Panasonic GH6

[Video Hummus]

On 2/17/2022 at 4:54 AM, Mmmbeats said:

I think the ARRI / Canon method is basically a frame blend too.  They've just done a much, much better job of marketing it!

Well the DGO method is using hardware and the other method is doing two different exposures at different shutter speeds and then blending them. I think ZCAM tried to do this and this is also what phones are doing with fast shutter speeds and fast readouts (looks bad in my opinion).

[kye]

11 hours ago, Ian Edward Weir said:

I have solely used GH cameras. GH2, GH3, GH4 and GH5. When I look at the GH6 I say to myself I have the GH5 and it still kicks ass. I only use manual lens and could care less about auto focus. Being that the GH5 holds down MFT for me and some thing like the Z-Cam M4 looks like next level MFT. I'm really thinking of jumping into a S5 or the next version of the SHII. Anyone else feel the same way? The size of the GH6 looks around the same size as full frame cameras. The idea of seeing my vast collection of adapted lens in their native FOV and having the low light capabilities are more appealing than GH6. Anyone else feel this way? 

If you've built a FF glass collection then that's definitely a bridge to a mirrorless S35 or FF, so in a way you were always keeping your options open.

In terms of low-light I'll be very interested in the GH6's low light as that's one of the aspects of the GH5 that I really push.  For example, here's a shot I took with the Voigtlander 17.5mm F0.95 lens wide open - the scene is solely lit by the lights on the river bank:

I'd recommend you wait for the sample footage and tests as the specifics of how they have implemented the sensor tech will really matter.

11 hours ago, Jimmy G said:

@kye @Django re: ARRI

Seeing how the vast majority of recent Sundance films used their cameras, I can certainly understand a potential camera buyer's curiosity and desire to know how well a particular camera holds up against ARRI's offerings...

The Cameras Behind Sundance 2021: ALEXA Mini Rules, and the BMPCC Chimes-In - Y.M.Cinema - News & Insights on Digital Cinema

https://ymcinema.com/2021/01/31/the-cameras-behind-sundance-2021-alexa-mini-rules-and-the-bmpcc-chimes-in/

...methinks the ARRI image/look is (has become) the, cough, "gold standard" within certain creative circles.

The Alexa has definitely become the gold standard within some circles.  Those circles are basically people who appreciate great colour.  I actually don't want ARRI to have the best colour - I'd prefer if the GH6 ends up with the worlds best colour, having the best colour come from a camera I can't afford, couldn't carry, and couldn't realistically use would be a completely stupid wish.

I just find it odd that people can say "Sony might include the Venice colour science in their next A7S camera" and it's fine, but saying "I wish someone would make colour science approaching ARRI" somehow is crazy talk, as if the Alexa is a magical unicorn instead of a sensor and a processor in a box...  just like every other digital camera ever.

11 hours ago, Jimmy G said:

That said, I think what Panasonic is offering with their MFT lineup, and the GH line in particular, is meant to appeal to both professionals and non- alike. Being able to shoot untethered clearly has it's advantages for numerous environments (bumper cams, bike cams, underwater, spelunking, hiking, street, immediately come to mind), it would be nice to see them include all flavors of output (read: RAW, ProRes, etc.) internally, as well.

Early "leaked" specs (if they are to be believed) seem to indicate a move in that direction, some of us have absolutely zero desire for an external recorder. This S1 shooter is, in fact, still quite pissed at Panasonic that the XQD slots have never been used to their potential, 1400/1700MBs sustained write speeds would easily accommodate ProRes HQ and RAW. (Not too late for a firmware update, folks! 😉 )

I agree.

People talk about high resolutions like a "just in case / when you need it it's there" kind of thing.  I see that "ready for anything" aspect as the design brief of the GH line.

Do people need GH5-level IBIS on every shot?  No.  Most of the time a lesser-IBIS would suffice, but are there times when you need it?  Absolutely.  I routinely push the GH5 IBIS past its limits on trips - maybe because I'm cold or low-blood sugar or I'm filming from a helicopter with the door open at 200kph or whatever.  

Do people need 10-bit on every shot?  No.  8-bit cameras make gorgeous footage in when exposed properly and under modest-DR situations, but are there times when you need the flexibility in post?  Absolutely.  I shot with the XC10 in 8-bit C-Log on a 5-week trip to Italy and have really struggled to clean up the footage because it doesn't have the latitude the GH5 has.

Same logic for high-bitrates.  Every now and then you film in a situation where there's lots of chaotic movement like in rain or snow or with trees moving in the wind or whatever.  Also, sometimes you want to crop in post a bit and not reveal compression nasties.

+1 about not wanting an external recorder.  People who don't care about camera size seem incapable of understanding that everyone isn't like them, it's rather odd.  Its like saying you prefer chocolate ice cream and them saying "no you don't".  

Thanks for your thoughts on the ALEV vs the GH6 tech.  Obviously the proof is in the pudding with the images, which I am really looking forward to, but once it arrives I must admit I'd be very interested in learning more about how it works 🙂 

8 hours ago, Davide DB said:

I don't understand... Granted that we will then need to see the files generated, I don't understand how you can compare the two cameras on video specs.

It seems to me that the OM-1 has almost the same specs as the current GH5 except for 240 fps in Full HD and external ProRes.

Just fto name a few:

OM-1

H.265 Long GOP/MOV 4:2:0 10-Bit

DCI 4K (4096 x 2160) at 23.98p/25p/29.97p/50p/59.94p [77 to 152 Mb/s]

H.264 Long GOP/MOV 4:2:0 8-Bit

DCI 4K (4096 x 2160) at 23.98p/25p/29.97p/50p/59.94p [102 to 202 Mb/s]

GH6

K120p 4:2:0 10bit 300Mbps
4K 60p 4:2:2 10bit 800Mbps/600Mbps
5.7K 60p 4:2:0 10bit 300Mbps
C4K 120p 4:2:0 10bit 300Mbps
C4K 60p 4:2:2 10bit 800Mbps/600Mbps

 

If all you care about is specs then you can make a complete comparison from the spec sheet.

Some people will be comparing specs until we see the images, but sadly, other people here only care about specs, and when talking about cameras seem uninterested or incapable of understanding the various other considerations that go into making an engaging end product.

4 hours ago, mercer said:

For years, people complained that they couldn't get internal raw video on their cameras and now Panasonic is releasing a camera with internal ProRes (the next best thing to raw) and people aren't impressed?

ProRes is a professional codec.

With that said, is $2500 too expensive for a GH camera... YUP!

Can you imagine what the S2H will cost? $5000? Way too much for that camera too... even with ProRes.

I'm curious why more of the pros, on this site, aren't buying a Komodo or C300 Mark III?

Agreed - Prores is a big deal.

One thing that people probably aren't aware of is how good a codec Prores really is.  For reference, a very large proportion of the movies that people saw in the cinema between the mid 1990's and the mid 2010's probably went through Prores HQ, and a good chunk of those would have been Prores HQ in 1080p.  

It was the bread-and-butter codec for Hollywood and often still is.  People don't seem to understand that.

[kye]

10 hours ago, webrunner5 said:

I have just been on my computer looking at my multiple hard drives and sifting through 20 different cameras I have had in the last several years and the GH5 stuff has to be the least favorite of all of them looks wise. It is just flat and lifeless compared to everything else I have owned. I have thousands of pictures and videos and it is far and away the least favorite. YMMV.

I'm not the biggest fan of the GH5 colour science either.

Which makes me happy that Panasonic seem to be chasing a better image with the GH6 instead of just chasing endless resolution at the expense of everything else.

The colour from the more recent Panasonic cameras has all been incremental improvements over the GH5, so I'm optimistic about what they'll do with a new sensor.

The other aspect to making images pop is lenses, which there are more and more available all the time now with third-party manufacturers like 7artisans, TTartisans, Meike, Mikaton, etc making interesting offerings.  To me they're interesting because they are a perfect-combination of features - they have simpler optical designs and simpler coatings that are reminiscent of vintage lenses that are now climbing radically in price, but due to cheap Chinese manufacturing are both low-cost and also relatively high-quality.  
Unlike modern high-resolution high-precision zero-distortion lenses which have a very dull and lifeless rendering, these third-party primes tend to exhibit the aberrations that make lenses like the "Zeiss 28mm f2 'Hollywood'" lens famous, plus with MFT or APS-C lenses you're seeing more of the edges of the image circle than you do from FF lenses and so you're getting more of those character-providing flaws.  I'll be talking more about this in coming weeks, but there's a lot to talk about, let's just say that.

 

[webrunner5]

I don't understand it to be honest. The LX100 and the G7 I have had were really good, damn good CS wise. Even the GH3 was not bad, but man the GH4, GH5, well not worth a damn I think. They went backward.

[kye]

1 hour ago, webrunner5 said:

I don't understand it to be honest. The LX100 and the G7 I have had were really good, damn good CS wise. Even the GH3 was not bad, but man the GH4, GH5, well not worth a damn I think. They went backward.

Well, there's the gradual Sony-fication that has been happening in camera colour science.  Hopefully this is a departure from that, more in the direction of manufacturers-that-cannot-be-named.

[Django]

Ok rather than blame Sony or high-resolution etc.. Let's try and actually tackle the actual IQ/CS situation to understand the bottlenecks and what's at stake shall we?

There is much more to CS than sensor manufacturing. Sony make sensors for Sony/Nikon/Fuji/Panasonic yet they all fine tune to their respective CS through image processing. 

And CS evolves.. for better or for worst depending on personal preference. Some people prefer pre-GH5 color gamuts. Same with Canon, some miss the 5D3/C100/1DC CS.. fortunately Canon bodies allow to set the log color matrix to "Cinema EOS original" or the more ARRI inspired CS "neutral". You get to decide.

Every camera manufacturer has its own golden standard. For Sony its Venice and A7S3/FX3/FX6/FX9 allow proper Venice S-gamut3.cine in log but also S-cinetone profile for a baked-in Venice look. 

For Panasonic, their gold standard is obviously Varicam. But what some may not realise is that GH5's Vlog-L doesn't operate in V-Gamut but in Rec709. So it's kind of a fake Varicam CS. Also seems to dead-end ACES workflow with no appropriate IDT support from Panny AFAIK. You have to step up to S1H for proper V-Log/V-Gamut.

ARRI -the Hollywood gold standard- Log C uses Alexa Wide Gamut (AWG) with either 10-bit ProRes 422 or 12-bit ProRes 4444 XQ. AWG works in a much larger colour space than Rec709. The encoding was done with a complete and deep ACES workflow in mind. This whole relationship is what makes for a class-leading grading platform.

Other important IQ related aspects are in-camera sharpening and NR. Both seemed to plague GH5.

Will be interesting to see if GH6 tackles some of these issues.. we still know nothing of the actual ProRes flavours or if Log is still Vlog-L or proper Vlog/VGamut...

 

[billdoubleu]

8 hours ago, mercer said:

For years, people complained that they couldn't get internal raw video on their cameras and now Panasonic is releasing a camera with internal ProRes (the next best thing to raw) and people aren't impressed?

ProRes is a professional codec.

Sure, people were complaining about not having professional editing codecs available internally in cameras like the GH5, but, those complaints began at least five years ago when Prores was already 10 years old. Prores was an option in every blackmagic camera alongside CinemaDNG starting in 2012, so, there's no logical argument that those codecs couldn't have been included in the GH5. The GH3 should've been the last GH camera sans Prores.

Here we are in 2022 and there is a growing list of RAW/ RAW(ish) codecs that didn't exist five years ago. ProresRAW, BlackmagicRAW, TICO-RAW are innovations sprung from the limitations imposed by RED's patents. These are all available internally in one or more cameras. Along with uncompressed RAW in the Sigma fp. So, while Prores is a welcome addition, no one should be impressed by it. Panasonic should either have developed a RAW technology by now or be licencing one.

[Django]

Agreed. Competition has caught up since 5 years ago and playing field has levelled as far as codecs. Just about every cam does high bitrate 10-bit 4:2:2. Nikon now has internal NRAW/ProRes, Canon has CRL/XFAVC. BMD BRAW/ProRes. 

Its great GH6 has ProRes but its basically to be expected. I also think they should have RAW internal by now. So should Sony (especially on FX6/FX9).

9 hours ago, mercer said:

I'm curious why more of the pros, on this site, aren't buying a Komodo or C300 Mark III?

There are couple Komodo users here. 

Personally I'd rather rent when it comes to those types of sensor in a box cine cams. No NDs, XLR inputs, EIS etc means a crew and when you're at that budget, renting makes more sense imo.

C300 mk3? Well other than costing $11K and weighing a ton, the C70 is a much wiser and popular investment especially now that it's getting RAW via free firmware update next month. You're basically getting C300mk3 IQ for half the price. Combine that with an R5C and you've got a pretty versatile and future proof hybrid combo for less than a C300mk3.

[kye]

 

53 minutes ago, Django said:

Ok rather than blame Sony or high-resolution etc.. Let's try and actually tackle the actual IQ/CS situation to understand the bottlenecks and what's at stake shall we?

There is much more to CS than sensor manufacturing. Sony make sensors for Sony/Nikon/Fuji/Panasonic yet they all fine tune to their respective CS through image processing. 

Absolutely, and that's perhaps the biggest reason that I keep referencing ARRI.  Their sensor is great, but it's the processing that they do to the image that really sets them apart.  In fact, it's so valuable to them that they apply it in-camera rather than in LUTs that can be pulled apart and analysed.

I'm not sure how much you know about colour science, but I have done many deep dives into pulling apart the colour science from a number of cameras, including Alexas, REDs, BMs, Canons, and Panasonics.  I have spent a lot of time on the colourist forums reading their responses, reading the articles they link to, and studying their methods (and when I say "studying" I mean opening Resolve and trying to re-create the methods they explain, and then using those methods on my own footage to get a feel for what is going on - like how you do when you do assignments at school - literally studying).

What I have found is that:

• most colour science is a long way from neutral, but are almost universally pushing the colours in the same ways, typically in the ways that film does
• you can take clips from multiple cameras and match them (and I'm talking about footage with larger colour checkers with lots of patches) and they look the same, but the ARRI or RED will have magic that the other simply won't have

I have also observed time and time again that the colourists are doing very complicated adjustments (often in alternate colour spaces that work in very different ways) and applying them very subtly.  

What I conclude from both the comparisons I have done and the little tweaks that the colourists are willing to share (there is a lot they're not willing to share too) is that the magic is in the tiny little adjustments.  Like in cooking how some chefs can add tiny amounts of various seasonings that are so subtle you can't pick them out but they really lift the flavour.  

You are absolutely right that each manufacturer has the opportunity to be building these things into their colour science (and not relying on their sensors), but the problem is that they just don't.  Year on year they are getting incrementally better but really aren't closing the gap between their $2K-5K cameras and what the leaders are doing.

The end result is that we're getting food that has come from the same ingredients (Sony sensors) and has only been seasoned with salt and pepper and therefore tastes rather bland in comparison to ARRI/RED who are demonstrating mastery in their use of spices.

53 minutes ago, Django said:

For Panasonic, their gold standard is obviously Varicam. But what some may not realise is that GH5's Vlog-L doesn't operate in V-Gamut but in Rec709. So it's kind of a fake Varicam CS. Also seems to dead-end ACES workflow with no appropriate IDT support from Panny AFAIK. You have to step up to S1H for proper V-Log/V-Gamut.

I don't have V-Log on my GH5 for precisely this purpose - it wouldn't get me anything.  That's why I've been shooting HLG and testing it (it's not exactly either rec2020 or rec2100, but it's close enough to rec2100 to use that in Resolve).

It would be great if the GH6 had real V-Log.

53 minutes ago, Django said:

ARRI -the Hollywood gold standard- Log C uses Alexa Wide Gamut (AWG) with either 10-bit ProRes 422 or 12-bit ProRes 4444 XQ. AWG works in a much larger colour space than Rec709. The encoding was done with a complete and deep ACES workflow in mind. This whole relationship is what makes for a class-leading grading platform.

I'm very keen to see how they go about using the Prores.

Currently the GH5 HLG implementation is 10-bit rec2100-like colour and gamma, which isn't too bad to work with.  The extra bit depth of Prores 4444 would be most welcome.

53 minutes ago, Django said:

Other important IQ related aspects are in-camera sharpening and NR. Both seemed to plague GH5.

Will be interesting to see if GH6 tackles some of these issues.. we still know nothing of the actual ProRes flavours or if Log is still Vlog-L or proper Vlog/VGamut...

In camera NR and sharpening are definitely an issue and I'd hope that implementing Prores will mean they'll tune the image to that codec and the expectations that pros would have.  I don't think the idea that Prores isn't sharpened is true - I read somewhere that as Prores is compressed its best to add a small amount of sharpening to match the look of RAW.  I can't remember where I read that but I remember it coming from a source beyond questioning - perhaps ARRI or RED or the like.  It makes sense, as does the idea they would match the perceived sharpness of RAW.

In a sense, Prores isn't just a codec, but a complete approach to the processing of the image.

The flavours of Prores will be interesting to see.

44 minutes ago, billdoubleu said:

Sure, people were complaining about not having professional editing codecs available internally in cameras like the GH5, but, those complaints began at least five years ago when Prores was already 10 years old. Prores was an option in every blackmagic camera alongside CinemaDNG starting in 2012, so, there's no logical argument that those codecs couldn't have been included in the GH5. The GH3 should've been the last GH camera sans Prores.

Here we are in 2022 and there is a growing list of RAW/ RAW(ish) codecs that didn't exist five years ago. ProresRAW, BlackmagicRAW, TICO-RAW are innovations sprung from the limitations imposed by RED's patents. These are all available internally in one or more cameras. Along with uncompressed RAW in the Sigma fp. So, while Prores is a welcome addition, no one should be impressed by it. Panasonic should either have developed a RAW technology by now or be licencing one.

It is unfortunate that Prores wasn't included in the GH4 and GH5, but the bitrates might have been more than they could handle.  With h264 there's no "right" bitrate, but with Prores there are standards, and it doesn't look so good on marketing if you're only giving people Prores LT, even though the bitrate of 4K Prores LT is 328Mbps - more than most other cameras and almost as much as the headline grabbing 400Mbps GH5 ALL-I codec.

Marketing is real, and often irrational, unfortunately.

In terms of saying prores doesn't matter because other cameras have internal RAW is just ridiculous.  It's like someone saying that their Ferrari doesn't have cupholders and someone else saying that most family sedans now have cupholders.  A different camera having a good codec doesn't matter if that other camera doesn't meet other criteria.  I can't go outside and capture images using the sensor of one camera, the colour science of a second camera, and the codecs of a third camera.

RAW is also different to Prores in that RAW tends to be a 1:1 sensor read-out, meaning that you either have to have the huge resolution and huge file sizes of the full sensor read-out or cope with some kind of crop which screws up your whole lens collection.  Lots of people shoot with a lower-resolution codec than their sensor and enjoy having the benefits of downsampling.  I am one of them, so RAW isn't of that much interest.  

One of the other benefits of Prores is that it was designed to be mostly indistinguishable from RAW under most conditions, so it's a very practical thing.  Otherwise, why would every / most cinema cameras offer it in addition to shooting RAW?

[Django]

2 hours ago, kye said:

Their sensor is great, but it's the processing that they do to the image that really sets them apart.  In fact, it's so valuable to them that they apply it in-camera rather than in LUTs that can be pulled apart and analysed.

Arri's overall image is an ensemble of things:

To create such outstanding images, all components of the imaging chain are custom designed by our engineers and carefully tuned for optimal performance, starting with the optical low pass filter, the CMOS sensor, the imaging electronics, and the image processing software.

The image processing step comes after the RAW image data:

Nothing is "baked" into an ARRIRAW image: Image processing steps like de-Bayer, white balance, sensitivity, up-sampling or down-sampling, which are irreversibly applied in-camera for compressed recording, HD-SDI outputs, and the viewfinder image, are not applied to ARRIRAW. All these parameters can be applied to the image in post.

ARRIRAW converter allows you to apply these image processing parameters to output ProRes 422/444 LogC files. You can also do it in-camera directly when shooting ProRes.

Their actual color science isn't secret:

 We openly publish our color science as standard SMPTE documents (Recommended Disclosure Document (RDD) 30:2014 and RDD 31:2014).

2 hours ago, kye said:

I'm not sure how much you know about colour science, but I have done many deep dives into pulling apart the colour science from a number of cameras, including Alexas, REDs, BMs, Canons, and Panasonics.  I have spent a lot of time on the colourist forums reading their responses, reading the articles they link to, and studying their methods (and when I say "studying" I mean opening Resolve and trying to re-create the methods they explain, and then using those methods on my own footage to get a feel for what is going on - like how you do when you do assignments at school - literally studying).

Ok but have you actually worked using Alexa/RED/BM/Canon RAW footage on an actual project? Because I have. I ask this, because out of all your researching you still seem to not grasp many basic aspects concerning RAW, Log & ProRes.

It sounds to me like you are mostly interested in matching your (GH5) camera footage to.. an Alexa. That's a fair intent and I'm sure you learn a great deal about CS & colour grading in the process but if that is your ultimate goal then perhaps buying Emotive Color matrix power grade will help you more?

2 hours ago, kye said:

You are absolutely right that each manufacturer has the opportunity to be building these things into their colour science (and not relying on their sensors), but the problem is that they just don't.  Year on year they are getting incrementally better but really aren't closing the gap between their $2K-5K cameras and what the leaders are doing.

The end result is that we're getting food that has come from the same ingredients (Sony sensors) and has only been seasoned with salt and pepper and therefore tastes rather bland in comparison to ARRI/RED who are demonstrating mastery in their use of spices.

Well, first off not every camera uses Sony sensors. Canon produce their own. And second, each brand and subsequent camera model has its own image processing, codec choice, log curve, colour matrix, optical lowpass filter (or lack of i.e Fuji XTrans) etc. 

It took a while for DSLR/MILC hybrids to disengage from the photo side and fully bring all the cine elements to their video modes but we're getting there now with RAW/10-bit Log/ProRes etc. DR/RS keep getting better.

I do think the gap is closing, even if ARRI/RED are still ahead.

2 hours ago, kye said:

In camera NR and sharpening are definitely an issue and I'd hope that implementing Prores will mean they'll tune the image to that codec and the expectations that pros would have.  I don't think the idea that Prores isn't sharpened is true - I read somewhere that as Prores is compressed its best to add a small amount of sharpening to match the look of RAW.  I can't remember where I read that but I remember it coming from a source beyond questioning - perhaps ARRI or RED or the like.  It makes sense, as does the idea they would match the perceived sharpness of RAW.

In a sense, Prores isn't just a codec, but a complete approach to the processing of the image.

ProRes is just a codec. A damn good one (especially ProRes 4444 XQ) but it's a lossy baked-in codec as are h264/h265.

Now how a camera decides to encode these codecs from RAW is where the whole image processing happens.

Adding sharpness to match the look of RAW on a compressed codec is a well known technique. Has nothing to do with ProRes itself.

Personally, I like to convert RAW myself and have full control on things like sharpness or NR.

ARRIRAW has a bunch of conversion settings, including precise sharpness, detail & NR but also various de-bayering algorithms that all shape the look.

2 hours ago, kye said:

One of the other benefits of Prores is that it was designed to be mostly indistinguishable from RAW under most conditions, so it's a very practical thing.  Otherwise, why would every / most cinema cameras offer it in addition to shooting RAW?

Again ProRes is just a codec. It's a pro standard and insures certain bitrates and bit-depths.

Like any compressed codec it is useful for file-size and fast turnaround. 

Not every project allows you the luxury of storing & converting RAW footage.

That said, ProRes remains lossy and image processed vs. RAW.

Saying it's indistinguishable from RAW is plain wrong. RAW can be pushed in almost infinite ways compared to ProRes. Especially if you bake in a Rec709 profile instead of Log to your ProRes output.

[kye]

3 hours ago, Django said:

Arri's overall image is an ensemble of things:

To create such outstanding images, all components of the imaging chain are custom designed by our engineers and carefully tuned for optimal performance, starting with the optical low pass filter, the CMOS sensor, the imaging electronics, and the image processing software.

The image processing step comes after the RAW image data:

Nothing is "baked" into an ARRIRAW image: Image processing steps like de-Bayer, white balance, sensitivity, up-sampling or down-sampling, which are irreversibly applied in-camera for compressed recording, HD-SDI outputs, and the viewfinder image, are not applied to ARRIRAW. All these parameters can be applied to the image in post.

ARRIRAW converter allows you to apply these image processing parameters to output ProRes 422/444 LogC files. You can also do it in-camera directly when shooting ProRes.

Their actual color science isn't secret:

 We openly publish our color science as standard SMPTE documents (Recommended Disclosure Document (RDD) 30:2014 and RDD 31:2014).

I'm keen to read more about this - can you link the source?

3 hours ago, Django said:

Ok but have you actually worked using Alexa/RED/BM/Canon RAW footage on an actual project? Because I have. I ask this, because out of all your researching you still seem to not grasp many basic aspects concerning RAW, Log & ProRes.

I'm not explaining the entire image pipeline in every post that I make, but even if I did no-one understands the details anyway so it wouldn't help.  Perhaps instead of just saying I don't understand things, speak to something tangible that I can talk to.

In terms of having used RAW footage?  Sure.  Probably more than a dozen.  I own three cameras that shoot RAW - one Canon and two BM.

What's your point?

3 hours ago, Django said:

It sounds to me like you are mostly interested in matching your (GH5) camera footage to.. an Alexa. That's a fair intent and I'm sure you learn a great deal about CS & colour grading in the process but if that is your ultimate goal then perhaps buying Emotive Color matrix power grade will help you more?

I'm trying to get the best colour possible.

For that I have pulled apart the colour science from many brands, trying to understand what they are doing.  I talk about ARRI mostly because they give the nicest colour by just applying their official LUT.  RED is right up there too, along with BM 2012.  BM 2018+ and Canon (RAW) are nice too, but aren't really at the highest level.

I own the Emotive Colour matrix and don't talk about it much here, partly because I've pulled it apart and don't want to give away too much, and partly because the it's so fragile to use - if the stars didn't align then you're not getting good results.  It doesn't like exposure changes much and doesn't deal with mixed colour temperatures basically at all, which is present in most situations I film in.

I have also bought film simulations from Juan Melara, downloaded DCTL plugins from professional colourists, LUTs from post production houses, and many other things.

3 hours ago, Django said:

Well, first off not every camera uses Sony sensors. Canon produce their own. And second, each brand and subsequent camera model has its own image processing, codec choice, log curve, colour matrix, optical lowpass filter (or lack of i.e Fuji XTrans) etc. 

It took a while for DSLR/MILC hybrids to disengage from the photo side and fully bring all the cine elements to their video modes but we're getting there now with RAW/10-bit Log/ProRes etc. DR/RS keep getting better.

I do think the gap is closing, even if ARRI/RED are still ahead.

Did you see the difference in image between the Alexa and the LF from that guy who did the comparison by splitting the light with a piece of glass to duplicate the cameras position between the two?

The other manufacturers may have been closing the gap but that was ARRI leaping ahead by a mile.  You don't seem to really acknowledge the differences, but I read lots of comments from people who are amazed at how much of an improvement they were able to make to what was already a world-leading performer.

I suspect colour matters more to me than to you.  Sadly, in the blind tests I always pick the most expensive cameras, even if I watch the comparison in 480p on YT.

3 hours ago, Django said:

ProRes is just a codec. A damn good one (especially ProRes 4444 XQ) but it's a lossy baked-in codec as are h264/h265.

Now how a camera decides to encode these codecs from RAW is where the whole image processing happens.

Adding sharpness to match the look of RAW on a compressed codec is a well known technique. Has nothing to do with ProRes itself.

You missed my point.

If I was a manufacturer including h264 in a camera, I know it's going to be targeted at consumers and videographers, so I will go a bit heavier on the processing because that's what this audience wants.  If I then decide to include Prores, I know it's for a different target market who have different expectations about the image, and so I'm likely to apply far less processing when the camera is set to record Prores rather than h264.

The anamorphic mode in the GH5 applies less processing than the 16:9 modes, so Panasonic clearly understand that it's for a different target audience.

Panasonic would be insane to include consumer amounts of processing and NR when the camera is set to record Prores.

3 hours ago, Django said:

Personally, I like to convert RAW myself and have full control on things like sharpness or NR.

Cool.  Personally I record in a lower resolution than the sensor and don't want the camera to crop.  For this, Prores is the winning option.

3 hours ago, Django said:

Saying it's indistinguishable from RAW is plain wrong. RAW can be pushed in almost infinite ways compared to ProRes. Especially if you bake in a Rec709 profile instead of Log to your ProRes output.

Saying that would be plain wrong.  Good thing I didn't say it!  Thanks for pointing that out?

I've done tests on a number of RAW cameras comparing the various bit-depths and also comparing the latitude of RAW vs Prores and mostly I'm ok with 10-bit Prores recording a log profile.  I'd prefer a 12-bit Prores if it's available, but I'll happily take Prores HQ.  

Colour science quality (and image quality in general) has two factors for me, the first is how good things look when under the optimal conditions.  This is how (you'd hope) most professionals working on controlled sets are working.  The other main factor of colour science is how robust the image is when conditions are far from perfect.  This is probably something you're not very experienced with, but it's the vast majority of clips that I shoot.

I've mounted my BMMCC and GH5 together and put them through a number of sub-optimal situations and then pulled the footage into Resolve and graded them side-by-side and the results are eye-opening.  The BM footage just does what you tell it to do, whereas the GH5 footage suffers almost immediately.  The sweet spot of the GH5 is has pretty good colour, not great but good, but that very quickly disappears when pushing and pulling the footage, even just adjusting the WB reveals that the small amount of colour magic it does have is pretty fragile.
I've graded files from the S1 and the colour in the sweet spot was nicer, the DR was higher, but the magic was still quite fragile and the footage felt like the GH5.  

I'm hoping that with colour science improvements and Prores that the files will have more colour science magic and that the magic will be more robust when graded.

[Django]

13 minutes ago, kye said:

I'm keen to read more about this - can you link the source?

Arri's website:

 https://www.arri.com/en/learn-help/technology/image-processing

15 minutes ago, kye said:

Did you see the difference in image between the Alexa and the LF from that guy who did the comparison by splitting the light with a piece of glass to duplicate the cameras position between the two?

Already gave you my thoughts on that.. 

17 minutes ago, kye said:

You don't seem to really acknowledge the differences, but I read lots of comments from people who are amazed at how much of an improvement they were able to make to what was already a world-leading performer.

I suspect colour matters more to me than to you.  

 I shoot RAW/Log, have certified Davinci Trainer experience, work in an ACES workflow, create my own LUTs from scratch, use power grades, film curves, film stock etc.. and when that's not enough, I hire a pro colourist.. so yeah colour matters to me lol.. but does X/Y/Z CS hold me back in any way? Nope. I think the difference between us is as a pro I'm simply not obsessed with matching ARRI CS. Neither are all those you referenced on Netflix that shoot on RED, Varicams, Venices etc. 

The goal is to have a clear vision for each project/scene. It's then up to DoP, lighting & color grading to take us there. 

Obviously this greatly varies depending on the budget/job type/talent. Another way to go is creating and loading custom LUTs directly in-cam. 

27 minutes ago, kye said:

Panasonic would be insane to include consumer amounts of processing and NR when the camera is set to record Prores.

Well they messed up / crippled V-Log, a paid upgrade. So until we know more details, I'll remain skeptical.

30 minutes ago, kye said:

Colour science quality (and image quality in general) has two factors for me, the first is how good things look when under the optimal conditions.  This is how (you'd hope) most professionals working on controlled sets are working.  The other main factor of colour science is how robust the image is when conditions are far from perfect.  This is probably something you're not very experienced with, but it's the vast majority of clips that I shoot.

Again with these assumptions. I'm not always working under controlled environments. Plenty of run&gun / docu experience under my belt. I'd never pick a BM for that type of work though. Uncontrolled environment is when I want built-in ND's, DR latitude, flexible ISO/DGO, IBIS/EIS, PDAF etc.  

[Marcio Kabke Pinheiro]

15 hours ago, webrunner5 said:

I don't understand it to be honest. The LX100 and the G7 I have had were really good, damn good CS wise. Even the GH3 was not bad, but man the GH4, GH5, well not worth a damn I think. They went backward.

Each one is really each one, hehehehe - I still have the LX100 (will never sell it), but I think that its colors have a very greenish cast (like the GH4, who is in the same generation). My GX85s and GX9 (specially the GX9) have m uch better colors for my tastes.

[Caleb Genheimer]

I agree that the Alexa sensor is definitely “gold standard,” but also that expecting that kind of performance out of something like a GH6 is a bit unrealistic. 
 

That being said, I wish the prosumer camera manufacturers would throw in the towel at around 6K resolution, and begin focusing on dynamic range in video. That directly corresponds to the technology previously discussed: dual gain output photosites. The current limiting factor for basically all prosumer camera DR is the ADC (analog to digital converter) bit depth of 12 bits, which limits performance to 12 stops of DR. Fancy processing can squeeze out a bit extra (like the ~12.7 stops of the S1H,) but that’s the limit.

The Alexa sensor’s photosites have TWO ADC each, at different gain settings, in parallel, for simultaneous capture (at the same moment.) The signals from these offset sensitivity ADCs are then combined by the image processor to cover a resulting higher DR. This is entirely possible in a prosumer camera at reasonable bitrates via LOG scale encoding. The real bottleneck is the ADC tech, which is linear in nature. 
 

For the vast majority of applications/users, 12-and-change stops is plenty enough to get a satisfactory image with room for flexibility, but those wanting to accurately simulate DR-dependent effects like Bloom, Halation, and flare, it is important to have as much differentiation in the highlights as is possible. I’ve found the S1H to have… enough DR for pretty reasonable halation approximation, providing I expose for the highlights. I’d take two more stops though if another prosumer camera offered it.

It sounds like the GH6 is using an entirely different approach to increase DR: sequential offset exposure. That’s not bad, necessarily, but it’s far more computational/AI in nature, and could definitely have motion artifacts. I feel we’re finally at the juncture where the internal codecs and bitrates on prosumer cameras have caught up to the capabilities of the sensors, and it’s time for attention to swing back to the underlying sensor tech for significant improvement. 
 

I’m also a bit sad to see the continuing use of DFD AF. Prediction: “3X faster processing” will just mean “3X faster micro jitter.” I’m ready for Tilta or DJI to make a stand-alone TOF/LiDar wireless follow focus AF/focus aid device. The tech is there in the Ronin 4D, and I’d willingly pay several thousand dollars to remove AF from all my future camera body purchase decisions.

[Django]

29 minutes ago, Caleb Genheimer said:

The Alexa sensor’s photosites have TWO ADC each, at different gain settings, in parallel, for simultaneous capture (at the same moment.) The signals from these offset sensitivity ADCs are then combined by the image processor to cover a resulting higher DR. This is entirely possible in a prosumer camera at reasonable bitrates via LOG scale encoding. The real bottleneck is the ADC tech, which is linear in nature. 

Canon have managed to reverse-engineer the tech:

Canon has published its white paper regarding the technology implemented on its newest cameras (C300 Mark III and C70) which is the Dual Gain Output (DGO) that allows 16 stops of dynamic range. The technology utilizes the same concept being used in the ARRI cameras for more than 10 years. Learn more below.  

https://ymcinema.com/2020/09/27/canon-releases-the-dual-gain-output-dgo-4k-sensor-white-paper/

[Caleb Genheimer]

9 minutes ago, Django said:

Canon have managed to reverse-engineer the tech:

Canon has published its white paper regarding the technology implemented on its newest cameras (C300 Mark III and C70) which is the Dual Gain Output (DGO) that allows 16 stops of dynamic range. The technology utilizes the same concept being used in the ARRI cameras for more than 10 years. Learn more below.  

https://ymcinema.com/2020/09/27/canon-releases-the-dual-gain-output-dgo-4k-sensor-white-paper/

Yes, Canon is doing something similar. They’ve definitely broken the 12-stop barrier, but most tests show the C70/C300III still underperform versus the ALEV sensors. Regardless, I’d rank those two cameras over even RED/Sony/Panny cinema cameras for this exact reason. I debated heavily on S1H vs C70 last year, but landed on the S1H for the 24X36 Open Gate mode, because I primarily shoot anamorphic, and wanted to get into large format scope capture. I’m happy enough with my S1H to never sell it. The only other digital camera I’ve kept is my GH2. A Samsung NX1, GH5S, and Pocket 4K have all come and gone. 
 

I also agree that the GH6 really needs to have properly implemented V-Log to differentiate from previous models. Going from V-Log L on the GH5S to full V-Log on the S1H was a paradigm shift, and I wouldn’t go backwards after having it. I don’t mind M4/3 sensors from a size perspective, but they DO struggle to implement the same imaging pipelines that often accompany larger sensors. I’d be considering a GH6 for b-cam if it had full V-Log, but without that, I’ll pass.

[Django]

8 minutes ago, Caleb Genheimer said:

Yes, Canon is doing something similar. They’ve definitely broken the 12-stop barrier, but most tests show the C70/C300III still underperform versus the ALEV sensors. Regardless, I’d rank those two cameras over even RED/Sony/Panny cinema cameras for this exact reason.

Indeed CineD reports 12.8-13.9 of DR on Canon's DGO sensor vs 13.8-14.8 on Alexa ALEV sensor.

That's a stop under for the Canon's which is pretty impressive.

On latitude test the Canon DGO also comes closest to ALEV with a 2-stop difference.

What's most impressive is that these Canon results are on the 10-bit XFAVC codec. 

Canon claim upcoming RAW update on C70 will increase DR, could be PR we'll have to see..

RED/Sony/Panny are still imo worthy competitors. For RS (or lack thereof on Komodo) and resolution.

You do make a strong point about +6K resolution bottlenecking DGO type DR. Makes better sense why C70/C300III didn't aim above 4K..

[hoodlum]

2 hours ago, Caleb Genheimer said:

I also agree that the GH6 really needs to have properly implemented V-Log to differentiate from previous models. Going from V-Log L on the GH5S to full V-Log on the S1H was a paradigm shift, and I wouldn’t go backwards after having it. I don’t mind M4/3 sensors from a size perspective, but they DO struggle to implement the same imaging pipelines that often accompany larger sensors. I’d be considering a GH6 for b-cam if it had full V-Log, but without that, I’ll pass.

The skates specs leaked this morning suggests it will have V-Log and will likely be priced at $2300 US

“The LUMIX GH6 contains V-Log/V-Gamut for the first time in LUMIX G Micro Four Thirds cameras and provides 12+ stops, and even 13+ stops, of wide dynamic range using Dynamic Range Boost mode. “

[kye]

10 hours ago, Django said:

Arri's website:

 https://www.arri.com/en/learn-help/technology/image-processing

I'm still not convinced about this.

Yes, they do say:
"Nothing is "baked" into an ARRIRAW image: Image processing steps like de-Bayer, white balance, sensitivity, up-sampling or down-sampling, which are irreversibly applied in-camera for compressed recording, HD-SDI outputs, and the viewfinder image, are not applied to ARRIRAW. All these parameters can be applied to the image in post."

However, immediately before that, they say:
"For the absolute best in image quality, for the greatest flexibility in post, and for safest archiving, the 16-bit (linear) raw data stream from the sensor can be recorded as 12-bit (log) ARRIRAW files."

So in this sense, the "nothing" baked into the ARRIRAW includes the combination of two streams of ADC and also a colour space conversion.  It's perfectly possible to do whatever you like to the image and still have the first statement be true in a figurative sense, which is how they have obviously intended it - "nothing you don't want baked in is baked in".  The ARRI colour science could still well be baked in and no-one would be critical and the statement they make would still be true in that figurative sense.

To me, the proof is in the pudding, and even that Alexa vs LF test included small and subtle shifts between the two images that is unexplained by the difference in lenses.

10 hours ago, Django said:

 I shoot RAW/Log, have certified Davinci Trainer experience, work in an ACES workflow, create my own LUTs from scratch, use power grades, film curves, film stock etc.. and when that's not enough, I hire a pro colourist.. so yeah colour matters to me lol.. but does X/Y/Z CS hold me back in any way? Nope. I think the difference between us is as a pro I'm simply not obsessed with matching ARRI CS. Neither are all those you referenced on Netflix that shoot on RED, Varicams, Venices etc. 

That doesn't mean you care about colour as much as I do, or see it in the way I do.  The fact you can hire a colourist means that the colour science from the manufacturer means less as you can cover any shortfall by hiring a pro, I don't have that luxury unfortunately, and thus why the colour science has a greater impact to me.

I suspect most people buying a GH6 also probably don't have regular access to a colourist to take up any shortfalls.  Ironically, the cheaper the camera the more that someone would need to have more latitude, great colour science, and a solid codec to work with.  By having the best image come from a camera that costs $100K you're giving the most robust image to the very people who need it least as they can afford to use the camera at its exact sweet spot and not need any of its latitude.

10 hours ago, Django said:

Well they messed up / crippled V-Log, a paid upgrade. So until we know more details, I'll remain skeptical.

Agreed, and being skeptical makes sense.  Realistically I'd be happy with any log colour space as long as it's a standard that is supported by ACES or RCM.

10 hours ago, Django said:

Again with these assumptions. I'm not always working under controlled environments. Plenty of run&gun / docu experience under my belt. I'd never pick a BM for that type of work though. Uncontrolled environment is when I want built-in ND's, DR latitude, flexible ISO/DGO, IBIS/EIS, PDAF etc.  

Like I said, the folks with access to higher-end equipment and colourists for backup and troubleshooting swap to the higher budget stuff when the going gets tough, but those of us who don't have that luxury are stuck with what we have and have to make the best of it without any backup.

But when we suggest that we'd prefer things that make our lives easier (robust colour science, codecs, DR, ISO performance, etc) rather than things that don't really help in difficult situations (resolution, etc) somehow that doesn't make sense to the people that aren't in our shoes?

I think the top comment on the Alexa vs LF video is the most telling.... "Based on image aesthetics I'd go with the 65, but based on my budget I went with my Panasonic G7."

8 hours ago, Caleb Genheimer said:

I agree that the Alexa sensor is definitely “gold standard,” but also that expecting that kind of performance out of something like a GH6 is a bit unrealistic. 

That being said, I wish the prosumer camera manufacturers would throw in the towel at around 6K resolution, and begin focusing on dynamic range in video.

Expecting ARRI level images from a GH6 is definitely unrealistic.  But I mention it for precisely the same reason you mention wanting more DR - the better they can make it the better our results will be.

8 hours ago, Caleb Genheimer said:

That being said, I wish the prosumer camera manufacturers would throw in the towel at around 6K resolution, and begin focusing on dynamic range in video. That directly corresponds to the technology previously discussed: dual gain output photosites. The current limiting factor for basically all prosumer camera DR is the ADC (analog to digital converter) bit depth of 12 bits, which limits performance to 12 stops of DR. Fancy processing can squeeze out a bit extra (like the ~12.7 stops of the S1H,) but that’s the limit.

The Alexa sensor’s photosites have TWO ADC each, at different gain settings, in parallel, for simultaneous capture (at the same moment.) The signals from these offset sensitivity ADCs are then combined by the image processor to cover a resulting higher DR. This is entirely possible in a prosumer camera at reasonable bitrates via LOG scale encoding. The real bottleneck is the ADC tech, which is linear in nature. 

For the vast majority of applications/users, 12-and-change stops is plenty enough to get a satisfactory image with room for flexibility, but those wanting to accurately simulate DR-dependent effects like Bloom, Halation, and flare, it is important to have as much differentiation in the highlights as is possible. I’ve found the S1H to have… enough DR for pretty reasonable halation approximation, providing I expose for the highlights. I’d take two more stops though if another prosumer camera offered it.

Magic Lantern gives options for 14-bit RAW on the 5D and also on my 700D, so I'm assuming that the Canon sensors have a 14-bit output.  I'd assume that means it's not that difficult to do - after all the 5D isn't a new camera by any means!

According to CineD the GH5 gets 9.7/10.8 stops, the OG BMPCC gets 11.2/12.5, the S1H gets 12.7/13.8 and the Alexa gets 14/15.3.  I really notice a difference between the GH5 and OG BMPCC, and obviously the S1H is significantly more again.

Have you shot with an Alexa?  If so I'm curious to hear if you noticed any differences from the extra DR when using them?  Obviously there's an amount of DR that would be "enough" for almost all situations, and I'm curious to know where that amount is.

8 hours ago, Caleb Genheimer said:

It sounds like the GH6 is using an entirely different approach to increase DR: sequential offset exposure. That’s not bad, necessarily, but it’s far more computational/AI in nature, and could definitely have motion artifacts. I feel we’re finally at the juncture where the internal codecs and bitrates on prosumer cameras have caught up to the capabilities of the sensors, and it’s time for attention to swing back to the underlying sensor tech for significant improvement. 

I'm also skeptical about motion artefacts that would come from non-sequential exposures.  I guess we'll see.  It might be more useful in situations where motion blur isn't significant and extra DR is, for example locked-off shots of architecture and the like.  You could easily have it take a set of images around the 1/1000s mark, perhaps at 300fps, and then combine them.  That would give you great DR to include the highlights in the sky etc.

Based on your logic about 12-bit ADC being a limitation and new sensor tech being required it will be interesting to see what this new sensor can do.

Only a few sleeps left until the formal announcement!

[hyalinejim]

16 hours ago, kye said:

the GH5 is has pretty good colour, not great but good, but that very quickly disappears when pushing and pulling the footage, even just adjusting the WB reveals that the small amount of colour magic it does have is pretty fragile.

Have you tried WBing GH5 10bit VLog in an ACES colour space? You should be able to do it with relative impunity.

I know there's no GH5 IDT, but you can sandwich Vlog to ACES to VLog and get your footage back to what it was, but with WB fixed.

Incidentally, I don't think that GH5 VLog should be treated as Rec709 gamut. Admittedly, converting to V-Gamut gives slightly wonky colour (slightly pink skin) but leaving it as is gives much worse (horribly green skin).