How will the SI-1920HDVR compete against RED

[Glenn Gipson]

I'm not trying to start a hardware flame war or anything. Trust me, I'm not. I'm only bringing this up because I want to see the SI-1920HDVR succeed. But I was wondering how the makers of this camera hope to combat the RED Camera. Yes, I know that the RED camera is not out yet, but if it does live up to its specs and its price, how can the SI-1920HDVR hope to compete? I'm really having trouble seeing how the SI-1920HDVR has any advantage over a theoretical RED camera. After all, the SI-1920HDVR cost more than the Red Camera, and does a lot less too. It seems to me that the SI-1920HDVR's only hope is to come down in price if it hopes to compete, no?

[Brian Drysdale]

The price quoted for the RED is only for the camera section, it doesn't include a method of recording or other accessories. The SI offers the same set up for $12.5K.

RED's unique selling point is the 4k 35mm size sensor. However, making full use of that feature won't be cheap. However, I suspect a lot of people will be using 16mm lens and shooting 2k on the RED and then the SI's price is competitive. The cost of a camera is more than just one part, it's the cost of the whole system needed to shoot a film.

At the moment SI's advantage is they have working prototypes that are currently being field tested on productions. Currently, the RED is in the R & D stage on the bench, it's big advantage is the marketing.

[Glenn Gipson]

>>RED's unique selling point is the 4k 35mm size sensor. However, making full use of that feature won't be cheap. However, I suspect a lot of people will be using 16mm lens and shooting 2k on the RED and then the SI's price is competitive. The cost of a camera is more than just one part, it's the cost of the whole system needed to shoot a film.<<

I thought you had to use a 35mm size lens with the RED Camera?

[Jason Rodriguez]

After all, the SI-1920HDVR cost more than the Red Camera, and does a lot less too.

One needs to think about what "a lot less" means . . . .

Yes, the RED has a larger sensor . . . but notice that on-board you can only shoot 2K resolution, if you want 4K, you're going to have to cough up a lot of cash to create a recording solution that can ingest via infiniband and record at up to 1GB/s (10Gb/s) for 4K/60p.

Also we are going to have a higher signal-to-noise ratio with our sensor than what they are quoting, so our images should be cleaner/less noisy, which transfers into more dynamic range.

Additionally our camera will be 2K. So basically for most intents and purposes, you're looking at a 2K Silicon Imaging camera vs. a 2K RED.

RED does have the advantage that their 2K will be a 2x oversampled 4K image . . . but they will not be shooting RAW like we will (it will be processed RGB or YUV), especially in a format like Cineform RAW that gives you 4 streams of real-time effects at 2K on "decent" dual-Opteron workstation.

If that's the case, then doing things like re-doing the color matrix, re-dogin the white-balance for optimal dynamic range in the highlights, going back to photometrically linear, etc. is impossible with the RED camera.

Another thing to consider is that they have to-do a real-time demosaic in-camera to get the 2K RGB images. We have the luxury of doing some very complicated algorithms in post that will give you super-high quality images . . . that will mitigate the advantage of their 2x oversampled 2K, since the method by which they are generating the RGB images will be inferior to how we will be demosaicing . . . as a result, our 2K vs. their 2x oversampled 2K images will be on a more level playing field than the numbers would suggest. A 2x oversampled image from a weaker real-time demosaic algorithm vs. a non-oversampled image from a very strong adaptive algorithm could end up being quite even in overall resolution, lack of image artifacts, and sharpness.

We will have an uncompressed RAW mode (up to 2K again) that will record 12-bit linear, and an exporter to DNG file sequences so you can ingest the footage straight into After Effects or any other third-party RAW converter on the market that accepts DNG files (a number of them do for still cameras).

We have some very nice playback capabilities, and there are also a number of other tricks that we can do because we're a computer.

For instance, with the RED, while you're recording, can somebody else be playing the files off the machine, adding metadata as they play it back, etc.? Can you be editing the files *right on the camera* while your DP is shooting (if you wanted to-do that)? In this case, the camera *is* the computer, it *is* (or can be) the central nervous system of the shoot . . . the camera is not a separate ingest point. Red AFAIK doesn't have shared storage capabilities. So when it comes to group workflows on set, this is a very powerful feature that I don't think should be ignored, since we're now talking about very powerful colaborative features. Sure you can take an HD-SDI feed off the RED and pipe that around the set, but I don't think that's as efficient as workign directly with the RAW files right from the source of recording (like Spoon is doing right now).

And again, don't ignore the fact that we're shooting RAW, while RED is shooting to processed RGB/YUV (they will have a RAW mode, but their codec is not going to be as good as our Cineform RAW codec which lets you work with the RAW material in real-time in Premiere/After Effects, Final Cut, or any QT capable application. Also you can't record RAW mode on-camera). RAW is enabling you to basically go back to the source, and manipulate the data as you need, not how the camera does it on the day of the shoot. Don't like the white-balance . . .change it. Don't like the gamma curve . . . okay, change it. Don't like the color matrix (and default color "look") . . . no problem, change it. Hate the demosaicing algorithm . . . what the heck, change it.

And the REALLY nice thing is that since we're software, if RED does come out with a functionality feature that everyone really likes, chances are, we can add it free of cost to the end user in a very quick time-frame :) And of course on less of a copy-cat note, we can innovate and evolve our software much faster to the needs of the end user than RED can do on their camera (how many firmware upgrades have you seen that actually do major functionality changes to your hardware on the same scale as software changes?).

So you can see how there's lots of flexiblity that we'll be providing with our camera/software combination that RED can't provide.

[Brian Drysdale]

Originally Posted by Glenn Gipson

>>

I thought you had to use a 35mm size lens with the RED Camera?

You can have a window for other image sizes. RED currently seems to just offer a PL mount for the lens, which is also used on modern 16/S16 lenses.

The SI also offers C mount, which is an extremely useful feature if you've got a limited budget.

[Bob Grant]

I can't see much point in putting a 16mm lens on a 35mm sensor, putting a 16mm lens on a cheaper 2/3" sensor does make sense.
I suspect there's not a huge difference cost wise between a 2/3" sensor as used in the SI camera and a 35mm sensor used in the RED.
Where there is a big difference is between 16mm and 35mm cine lenses.
Where there's another big difference is in handling the amount of data coming out of a 35mm sensor, not just in recording it but handling it in post. SI and Cineform have demonstrated a workable, relatively affordable way to do it. The total cost of ownership is a known quantity.
RED can in all honesty say their solution is afforable if it's 50% of the current offerings but that still leaves it way more expensive than the SI offering.
To put it another way, if RED can deliver an end to end solution at 4 times the cost of the SI solution then it is revolutionary as that's roughly a linear cost scaling, typically the cost scaling is exponential.
The big question is how many can afford the SI solution?
How many can afford the RED solution at lets say 4x the price?

How many will be offering either for rental, who will supply support / service networks, training for technicians to keep the things running right and spare parts.

This is a big part of the hidden cost structure in any piece of expensive hi tech gear.

[Brian Drysdale]

If you want to shoot a documentary using a RED camera, filming with a much smaller 10 to 1 16/S16 zoom lens makes a lot more sense than the much larger equivalent 35mm zoom lens.

[Glenn Gipson]

Jason, thanks for that explanation, that clears a lot of things up for me. I’m very much looking forward to your camera.

[Don Donatello]

"but their codec is not going to be as good as our Cineform RAW codec which lets you work with the RAW material in real-time in Premiere/After Effects, Final Cut, or any QT capable application. Also you can't record RAW mode on-camera). "

unless you know something we don't this is really UNKNOWN as the RED codec/workflow is not released ... as of today and until at least till Q3 the RAW cineform material will NOT play real time (nor be put in time line) in FCP or any QT application ...

The SI camera & RED IMO will co exist together .. just like the 4 hand size HDV/HD camera's.. they all offer something slightly different ...

some will buy/use both SI & RED camera's ... IMO both will be winners

[Obin Olson]

Problem is, RED has a 35mm sized sensor. Silicon Imaging has 2/3rd inch cmos

That feature alone is worth the non raw images and the firmware based machine to me. But I am a flat deep DOF hater too so...;) If Red uses edge sharpening though...I will RUN as far and as fast as I can away from that camera!


Who knows, time will tell. In the mean time I am happy as heck with my little hvx200+redrock+rails+followfocus+Cavision matte box. Our production company can now make "real" work without having to rent every time we need a greenscreen shot or a plate or footage that looks "pro" this cam ROCKS for the price IMHO

[Jason Rodriguez]

Originally Posted by Don Donatello

"but their codec is not going to be as good as our Cineform RAW codec which lets you work with the RAW material in real-time in Premiere/After Effects, Final Cut, or any QT capable application. Also you can't record RAW mode on-camera). "

unless you know something we don't this is really UNKNOWN as the RED codec/workflow is not released

The major point you're missing which has already been revealed is that REDCODE is not compressing RAW data like Cineform RAW . . . the RED team has mentioned nothing about compressing RAW data with REDCODE.

4:4:4 and 4:2:2 are NOT RAW color-spaces . . . in order to get a 4:4:4 or 4:2:2 image, you have to demosaic.

Once you're out of the RAW colorspace (and erase the per-pixel bayer information), you loose the ability to do stuff like re-do the color matrix, re-do white-balance, re-do the demosaicing algorithm, go back to photometricallylinear, re-do the gamma curves, etc. RAW is the latent electronic image from the sensor . . . once you "develop" it, and compress it, that information is forever gone.

Cineform RAW allows us to preserve all that information and dynamic range, and deliver it for manipulation in post with multiple real-time streams. I think that's a key benefit over REDCODE. It may do a great job compressing to 4:4:4 or 4:2:2 10-bit data, but it's not going to be RAW data which gets you so much more mileage and is passing around a TRUE digital negative-the advantages of having the actual digital negative vs. a "developed" image for DI are enormous.

[Jason Rodriguez]

Originally Posted by Obin Olson

Problem is, RED has a 35mm sized sensor. Silicon Imaging has 2/3rd inch cmos

That feature alone is worth the non raw images and the firmware based machine to me. But I am a flat deep DOF hater too so...;)

Hi Obin,

If you read our FAQ, you will see that you can get the same DOF on a 2/3" sensor to a f4-5.6 split 35mm sensor if you use Zeiss primes at f1.2 (wide-open, which is T1.3).

While F4-5.6 split is not f2.8 or less, it's still fairly shallow DOF. While that is a concern, I don't think it's as big a concern as one would think.

If you go over to our footage on the website, you'll see that we're not having a "deep DOF" issue with our sensor. A number of the shots are quite shallow acutally.

[Jason Rodriguez]

BTW, Everyone should again keep in mind that to utilize the 4K images from the RED camera, you're going to have to come up with a recording solution of your own . . . you cannot do that on-camera (only 2K, which we will be doing as well).

So if you want to make it a purely resolution game, again, we're talking a real-time linear demosaiced 2x oversampled 2K RGB/YUV image from the RED, vs. our 2K RAW bayer image with post-based non-linear adaptive demosaicing (and the abilty to add your own demosaicing engine via a simple DLL interface since we're just recording RAW).

[Tim Holtermann]

Jason, makes excellent points. Remember many films have been shot in Super 16mm and Shallow DOF being less than 35mm has never really been an issue. The SI seems to bring plenty of control for DOF.

Side note: We were ready to shoot our indy feature on the HD100, but I may have to look more into the SI solution. I wish I would have stopped by at NAB.

[Graeme Nattress]

Interesting thread, but I'm not here to get into a my camera is better than yours - that's just silly. However, Jason's comments are based upon his supposition of how the RED camera will work, not facts.

Graeme