Red, can you make this happen? 1920x1080 option?

[Joseph Hutson]

I don't know if this is a thing you guys need to yet learn, or you can but haven't decided to include it, but if anything else could be changed, I would like the capability to record HD 1920x1080 within the Scarlet without having to downrez in post.

I am not asking for higher framerates because of lower resoulution...just the option to record in HD.

Jim, you know more than me so just tell me if this is something not possible.

Thanks for listening to people like us who express their ideas.

P.S. I remember you saying interchangeable lenses on the 2/3" Scarlet wasn't gonna happen. Maybe this just another one of those things that aren't...but could before the release.

 

[Gary Stone]

i believe the scarlet will have (at some point) 1080p out through HD-SDI so you could just find a way to record off that output. somebody chime in if i'm an idiot and have this wrong.

 

[Joseph Hutson]

Yeah, I am sure you could do that, but that's not really what I am asking.

I am not demanding this feature as some people might make demands. This is just a "humble" request. :biggrin:

 

[Joe Carney]

Yeah, I am sure you could do that, but that's not really what I am asking.

I am not demanding this feature as some people might make demands. This is just a "humble" request. :biggrin:

Scarlet will offer 2K that you can downrez to 1080p on final render. Should work out fine for you.

 

[Joseph Hutson]

Scarlet will offer 2K that you can downrez to 1080p on final render. Should work out fine for you.

Only the S35 Scarlet gives the option of 2K recording.

 

[David Mullen ASC]

Just note that 1080P 4:4:4 is not all that less, data-wise, than 4K RAW, assuming the same compression rate, so even if they give you that option, it won't necessarily mean a lot of advantages in recording times, frame rates, etc. over 4K RAW.

And converting 2K RAW in-camera to 1080P 4:4:4 (RGB) will generate even more data to record than keeping it 2K RAW. 2048 x 1152 (16x9) is 2.36MP to record, whereas 1920 x 1080 4:4:4 is 3 x 2.07MP, or 6.22MP to record -- a lot more data.

 

[Jeff Kilgroe]

Hmmm.... Native 1920x1080p recording...

But would we really want this option? If it's a cropped record mode, then we're essentially gaining little over other < 2/3" HD cameras. Better compression and the RAW workflow, yes, but no advantages for resolution. 3K provides excellent oversampled down-rez to 2K / 1080p.

Why is it that you want native 1080p recording? By the time Scarlet ships, we will have more media options to choose from, like 32GB CF cards. The difference between transcoding 2K/1080p and 3K isn't that big. Or at least not big enough to warrant another mode that would just be a crippled version of what the camera is designed to do.

If you care about image quality, and you obviously do if you're buying a Scarlet, then you will want all the resolution you can get. The best looking 1080p deliverables are derived from sources with greater than 1080p HD resolution. I'm just curious about why you want this feature -- like many of the features being requested since the 11/13th announcement, a lot of them would be a step backward in relation to what RED is trying to do here.


I would imagine that the Scarlet will offer 720p and 1080p playback from your 3K recording. If that helps.

 

[Mark K.]

Just note that 1080P 4:4:4 is not all that less, data-wise, than 4K RAW, assuming the same compression rate, so even if they give you that option, it won't necessarily mean a lot of advantages in recording times, frame rates, etc. over 4K RAW.

And converting 2K RAW in-camera to 1080P 4:4:4 (RGB) will generate even more data to record than keeping it 2K RAW. 2048 x 1152 (16x9) is 2.36MP to record, whereas 1920 x 1080 4:4:4 is 3 x 2.07MP, or 6.22MP to record -- a lot more data.

David, that's the main reason I can't see much point in having on-board RGB recording in the cameras. However down-scaled 720p and 1080p REDCODE would offer the sort of benefits in file-sizes and recording times that some people are looking for.

I know that most of my work still only requires standard-definition delivery, so smaller file sizes and shorter processing times certainly have a lot of appeal for me (and I imagine many others too).

Hmmm.... Native 1920x1080p recording...

....

If you care about image quality, and you obviously do if you're buying a Scarlet, then you will want all the resolution you can get. The best looking 1080p deliverables are derived from sources with greater than 1080p HD resolution. I'm just curious about why you want this feature -- like many of the features being requested since the 11/13th announcement, a lot of them would be a step backward in relation to what RED is trying to do here.

Jeff, the appeals of the smaller formats are for those of us who know we're going to have a lower-end delivery format for projects we're shooting. I think the mantra "keep it simple stupid" really factors in here. Because if I know that something I'm shooting is only ever going to see a run of SD DVDs, then working with 720p REDCODE and downscaling it to a crisp-looking SD makes significantly more sense (for a streamlined workflow) than downscaling to SD from a source file that's either 4x or 12x bigger than 720p for what will probably be a negligible increase in image quality in the final SD output.

It's not that we don't want to shoot 3k/5k as much as possible, it's just that those sorts of resolutions (and their accompanying file-sizes) don't make sense for certain projects.

Now I know that some people will say: "just shoot on video then", but with the price of a camera like Scarlet looking extremely reasonable compared to those of conventional video-cameras. For those of us wanting to own our own cameras, having that flexibility remains important.

 

[Stephen Gentle]

Hmmm.... Native 1920x1080p recording...

But would we really want this option? If it's a cropped record mode, then we're essentially gaining little over other < 2/3" HD cameras. Better compression and the RAW workflow, yes, but no advantages for resolution. 3K provides excellent oversampled down-rez to 2K / 1080p.

Why is it that you want native 1080p recording? By the time Scarlet ships, we will have more media options to choose from, like 32GB CF cards. The difference between transcoding 2K/1080p and 3K isn't that big. Or at least not big enough to warrant another mode that would just be a crippled version of what the camera is designed to do.

If you care about image quality, and you obviously do if you're buying a Scarlet, then you will want all the resolution you can get. The best looking 1080p deliverables are derived from sources with greater than 1080p HD resolution. I'm just curious about why you want this feature -- like many of the features being requested since the 11/13th announcement, a lot of them would be a step backward in relation to what RED is trying to do here.

I agree - I don't really see a problem with shooting 3K even though I'll mainly be going to 1080p after post. I'll still work at 3K though so you can 1) make a higher quality version of your work when a higher resolution delivery format comes out, and 2) scaling the whole project down at the end will make titles and effects look better as well as the footage.

 

[Jeff Kilgroe]

Jeff, the appeals of the smaller formats are for those of us who know we're going to have a lower-end delivery format for projects we're shooting. I think the mantra "keep it simple stupid" really factors in here. Because if I know that something I'm shooting is only ever going to see a run of SD DVDs, then working with 720p REDCODE and downscaling it to a crisp-looking SD makes significantly more sense (for a streamlined workflow) than downscaling to SD from a source file that's either 4x or 12x bigger than 720p for what will probably be a negligible increase in image quality in the final SD output.

I understand what you're thinking / saying. However, I don't think the compromises would make the lower res shooting modes practical or even possible. The 2/3" Scarlet is designed to shoot at 3K resolution... Less than that and you are cropping the sensor. 1920x1080 crop of the 3K sensor gives you a sensor area that is smaller than a 1/3" consumer camcorder! Yes, really, like 0.265", just slightly bigger than 1/4". Now just think what that does for focal lengths, FOV, perception of DOF, etc.. If there was a compromise, by only cropping down to 2K, that brings the recorded sensor area up to about 0.29". Seriously, pulling 2K / 1920x1080 from a cropped region of the 3K 2/3" sensor makes no sense.

OK, so then what about scaling the 3K image to 1920x1080p in-camera? Problems... How do you scale the RAW bayer data? You have to quantize and process the photosite data into something meaningful as a rasterized image in order to scale it. This essentially means that you will be getting a 12 to 16 bit RGB or YCbCr image at 3K, scaled to HD res and then compressed with something similar to REDCODE RAW -- perhaps something like the long lost REDCODE RGB modes once theorized for the RED One. These will take a LOT of headroom and horsepower in-camera and will eat up more recording space per pixel than simply recording the 3K REDCODE RAW. Not to mention that the RAW image must be built into something more tangible in-camera with the appropriate color space and some form of LUT. You would be losing many benefits of working with RAW... May as well go shoot HDCAM or AVC.

Anyway, that's how I'm seeing it. And for what it's worth, for most of my SD DVD destined projects (probably half of what I shoot), I normally shoot 4K on the RED One. One-light and process through RED Alert and batch out via Redrushes to 720p or 1080p ProRes at half-res high debayer and it goes quick. We can also now wrap right into native R3D QT at 2K res from 4K footage and edit away in FCP. But the ProRes transcode I mentioned above is quick and painless and I tend to batch out to 720p ProRes all the time and use that for offline editing rather than the inefficient and restrictive R3D proxies. ...3K to 720p HD is not a big deal in post. In another 6 months from now, the situation will be even better with more options.

 

[Mark K.]

Jeff, I'm not referring to shooting smaller formats using crops of the sensors; rather, using the full sensor but recording the RAW RedCode files at smaller resolutions.

Canon are already doing this (reduced resolution RAW output) in their 50D and their 5DII DSLRs. So I'm working on the assumption that same technology could be implemented for the Red cameras.

The key would be maintaining the same RedCode RAW workflow that you'd normally use with the camera, but simply in lower resolutions.

 

[Jeff Kilgroe]

Jeff, I'm not referring to shooting smaller formats using crops of the sensors; rather, using the full sensor but recording the RAW RedCode files at smaller resolutions.

You didn't read all of my post.

Canon are already doing this (reduced resolution RAW output) in their 50D and their 5DII DSLRs. So I'm working on the assumption that same technology could be implemented for the Red cameras.

They cheat. They're not actually recording true scaled RAW. Jim referred to it as a "parlor trick" and something RED would not do.

The key would be maintaining the same RedCode RAW workflow that you'd normally use with the camera, but simply in lower resolutions.

And how does one actually scale RAW bayer data to retain all the benefits of a RAW workflow?

 

[Mark K.]

You didn't read all of my post.

I did read all of your post, but as I said my thoughts on the matter are based on the assumption that Canon are managing to do this with their DSLRs at the moment, and that the technology could be scaled for these Red cameras.

They cheat. They're not actually recording true scaled RAW. Jim referred to it as a "parlor trick" and something RED would not do.

I was unaware of this, and if it's true. Then that voids the assumption my thoughts were based on. Between what I'd heard of the reduced-resolution RAW files the Canon DSLRs could do, and the initial suggestions of Red of a "scaled" 2k option in the R1 I'd simply assumed that scaling the RAW bayer data was possible. If it's not possible, then I'll simply move on and transfer my energies into harping on about something else! :biggrin:

And how does one actually scale RAW bayer data to retain all the benefits of a RAW workflow?

I have no idea, I thought that was what Graeme's Magic Pixies were for.

 

[Jeff Kilgroe]

In a nutshell, what Canon is doing is firing off a pattern of photosites for each frame of the reduced resolution RAW and for video, not the entire frame. Fewer photosites in operation will decrease the read reset times. For video, this means a faster "rolling shutter" and less apparent skew. Even with that, the skew is horrible, but not as visibly bad as on the Nikon D90. The used photosites and their RAW data is binned together into a lower-res image, but this introduces huge problems. The spatial gaps of the non-fired photosites create mis-matches on fine details. This leads to huge aliasing and moire problems and a further loss of detail as these are smoothed out in their RAW processing, many can not be eliminated at all. Canon is not recording video in any form of RAW. and you will find that most of the pro (or should I say promotional) videos out there for the 5Dm2 are showcasing its low-light performance. The camera does indeed have very good low light ability, but shooting this sort of material hides many of the flaws introduced with their photosite skipping process. Most daylight shots I have seen are either highly over compressed for the web, over processed in post to be excessively contrasty or harsh in some way or they exhibit very obvious edge-aliasing and issues with finely detailed patterns.

As for Graeme's Magic Pixies, he's working their magic every day... If anyone can find a way to make this happen, it would be him. To be fair in this, I'll say that I'm making some assumptions too. Based on the situation with the RED One and other things that have been said by RED team members. I know that RED will not do what Canon is doing. However, they may be a real way to provide for scaling RAW bayer data... If there is, then we have to wonder if Scarlet will be powerful enough to handle the task -- it has already been staten on many occasions that it will have much less processing power than EPIC and in many cases fewer capabilities than the RED One. If RED does find a way to take 3K RAW and convert it to 1920x1080 RAW without any side-effects, then I'm all for it. But I don't see it as a top priority on their list when the current workflow and performance level for working with 3K REDCODE is not that bad, only getting better. I see the biggest benefit to lower resolution recording to be in media capacity -- 1080p would take up less than half as much space vs. 3K. But this is a problem that I see the industry solving within the next year as larger and faster FLASH media rolls into the market.

 

[David Mullen ASC]

RAW to RAW downscaling isn't going to happen anytime soon, and it's highly problematic. Just think about the Bayer patterned filters over photosites and then imagine what it would take to downscale that successfully without artifacts, without converting it to RGB first.

There are RAW cameras that offer 1080P outputs, like the ARRI-D21, but again, that's RAW to RGB, with very high data rates for 1080P using HD-SDI outputs.

Rather than RAW to RAW downscaling, you'd more likely want to convert RAW to RGB and then compress 1080P 4:4:4 or 4:2:2 into something more manageable, like Panasonic does for their P2 recordings. But then you'd be living with the quality of real-time RAW debayering.

 

[Tom Lowe]

RAW to RAW downscaling isn't going to happen anytime soon, and it's highly problematic. Just think about the Bayer patterned filters over photosites and then imagine what it would take to downscale that successfully without artifacts, without converting it to RGB first.

There are RAW cameras that offer 1080P outputs, like the ARRI-D21, but again, that's RAW to RGB, with very high data rates for 1080P using HD-SDI outputs.

Rather than RAW to RAW downscaling, you'd more likely want to convert RAW to RGB and then compress 1080P 4:4:4 or 4:2:2 into something more manageable, like Panasonic does for their P2 recordings. But then you'd be living with the quality of real-time RAW debayering.

Canon has something called sRAW, which is typically a 1/4 or 1/2 scale RAW image, with no crop factor.

I found this description of how it (supposedly) works:

“ It needs to be added that 1/4 size sRAW does pixel binning of 4 neighboring pixels in Bayern pattern:
G B
R G
into a single pixel with (R, G, B) values. The effect is pretty similar to what Foveon or film-capture does; since the need for demosaicing disappears (i.e. you don’t need to interpolate for the missing color values), the resultant image becomes quite sharp. I’ve been finding the 2.5mp sRaw option on 10mp 40D to be extremely sharp and noise-free.“

The trouble with this request for 1080p is that it will probably require a lot of computational horsepower that would drive up the size and cost of the 2/3" Scarlet, which should be kept as light and simple as possible, IMO.

 

[David Mullen ASC]

I can see how 1/4 RAW downscaling could be accomplished using that method, but of course, that makes 4K RAW into 1K RAW, not HD (1.9K)... and then you'd have to convert 1K RAW into 1.9K RGB...

The trouble is that there is no such thing as HD RAW anyway -- HD is a video format. So 1920 x 1080 RAW wouldn't be any existing HD format; it would still require conversion to an existing HD standard. And it sounds like what some people want is a RED camera that records an HD signal that doesn't need post processing or conversion, i.e. it records 1080P 4:2:2 or 4:4:4, not 1.9K RAW that needs further conversion.

And once you start talking about recording 1080P 4:2:2 or 4:4:4, you have to decide what compression scheme to use because uncompressed HD data rates aren't much lower than REDCODE-compressed 4K RAW.

 

[Stu Mannion]

RAW to RAW downscaling isn't going to happen anytime soon, and it's highly problematic. Just think about the Bayer patterned filters over photosites and then imagine what it would take to downscale that successfully without artifacts, without converting it to RGB first... you'd be living with the quality of real-time RAW debayering.

The trouble with this request for 1080p is that it will probably require a lot of computational horsepower that would drive up the size and cost of the 2/3" Scarlet, which should be kept as light and simple as possible, IMO.

Thanks for the insights guys. Perhaps 3k raw is the only way to go.

 

[Daniel Browning]

Canon has something called sRAW, which is typically a 1/4 or 1/2 scale RAW image, with no crop factor.

sRAW is the laughing stock of file compression methods. Half of it is wasted on two embedded jpegs, and the rest contains only a quarter of the resolution of the original.

If they removed the embedded jpegs, assigned bits on a curve, and used less bits for higher and higher ISO settings, they could reduce file sizes to 1/8th for ISO 3200, without losing *anything* from the image. Advanced wavelet compression is the next best step to reduce file size without compromising quality.

RAW scaling (AKA very lossy compression) is a good option when all the more basic compression methods have been exhausted and the camera is too weak to do any kind of l visually lossless compression.

 

[Gary Stone]

1080p is listed in all models of scarlet and epid. holy crap.