ProRes RAW

The moment Apple killed PC support for ProRes they killed ProRes.

I can't get excited for a new branch of a format that has zero future in the PC world (which accounts for the majority of video/film productions).

DNxHR is the universal future, and we all should be championing universal formats that don't exist just to sell Brand X computers.

I shoot mostly for the network/broadcast world and I have NEVER been asked for DNx(Avid). But I have lost count of the times I've been asked for ProRes.
 
The moment Apple killed PC support for ProRes they killed ProRes.

I can't get excited for a new branch of a format that has zero future in the PC world (which accounts for the majority of video/film productions).

DNxHR is the universal future, and we all should be championing universal formats that don't exist just to sell Brand X computers.

Not sure if this is trolling or something but.....really? ProRes is 'killed' in the postproduction world?
And DNxHR is 'the future'?

There sure are a ton of people who disagree with you judging by the number of times I've
been asked for ProRes. And what compressed format does the Alexa shoot to again?

I get it, you are a PC user and don't like Apple. But c'mon now.
 
This is definitely good for people shooting RAW ProRes - but I don't really believe performance is going to be what they say. RAW may be smaller in size and data rate over its alternatives but are we forgetting about debayering quality settings, color space and gamma conversion in "real-time"?

Okay so I read the white paper and a little more about this, turns out my initial speculation is quite inaccurate.

Regarding the real-time performance, Apple says: "ProRes RAW brings to raw video the same first-rate performance and ease of editing that ProRes has brought to conventional video.
Raw video playback requires not only decoding the video bitstream but also demosaicing the decoded raw image. Final Cut Pro includes a streamlined and GPU-optimized demosaicing algorithm designed to meet the performance demands of real-time, multistream video editing."

Atomos has also claimed on their website that ProRes RAW will be 5.7 times faster to render than Redcode RAW 3:1, and still 1.5 times faster than Canon C200 RAW Lite.

Immediately supported cameras so far are: Canon C300k Mark II/C500 and Sony FS5/FS7/FS7 II and even the FS700 and I wonder if any camera already licensing ProRes codecs may be able to include this via a firmware update? Or, is this a new encoding chip that is required?
 
What will you sacrifice in quality for a demosaic algorithm that is optimized for speed? IMO Resolve has hit the sweet spot for quality vs speed so far. The really top notch algorithms used by stills processors like Capture One or Raw Therapee can take seconds per frame.
 
I shoot mostly for the network/broadcast world and I have NEVER been asked for DNx(Avid). But I have lost count of the times I've been asked for ProRes.

I shoot segments every now and again for CBS' Late Late Show and their post folks have us transcode into DNx so they can ingest directly in to their Avids. For certain projects that will air same day, we have to use the specific flavors that will fast ingest, recorded onto Shoguns on the DIT cart.
 
Immediately supported cameras so far are: Canon C300k Mark II/C500 and Sony FS5/FS7/FS7 II and even the FS700 and I wonder if any camera already licensing ProRes codecs may be able to include this via a firmware update? Or, is this a new encoding chip that is required?
Since the Shogun Inferno can do it and it's a couple of years old, I doubt any new chip is required.
 
good point, but the device has to facilitate any update to programmable codec encoding chips, and what if there are cheaper encoding chips used which can not be reprogrammed?

just a thought anyway, I have no idea how it works or what limits there can be.
 
My big concern is that Apple has a habit of dropping products that they have decided they no longer want to support, and do so without much warning. I hate to rely upon them for anything outside of the Apple infrastructure. I'd prefer to see CDNG get more support, or use a truly open RAW that all platforms can use for free.

Another DNx is the EVS, mostly used in sports production, but it likes to export as DNx and can do other codecs as well. Internally it uses some kind of special sauce.
 
What's the likelihood Atomos can take uncompressed 10-bit 422 YCbCr from HDMI or SDI and convert to ProRes RAw??
 
Last edited:
My big concern is that Apple has a habit of dropping products that they have decided they no longer want to support, and do so without much warning. I hate to rely upon them for anything outside of the Apple infrastructure. I'd prefer to see CDNG get more support, or use a truly open RAW that all platforms can use for free.

Agreed:

Tombstones.jpg
 
I shoot segments every now and again for CBS' Late Late Show and their post folks have us transcode into DNx so they can ingest directly in to their Avids. For certain projects that will air same day, we have to use the specific flavors that will fast ingest, recorded onto Shoguns on the DIT cart.

Fair enough. I wasn't trying to say that no one uses it period(I try to avoid speaking in terms of absolutes in this industry), just that between the two, ProRes still seems to be the more preferred. Actually, you can't even say which is THE preferred codec(if there even is one) within the same company, sometimes. It depends on so much. The last shoot I did for CBS was for composite elements for the Super Bowl a few years ago. We were shooting 4K XAVC. Some stuff I do for NBC Sports, ProRes is their preferred codec. And another is all over the map: Some still say ProRes, most now don't care and are fine with XAVC, MXF, ProRes, DVCPROHD and some of the features guys shoot on RED and Amira. It depends on the producer and "where" it's going/being edited.

I understand a workflow and "pipeline" being in-place(and shooting and technical specs), but it cracks me up that in this day and age there are organizations(and I'm speaking more on my end of the world instead of narrative production) that still say the cannot handle format/codec "X". It has to be "Y". One of my friends will often retort to the "we can't handle X" with "Do you want me to take you over to the middle school next to my house so that the 13 year olds can show you how to do that on their laptops?".
 
What's the likelihood Atomos can take uncompressed 10-bit 422 YCbCr from HDMII or SDI and convert to ProRes RAw??

thats not how raw works. raw is exactly how it sounds. the raw data from the image sensor. Uncompressed is no longer that raw signal but converted to YUV video data with a lot of color tossed out to handle that 10bit, YCbCr conversion and the 4:2:2 conversion. In no way shape or form can you get back to a raw signal from that. You would need one heck of a custom conversion method in the device and it would be almost impossible to do due to the color loss. How can you take a single RGB pixel and convert it back to a set of 4 raw values accurately by using the surrounding raw values to determine what each raw pixel actually was?
 
What will you sacrifice in quality for a demosaic algorithm that is optimized for speed? IMO Resolve has hit the sweet spot for quality vs speed so far. The really top notch algorithms used by stills processors like Capture One or Raw Therapee can take seconds per frame.

Likely much better than the in camera processing used when recording ProRes formats. Apple may have their faults but quality usually isn't one of those faults. We will have to wait and see what the final outcome will look like but if I had to take a wild guess I would say its really good. Of course you can't speed up that process and compress raw to this level without some level of sacrifice so we will have to see what that actually looks like next week.
 
thats not how raw works. raw is exactly how it sounds. the raw data from the image sensor. Uncompressed is no longer that raw signal but converted to YUV video data with a lot of color tossed out to handle that 10bit, YCbCr conversion and the 4:2:2 conversion. In no way shape or form can you get back to a raw signal from that. You would need one heck of a custom conversion method in the device and it would be almost impossible to do due to the color loss. How can you take a single RGB pixel and convert it back to a set of 4 raw values accurately by using the surrounding raw values to determine what each raw pixel actually was?
Right, so that why I have my doubts about using ProRes technology, which uses Y'CrCb internally, applied to raw data.

From Apple's whitepaper:

"Apple ProRes RAW is based on the same principles and underlying technology
as existing ProRes codecs, but is applied to a camera sensor’s pristine raw
image data rather than conventional image pixels. "


It seems they simply mangled raw data into Y'CbCr, use ProRes compression and unmangle it back to raw.
I see many potential issues with that.
 
Back
Top