Hi,

I've posted some new clips from the SI camera in the footage gallery on the website.

Chris, I'm sorry to use this forum for the announcement if it's not the right place, but I just wanted to give everyone a heads-up on what's happening.

Thanks,

Jason

http://siliconimaging.com/DigitalCinema/gallery_footage.html

The footage is gorgeous. Now if the camera didn't have the ergonomic appeal of a cinderblock... For shooters who don't move from the tripod, or who can afford a really really expensive handheld support unit, this thing looks great, but for handheld work it's just not the thing.

Well, true enough it ain't gonna be a pocket-cam but the current case is only an engineering shell; the SI guys at NAB did specifically say that there would be a somewhat smaller and presumably more ergonomic and stylish housing on the final product.

The optical block is removeable. If you don't mind being thethered you could easily shoot with it away from the body.

I too am excited by the cam, but the current form factor needs work.

Also, after looking at one of the clips from Vegas, Jason, I think I'd like to see a modern 16mm lens on the cam. I know those older primes you had are not really the sharpest 16 glass available these days.

Or put another way, after looking at the new clips, things look softer than I'd expect. Might be the glass.

These may give you some better detail:

The QT is compressed with WinRaR and expands to 2GB (8-bit output only)!!!

Uncompressed Quicktime (8-bit, 970MB)
http://www.siliconimaging.com/DigitalCinema/Gallery/CML_Uncompressed.rar

For single frame 10-bit viewing, grab the individual frames:

RAW File (DNG)
http://www.siliconimaging.com/DigitalCinema/Gallery/Stills/SI-1920.dng

Cineon File (Adobe Camera Raw conversion)
http://www.siliconimaging.com/DigitalCinema/Gallery/Stills/SI-1920.cin

These files are also posted in the usuall GALLERY location:

I wanna be impressed by this camera, but I'm just not. The image quality doesn't look THAT much better than an HD100 to me.

Well, SI/Cineform, RED, and even a couple others who get less press are doing their best to leap-frog the mid-pro market squarely past the offerings of the big guys. I'm not going to worry myself about whether others pronounce entirely subjective and premature judgments on products that are still in development. I'll take the facts as they become available, which also means waiting for objective analyses of shipping cameras before deciding whether to be impressed or not.

Amen Pete, but I'm glad to see great footage from SI so far. I think it looks pretty good and I can't wait to see some more. (Hint, hint)

heath

I have some questions about the example images, im only really looking at the dng as i haven't been able to download the big file yet.

There is some quite bizzare noise in this, small vertical and horizontal streaks which im guessing are an artifact from the codec as they look very digital? If we had the proper codec are there methods to reduce this? Or is this an example of the quality we would expect?

The overall image seems very desaturated, grading the image seems to push this noise above into becoming very visible. I mean to get the colour charts looking, well, colour, pushes the noise in the shadows (esp blue) into dubious quantities.

I guess im asking, in the grand scheme of things how good is that sample dng, is it representative or will there be better samples?

thanks
paul

Those are NOT artifacts from the Cineform RAW compression, those are from the Adobe debayer algorithm itself.

Also there is no color matrix in the DNG file, it's a "true" RAW file, so you have to add saturation in the Adobe Camera RAW converter, or somewhere along the line . . . you are seeing the same results as if you were in Premiere Pro and clicked off the color saturation matrix.

You may want to try different RAW converters if you don't like the results from Adobe. Other examples are Dcraw (which has a couple different algorithms to choose from), XnView, etc.

Jason, thanks. I have rawshooter which is how i deal with my SLR photography and that shows the same noise. Plus the default colours it comes in as look like a magenta version of the viper :) but it's the noise that matters.

Why give out a .dng when the processed results don't match up to what you see internally? Also you have a .cin file generated by adobe camera raw for download too which exhibits the same noise. I simply take this stuff into AE7 make sure everything is 32 bit linear and play with colour correction and exposure to get a feel for the latitude - is there's something wrong with this process?

Because of this noise if you push the saturation it doesn't hold up well at all.

Would it be possible to supply a .cin/.dpx after cineform codec has decompressed it?

Or does the workflow not allow that. And by that i mean if you keep the files in cinform right up until final rendering can you only output an 8 bit file at the end?

Or could you take the cineform files and generate log or 32 bit linear files to then work on without needing the codec?

Because if you can't then it implies that all the grading has to happen using the matrixes within the codec?

thanks in advance
paul

Hi Paul,

Good questions.

Right now my "limitation" is that we've been working on the Premiere Pro Cineform RAW importer, but not the After Effects version. I haven't found a way to export 10-bit DPX from Premiere Pro, only 8-bit still-file formats. So what you're asking can be done once we've got the After Effects support. In other words, with After Effects, you can do the whole 32-bit floating point workflow with the 10-bit Cineform files and then export to a deep-pixel format at render out.

BTW, I replaced the Cineon with a new file, this time using a different algorithm.

Check that one out and you can see how different codecs create different results.

Also, the way you're working with the 32-bit formats is definitely "wrong", because our file format is already gamma corrected . . . you you're placing a gamma corrected file in a 32-bit linear color-space and then trying to set "exposure" on it . . . you can't do that inside After Effects. You can only do that if you have access to the linear RAW data, which is only in the Adobe Camera RAW plug-in. Once it's in After Effects, it's been gamma corrected, and using the 32-bit float features like exposure aren't necessarily going to give you accurate results, especially if you're thinking the footage is photometrically linear when it's not. Only the exposure slider in the Adobe Camera Raw plug-in on import gives you access to the photometrically linear data.

Jason,

thanks, i understand and good news about the AE plug in

Saying the world exposure was a bit of a misnomer, i was just using it in that case to brighten everything up (the button's easier to access than a filter!). Although you can take the dng in AE, reduce the exposure in the raw convertor, bring that in, then use the 32 bit levels (with white and black point clipping unchecked) and create some superbrights. Quick and dirty and not recommended for production! (i don't know how the DNG converter deals with the underexposed image, i believe it's 16bpc so when you increase the exposure back there's still enough info to bring it up without posterising)

The new cineon looks better, the noise is more agreeable that's for sure. Is this going to be the best we'll see or is there more development to do on the RAW convertor?

Paul

There's always room for improvement :)

Re-edit: Whoops, appears that I was replying to posts several hours old, apologies, for the duplication.


I wanna be impressed by this camera, but I'm just not. The image quality doesn't look THAT much better than an HD100 to me.

You have got to be joking ;) If you are a very skillful cinematographer and have very good lighting I suppose it is possible to get them close to similarly well composed shots done on the Altasens.

I've read past posts from other threads. A grave yard test was mentioned last year, so it could be from an early prototype, apart from being told that a less then desirable lens set was being used in certain shots. I have noticed some queer stuff on the grave stones stills (I can't afford to keep downloading footage) but I don't know if that is from the marbling of the grave stone themselves, or Bayer's reaction to it. But if these are early prototype images (that should be marked and dated as such) then they may have not been properly debayered and handled (not to mention problems from compression for the web listing). Maybe they should move the earlier prototype images into a separate area, and show the best.

But I have also noticed that the images seem to be coloured in a less than desirable style, maybe it is to appeal to the film market, but this should have a modest mildly superior to film look. They should not be overstated like HDV, but some very nice colour graded versions of a few images, too show customers what can be done, would not go astray . For instance, why is everybody so Yellow, I know that this was a problem on old NTSC TV, but now we can render it good.

About the HD100, look at the latitude in the Altasens images, how the really bright areas are not overly stressful, and the dark areas have detail, turn the colour, contrast and brightness up or down to get an idea of what can be done with it. Though I do think that more could have been done with the images here. I am not overly impressed with the HD100 images I've seen (colour and tonality) I would have preferred if it was a notch above what I've seen, I prefer the sultry quality of the 1/2inch XDCAM HD (off yeah, I like that).

Being here, and on the 35mm SLR adaptors and the digital cinema projects, has taught me a lot, and changed the way i view image quality, still only half way there though, maybe less.


Showing the benefit:

But this is the problem they will have with selling to video people. They are used to various video looks and feel that they are preferable. But some really well composed example frames rendered in polished pro video and cinema styles would give them a good idea of what can be done. I personally would go through all image issues, blow them up and circle out the problem areas compared to the SI, so they can see the benefits.


Have a good day.

Wayne.

I take some of that back. I've just been viewing the stills enlarged on a black background, and mucked around with brightness etc, and they look a lot better than I thought (though still too much yellow in a few).

The problem is that the stills are surrounded by a bright white background and are small, and the lights were on here, this causes the iris of the eye to close down and makes them look a lot darker etc. I forgot, doh.

The daylight stuff looks really good, though. That, I will say, looks better than an HD100. Maybe the interior stuff just needs better lighting.

When I viewed them on a black background with lights out, everything came out, and when I turned up the brightness on the monitor I could see completely into the shadows, pretty amazing, would love to know what the readings were.

I don't know what they were aiming to do, but it looks like they turned off the fill lighting in the interior to show off how it handled difficult lighting.

That early prototype system's color rendition wasn't as good since we hadn't nailed down the color matricies for that sensor . . . we are not using those shots to show off color, but dynamic range. And I think the dynamic range is pretty impressive for what you're seeing.

This camera sounds interesting. The website did not have any photos of the unit, which I understand is still in the design stage ? Any idea on a very rough cost ?

BTW, We've done a quick update to the front-page of the website with some photos of the current working prototype (which is in South Africa shooting a feature film called "Spoon", so yes, it really does work). We'll be re-working some of the mechanical design for the final production model, but at least this gives you an idea of where we're at.

The final price will be $20K with editing software (Prospect HD and Premiere Pro).

Very impressive for a pre-pre model. Will there be a route to get the files on to FCPro? or will you be limited to Premier Pro fro editing?

Full Quicktime support is scheduled for 3Q/06, so you will be able to edit in FCP too.

Can you disclose what codec that would use - could we finally be getting the excellent Cineform codec on a Mac.

Q3 is when the best guesses are at the moment for FCP 6

Steve,

Yes the CineForm codec is porting to QuickTime and the Mac for FCP integration.

David, that is very, very good news, if you can keep even half the efficiency of the codec in the Mac port then I will be a very happy man.

Now I have no reason to consider going back to PC based editing.

Yes, very good news.

Are you guys going to take advantage of GPU acceleration?

I am thinking of getting a MacBook (Ibook replacement) if it is not too bad. What sort of configuration are we looking at for cineform?

There is 965 chipset coming out with more full hardware GPU features then normally, and support fro DirectX 10. I don't know where ever it will get to Macbook, but I will probably lend up buying a machine with that chipset. Would cineform be able to get much GPU acceleration from that chipset?


Thanks

Wayne.

Intel 965 is excellent. For codec work there is no GPU acceleration needed.

Thanks David.

What is the minimum CPU/Spec then to be 1080p territory?

Is the GPU used at all?

I'm just making sure that whatever I buy it is future proofed.

Are you guys setup to take in h264 1080 and transcode to cineform, and transcode out. Forgive my naivety, as I do not have the experience in Cineform and h264. What's needed for real timed transcode?

Any fast Dual core or dual proc system will edit 1080p very well. GPU is not used for standard NLE opterations (can be useful for your compositor.)

Transoding from 264? Why? If a new format use 264 as an aquistion format yes we can convert it, but today it is a promising distribution only format. 264 is a poor choice for an aquistion format. For real-time transcode, you need enough CPU for the decode and enough left over from the encode (as 264 is a CPU hog, this will likely require a quad core system.)

But no GPU help.

There are 18Mb/s+ H264 cameras coming, much better then the Mpeg2 distribution format at that rate. And Panasonic is talking 50Mb/s H264. So another problem format change as with HDV. Even though pocket cam is first for H264, still useful camera to have.


Thanks for your help

Wayne.

No point in waiting for a future format, 264 is going to be a terrible post production format, so yes we will convert it. The beauty of working with an intermediate format, camera changes are much easier to deal with.

as an acquisition format, h.264 is a huge step up from hdv... what will happen in the near future is that software apps will continue to take advantage of purevideo decoding for both mpeg2 and h.264, so you'll be able to edit it natively a lot easier.

so far, i know that matrox, intervideo, nero, and cyberlink all have various products out that support purevideo h.264 decoding, and there is more to come:

"Latest Nero's 7.2 ShowTime 3 already really supports NVIDIA
h264 hardware acceleration. At least it works on my 7600GT. When HW acceleration is turned on, CPU usage drops approx. by 30% (less, than in DVD case)." http://forum.doom9.org/showthread.php?t=99687&page=11

i don't think that purevideo technology supports the encoding side of the fence, but if you have ever encoded with nero h.264, you know how fast it works already... the potential is here and now.

so the object is to acquire with h.264 at the same bitrate/framerate you will use to distribute it with on hd dvd... you'll use purevideo for real-time editing, and you'll only have to render transitions, titles, etc.

Another Carlsbad guy. H.264 is a fine distribution format, in fact a great one, however the further a codec is optimized for distribution the worse it is optimized for post in the professional markets. We see H.264 in cameras as just other format to fix for post, it will be good for business.

You can always output to H.264 for distribution and keep your original CineForm RAW (digital negative) and be ready for the next better distribution format using SVC.

-----------------------

If you want to see the Silicon Imaging SI-1920 D-Cinema Camera and Cineform RAW Workflow IN OPERATION at NAB2006. Take a look at the comprehensive video produced by Studio Daily:

http://www.studiodaily.com/main/news/headlines/6536.html

No point in waiting for a future format, 264 is going to be a terrible post production format, so yes we will convert it. The beauty of working with an intermediate format, camera changes are much easier to deal with.

The H264 camera were supposed to be announced in March, but didn't make it. I know the most likely release period, but is confidential. Actually I was meaning to email you about something around NAB, probably this week.

The possibility of at least 3 cameras this year from various H264 solutions, and Panasonic professional solution next year.

With the GPU DX10/11 processing enhancements, Clearspeed (If AMD adopts it) and cell etc H264 will probably become child's play in the coming years. Then they can get onto really efficient codecs, probably taking cues from the teleconferencing codecs etc, which is sort of remotely similar to the direction I was going. Hmm, I just had another really excellent idea, about a couple of other things and a new direction, a big easy pay dirt.

..H.264 is a fine distribution format, in fact a great one, however the further a codec is optimized for distribution the worse it is optimized for post in the professional markets. We see H.264 in cameras as just other format to fix for post, it will be good for business.

What do you think I am talking to you about it for ;)

There is no pocket cams with cineform yet, and this is the best format for them so far. As long as we can get this into cineform in realtime or faster, and similarly out again, h264 will serve it's purpose. But I agree with Dan, the format is preferable to Mpeg2, the better compression at similar bit rates should help in extreme movement (and low light) where there are problems in HDV (though 50MB/s could do away with it completely).


Thanks

Wayne.

yes, i wouldn't want to have to do a bunch of color-correction with h.264 footage >ack!< it would certainly put a dampner on the idea of aquisition and distribution both using the same codec and the same bitrate.

but the die was already cast with hdv, because it records at a lower bitrate per pixel than we currently distribute sd dvd's at... since my workflow doesn't usually require color or gamma correction, my h.264 dream camera would allow for recording at selectable bitrates... think about how short your post-production time would be, if you didn't need to re-encode the source footage before putting it onto dvd... shoot, edit, and distribute, all without re-encoding the camera original footage.

if i'm going to build an hd-capable editing pc, i would look at putting in a purevideo card over a standard video card, because of the editing r.o.i.... but money spent on an intermediate codec is not useable for anything other than editing; you'd still have to buy a video card for the pc.

"Adobe Premiere 2.0 integrates GPU accelerated desktop playback which dramitically enhances the workflow for editing HD video footage and Adobe After Effects 7.0 allows artists to use the GPU for both high fidelity preview and final rendering. Features include working in true 3D space and comprehesive GPU acceleration of motion blur, lights, shadows, blending, color correction, blurs and matting." http://www.videoguys.com/fx540.html

liquid edition has been leveraging the processing power of the computer's video card for quite awhile now, it's probably just a matter of time until they support h.264 as well(?).

Dan: do you see any signs of H.264 being tested for use in cameras comparable to the SI-1920, and when do you think we might see direct H.264 editing become commonplace? Seems to me the SI-1920 offers unique high-end capabilities with an established workflow which H.264 can't match yet -- is there a practical way to edit H.264 footage in real time at HD resolution?

By the way, count me as one of those people who likes to color-correct much of my footage. Sounds like that's another area where H.264 isn't up to speed yet as a production format compared to other options, but I admire your enthusiasm for the future potential.

I'm not really suggesting that anybody use it for the high end, just a question about cineform in general. H264 is upto Eng level. Even if there was an 10bit+ version, I don't think that precision was an objective. It is just the possibility of an improvement over mpeg2 at the same data rates, when shooting at the extremes (though double, or tripling, the Mpeg2 data rates I suspect might do better). Any sign of macro blocking at the target resolution I think makes a codec junk suitable for only the most basic consumer. The Sanyo HD1 9mb/s Mpeg4 for instance, at 720p resolution, not so great, as a downscaled transfer to SD resolutions on DVD, not so bad (resolution wise, but camera is cheap and nasty otherwise, they could have easily made it match a JVC HD1 with better low light, but not).

I saw something that looked like a video to H264 encoder hardware in a diagram of man ATI card at toms-hardware, or extreme, I posted something somewhere awhile ago. I had links in my technical thread, but somebody seems to keep deleting relevant on topic things, so it use as a technical links repository seems to be greatly diminished. I now saving copies when I post there.

Anyway, this is not the place to discuss such things in detail. A forum H264 basics forum would be nice.

BTW, H.264 was made to be a very efficient codec at lower bit-rates. That's where it's strength lies. Throwing bits at the problem and trying to make improvements in the picture quality that way is actually quite inefficient . . . same problem with Jpeg2K, and why certain manufacturers are learning the hard-way why J2K makes for a really inefficient editing codec.

Also consider that if you increase the bit-rate of H.264, it's no longer in a delivery specification . . . i.e., you can't have high-bit-rate 50Mb/s H.264 and dump that to a HD-DVD or use it for on-air broadcast . . . you're going to have to re-render and re-sample it to those format's specifications. So in essence, when you throw more bits at the codec, you've now created an intermediary codec out of what was supposet to be a delivery codec-only problem being the codec wasn't designed from the ground-up to be a intermediary format in the first place. So you might as well use an intermediary codec like Cineform that is optimzied for a post-production workflow to get the maximum quality out of your footage from start-to-finish.

Dan: do you see any signs of H.264 being tested for use in cameras comparable to the SI-1920

"Panasonic also announced it will offer an optional professional AVC-Intra (H.264 Compliant) codec for the AJ-HPC2000 in addition to their its supported DVCPRO HD codec... The new AJ-HPC2000 is equipped with a native HD progressive 2/3” 3-CCD system and 14-bit A/D processing."
https://eww.pavc.panasonic.co.jp/pro-av/sales_o/news_info/nab2006releases/nab06_16.html

kevin, do you know of any major camera manufacturers that have considered using cineform as an acquisition format?

as i pointed out in this thread, the problem with intermediate codecs is that they are now competing with pc video cards used in the gaming industry, which is a HUGE business... that means a lot of hardware-based video processing power for dirt cheap.

months ago, wayne and i both predicted on this website that h.264 would be used for acquisition, and this is where i think that the future of hd editing is headed... you have to understand the i.t. industry to see the trends.

putting cineform on this si-1920hdvr camera is a killer application, because it takes cineform out of it's dead-end future as an intermediate codec... i hope that somebody puts cineform on a portable sdi recorder, similar in size to the fs-4... it's the perfect app for the xlh1.

Well I definitely said the h264 was not mean to be precise etc. As an Eng format maybe OK, but that is still a level or more down from Cineform.

Here we go, the H264 future is here, Sony and Panasonic have announced a new upto 18Mb/s standard. Now we need cineform more than ever:

http://www.dvinfo.net/conf/showthread.php?t=67127

Yes, plenty of future business.

;)
.......

:) :) :) :)

months ago, wayne and i both predicted on this website that h.264 would be used for acquisition, and this is where i think that the future of hd editing is headed...

Yes, it's starting to become clearer now. I'll still remain skeptical until we see all the pieces fall into places for capture, editing and delivery, but this announcement puts teeth into what you've been saying. Sorry I gave you such a hard time about it...