one I posted came with a top handle, but it looks like smallrig is the only game in town for the x-t3.

The SmallRig top handle is cheap too. I think it's only like $22.00 for the cheap one and $56.00 for the Nato Rail one? I'm used to Zacuto, where everything is overpriced, I like the SmallRig stuff almost as much though

 

[Matt Gottshalk]

Not XT-3 but Fuji related:

I just pulled the trigger on a slightly used Fujifilm GFX medium format body.:2vrolijk_08:

 

[scorsesefan]

I can't see a difference. But I am looking at it on a 27" Retina iMac screen, maybe on a 65" OLED, you might? If you did, it would be very slight. The way I look at it, all of my 200Mbps comes into FCP X and is transcoded to Prores 422 at a data rate of 471 Mbps avg. Both the 200 and 400 Mbps are 10-bit, although 4:2:0. I only have two 128GB V60 ProGrade cards so why eat up all of my cards for a data rate that I cannot see any difference with? I find a much larger difference in shooting F-Log than an in-camera simulation, although I am having fun with those for quickie turnaround stuff or just playing with looks. If I was shooting green screen, I would buy a Ninja V for that to get the desired 4:2:2 but I have my C200 to shoot green screen. I think 200Mbps works great and gives me 160 minutes of UHD H265 Long GOP at 23.98 fps with the two 128GB V60 cards. When I can afford to buy a pair or a few pairs of V90 256GB cards, shooting 400 Mbps might make more sense. But the 200 in 10-bit looks amazing, not sure how much better you need? None of my clients are that picky, so far, they've all requested and been happy with H265 Long GOP at 200 Mbps and one of them is using it in a documentary that will eventually be in theaters.

The V90 256GB cards are still kind of pricey, although cheaper than CFast 2.0 cards for my C200.

Thanks, would something like this work: https://www.bhphotovideo.com/c/prod...al_pgsd128gbjbh_sdxc_uhs_ii_128gb_memory.html Is the write speed too slow? I've been working with the 28mbps c100 for so long I forgot about these specs

 

[NorBro]

80MB shall cover 50MB (400mbps) if the writing speed is solid and continuous (sustained).

 

[yoclay]

If you, or someone else on this thread can confirm that this does indeed work - and try the lenses on both mfg's camera's.... this would be HUGE - at least for me. It's been the only thing holding me back from getting at least one, if not both of the Fuji lenses.

I am receiving them next week and will check then. By the way, they were originally designed to have interchangeable mounts, but the discussions around that are still ongoing at the corporate level. Personally, I think the design has already integrated this because Duclos makes a 3rd party FZ mount for these lenses (with the blessing of Fuji so as not to endanger the warranty) and basically it really is only 4 screws to swap out the mount. The backfocus adjustment is primarily for this reason, but the macro capacity is a nice secondary benefit.

 

[jcs]

I can't see a difference. But I am looking at it on a 27" Retina iMac screen, maybe on a 65" OLED, you might? If you did, it would be very slight. The way I look at it, all of my 200Mbps comes into FCP X and is transcoded to Prores 422 at a data rate of 471 Mbps avg. Both the 200 and 400 Mbps are 10-bit, although 4:2:0. I only have two 128GB V60 ProGrade cards so why eat up all of my cards for a data rate that I cannot see any difference with? I find a much larger difference in shooting F-Log than an in-camera simulation, although I am having fun with those for quickie turnaround stuff or just playing with looks. If I was shooting green screen, I would buy a Ninja V for that to get the desired 4:2:2 but I have my C200 to shoot green screen. I think 200Mbps works great and gives me 160 minutes of UHD H265 Long GOP at 23.98 fps with the two 128GB V60 cards. When I can afford to buy a pair or a few pairs of V90 256GB cards, shooting 400 Mbps might make more sense. But the 200 in 10-bit looks amazing, not sure how much better you need? None of my clients are that picky, so far, they've all requested and been happy with H265 Long GOP at 200 Mbps and one of them is using it in a documentary that will eventually be in theaters.

The V90 256GB cards are still kind of pricey, although cheaper than CFast 2.0 cards for my C200.

If you're going to transcode to ProRes 422 at 471Mbps (UHD), I think you might find QuickTime (MOV container) with the (wavelet-based) CineForm codec at quality 4 (Film Scan) or for smaller files, quality 3, provides a more bang-for-buck in terms of quality per-bitrate vs. the older DCT-based ProRes. The cool thing about wavelet is you'll never see macroblock artifacts, and it's why Red and DCP (Digital Cinema/JPEG 2000) use wavelet-based instead of DCT. Since Cineform is now open source, it's really picked up adoption and the reliability is also excellent (that was my only reservation years ago when I purchased it (was one of the only options when dealing with stereoscopic 3D footage)). Real-time decode performance is also excellent, and it appears some apps even have GPU support for encoding (e.g. Resolve); probably has GPU support for decoding too since it's open source (and wavelet in general has proven to be fast to decode on the CPU).

QT + Cineform should be available on most NLEs that support QuickTime, except FCP X (just checked). After googling, it looks like Apple is even going to (or already?) drop support for Cineform. This makes sense since it appears to be a superior codec in terms of quality per bitrate and real-time decode performance. It appears Apple dropped support the moment it went open-source/free: Apple gets royalties for the closed-source ProPres.

So for those using Resolve and PP CC (perhaps Avid?), you're good to go with a better, faster, smaller, and license-free codec vs. ProRes

A quick search shows 310Mbps for Cineform Film Scan UHD vs. 471Mbps for ProRes 422. Cineform with 3 (minor) vs. ProRes 422 with 4 (some) quality loss (lower numbers are better): https://blog.frame.io/2017/02/13/50-intermediate-codecs-compared/ . This makes sense based on my informal tests and new wavelet tech vs. older DCT tech.

Additionally: Cineform: 140GB/hr vs. ProRes 422: 212GB/hr. To match Cineform Film Scan quality (loss: 3 (minor)), ProRes has to go up to 422 HQ at 707Mbps and 318GB/hr. Thus it's clear why Apple dropped (more like banned lol) Cineform when it became free: Cineform has more than double the quality of ProRes for the bitrate! Also likely that Cineform encodes and decodes faster.

Looks like Red users knew Cineform was superior to ProRes back in 2014: http://www.reduser.net/forum/showthread.php?121568-Cineform-superior-alternative-to-ProRes

I need to double check, Cineform can also replace ProRes for those looking to reduce file size for YouTube upload (and have been using ProRes).

 

[puredrifting]

If you're going to transcode to ProRes 422 at 471Mbps (UHD), I think you might find QuickTime (MOV container) with the (wavelet-based) CineForm codec at quality 4 (Film Scan) or for smaller files, quality 3, provides a more bang-for-buck in terms of quality per-bitrate vs. the older DCT-based ProRes. The cool thing about wavelet is you'll never see macroblock artifacts, and it's why Red and DCP (Digital Cinema/JPEG 2000) use wavelet-based instead of DCT. Since Cineform is now open source, it's really picked up adoption and the reliability is also excellent (that was my only reservation years ago when I purchased it (was one of the only options when dealing with stereoscopic 3D footage)). Real-time decode performance is also excellent, and it appears some apps even have GPU support for encoding (e.g. Resolve); probably has GPU support for decoding too since it's open source (and wavelet in general has proven to be fast to decode on the CPU).

QT + Cineform should be available on most NLEs that support QuickTime, except FCP X (just checked). After googling, it looks like Apple is even going to (or already?) drop support for Cineform. This makes sense since it appears to be a superior codec in terms of quality per bitrate and real-time decode performance. It appears Apple dropped support the moment it went open-source/free: Apple gets royalties for the closed-source ProPres.

So for those using Resolve and PP CC (perhaps Avid?), you're good to go with a better, faster, smaller, and license-free codec vs. ProRes

A quick search shows 310Mbps for Cineform Film Scan UHD vs. 471Mbps for ProRes 422. Cineform with 3 (minor) vs. ProRes 422 with 4 (some) quality loss (lower numbers are better): https://blog.frame.io/2017/02/13/50-intermediate-codecs-compared/ . This makes sense based on my informal tests and new wavelet tech vs. older DCT tech.

Additionally: Cineform: 140GB/hr vs. ProRes 422: 212GB/hr. To match Cineform Film Scan quality (loss: 3 (minor)), ProRes has to go up to 422 HQ at 707Mbps and 318GB/hr. Thus it's clear why Apple dropped (more like banned lol) Cineform when it became free: Cineform has more than double the quality of ProRes for the bitrate! Also likely that Cineform encodes and decodes faster.

Looks like Red users knew Cineform was superior to ProRes back in 2014: http://www.reduser.net/forum/showthread.php?121568-Cineform-superior-alternative-to-ProRes

I need to double check, Cineform can also replace ProRes for those looking to reduce file size for YouTube upload (and have been using ProRes).

Unfortuantely the Cineform codec probably wouldn't work for those of us in collaborative workflows where Prores is the standard and expected codec for almost everything coming and going. Would be fine for projects where we do everything.

 

[puredrifting]

Thanks, would something like this work: https://www.bhphotovideo.com/c/prod...al_pgsd128gbjbh_sdxc_uhs_ii_128gb_memory.html Is the write speed too slow? I've been working with the 28mbps c100 for so long I forgot about these specs

Those are the exact cards I am using. So far, so good. It is a little confusing about the 400 Mbps, some sources have said V90 but the owners manual says V60 or better I think? I have shot H265 400Mbps with these V60s and they have worked fine. But I have not tried 30 minute continuous takes for interviews at 400 Mbps, only at 200 Mbps.

 

[SJX]

Guys I need to confirm something. So the X-T3 doesn't record 10bit 422 4K internally? First by reading I had impression it did 10bit 420 60p 4K and 10bit 422 up t0 30p 4k internally. Now I'm thinking 422 is only HDMI out and internal is only 420? Thanks.

 

[Grug]

Guys I need to confirm something. So the X-T3 doesn't record 10bit 422 4K internally? First by reading I had impression it did 10bit 420 60p 4K and 10bit 422 up t0 30p 4k internally. Now I'm thinking 422 is only HDMI out and internal is only 420? Thanks.

Only 10-bit 4:2:0 internally (in 2k/4k/24p-60p). 10-bit 4:2:2 (in 2k/4k/24p-60p) is through HDMI.

 

[SJX]

Are these the Fuji lenses, Matt?



https://www.bhphotovideo.com/c/product/1021608-REG/fujifilm_16412188_xf_10_24mm_f_4_r.html?sts=pi

https://www.bhphotovideo.com/c/product/1110879-REG/fujifilm_xf_16_55mm_f_2_8_r.html?sts=pi

https://www.bhphotovideo.com/c/product/1080892-REG/fujifilm_16443060_xf_50_140mm_f_2_8_r.html?sts=pi

This is one downside for me. Having to invest in fuji mount lenses to get the full benefits of CA correction etc. Is proprietary mount and I have no idea if I will buy another fuji cam after. Like when I bought GH2 and a bunch of 4/3 glass and then never bought another 4/3 again. Wasted glass. I much more prefer use my PL lenses or buy some more mainstream like EF. But then no full benefit as with fuji lens.

 

[SJX]

If I were buying a camera right now, the hardest decision would be between the Pocket 4K and X-T3.

https://petrolicious.com/films/1969-porsche-911t-maximum-pleasure-minimalist-package

This is from the X-T2 and X-T20 (I supplied the X-T20).

*Edit:

Shot by Macgregor
Before the F-Log internal update so straight to a profile
No External Recorder etc.
I believe on Fuji lenses as well


I know what the Pocket 4K can do as well, and I have a pretty good idea of what the X-T3 can do. In the end I'd probably go Pocket 4K because of shooting formats and flexibility, but it'd be a very very close call... extremely close.

So this was shot by macgregor? Nice. Seems lot of work goes in these 5 minutes videos. Wonder how long it takes to film that? Also why there is no credits on the video for who shot it etc? I was expecting it after such nice production.

 

[SJX]

I can only disagree strongly. 4:2:2 subsampling throws aways half of your chroma information, that's not great, but in practice it leaves enough for a lot of purposes. 4:2:0 subsampling throws away half of that remaining half. It's recording only 25% of the chroma information that the sensor is providing.

If throwing away 75% of your chroma information isn't 'violently' compressing it. I don't know what is.

Yes, downsampling the image to 2k should yield a more robust image (relatively) on a per-pixel basis, but then you're losing your 4k image.

It remains to be seen how malleable the 4:2:0 image is going to be, and the impact it will have on maintaining colour consistency through a final graded scene. So Matt, if you fancy shooting a test for us, it'd be much obliged!

Yes. This is my main worry.

 

[SJX]

Does anyone have any 10-bit 4:2:2 material that they've recorded externally that I could perhaps download and look at?

I feel like I have a decent sense for how the internal 10-bit 4:2:0 holds up, but

So how do you think it holds up?

 

[SJX]

Are the adapters really trustworthy enough for paid work?

It sounds like most of the Fuji primes are inappropriate for most video work. Focus-by-wire (making manual focusing a nightmare), and just not smooth, silent or fast enough in autofocus modes (for anything much beyond holding a face in a sit-down interview).

Can you get more consistent or reliable focus from Canon lenses pushed through adapters? And are they consistent enough to rely on?

Or are most people simply adapting their manual glass onto the cameras?

I've got a couple of PL mounts on the way for mine. Metabones told me the excitement around the X-T3 has them considering doing a PL to X-mount Speedbooster (which is what I'm most excited for), but who knows when they'll get around to that.

I think it'd well worthwhile us all collectively detailing how each lens we have (either native or adapted) performs focus-wise.

So you find the AF not good enough?

 

[marvinhello]

So how do you think it holds up?

During decoding, 420 or 422 will all be reconstructed to RGB 444, and the algorithm is pretty good and mature now. For primary colour grade, colour subsampling does not matter at all. It’s more about compression and bit depth. It’s when you start doing secondary grading (pulling keys) that you get the benefits of 422/444, edges will be cleaner with better colour separation. That’s why VFX heavy productions usually demand RAW acquisition for RGB 444 and linear colour space.

 

[SJX]

During decoding, 420 or 422 will all be reconstructed to RGB 444, and the algorithm is pretty good and mature now. For primary colour grade, colour subsampling does not matter at all. It’s more about compression and bit depth. It’s when you start doing secondary grading (pulling keys) that you get the benefits of 422/444, edges will be cleaner with better colour separation. That’s why VFX heavy productions usually demand RAW acquisition for RGB 444 and linear colour space.

I asked because grug's comment about possible problems maintaining color consistency through a final graded scene with 420. Maintaining colour consistency is considered primary color correction right?

 

[scorsesefan]

Those are the exact cards I am using. So far, so good. It is a little confusing about the 400 Mbps, some sources have said V90 but the owners manual says V60 or better I think? I have shot H265 400Mbps with these V60s and they have worked fine. But I have not tried 30 minute continuous takes for interviews at 400 Mbps, only at 200 Mbps.

Thanks. So how much rec time would you get on 64gb card 200mbps 4K 23.98 h265?