sunyata

Ebrahim, for over a decade now 10bit log has been the baseline for digital to film for theatrical projection. It's only gone higher internally since. A Xenon projector is a higher quality gamut than what you see on your monitor, and consequently less forgiving. Unless you have a 10bit monitor to finish on, you might not realize the artifacts and banding until you're in a screening room. You also have to assume that the color will be different when projected, so you need to provide enough overhead for a colorist to make final changes. Today ACEs format is being pretty broadly adopted for finishing which might be overkill, but that's where we're headed. If you want to know the lowest common denominator for theatrical projection though, I would say 10bit log. Not sLog in linear space, but logarithmic gamma encoded. p.s. I'm not saying anything about 8bit is "useless", I'm just addressing your comment above regarding "high-end theatrical releases".

That is a photo from the Apertus / axiom camera website, they are working on an open source 4k camera and it's using the same CMOSIS sensor as the BMPC, as best I can tell. It looks the same and has identical stats and FPN noise issues. They were explaining with the photo how they were subtracting the noise in the FPGA though, I want to try doing it in a compositor.. but I don't have the footage. The main project is about to start crowd funding: https://www.apertus.org/ Where they talk about the FPN issues: https://www.apertus.org/merry-christmas-2013

Yea, that's awesome, only drawback is that it's rec709 in Video mode. It would be ideal to somehow negate the noise using Film mode too, I'm wondering if the pattern noise is static enough to take a test shot and subtract it in post. In Nuke I think you would just do a subtract and then do an expand on the noise to alter the amount to subtract. It might not be static enough to work though. If you wanted to post the footage to do this with I'd be happy to test it in flame or nuke, but you could use after effects too. Maybe Resolve's procedural cc window also. This is what supposedly the Axiom guys are doing with the same image sensor. They are trying to subtract the noise, but at the FPGA level. Your footage looks really good btw. I'd be interested to see a comparison with the GH4 and BMPC in Video mode since they would both be shooting rec709.

Just curious, has anyone tried recording the FPN against neutral grey and then using the footage in post to subtract out the noise?

Yea, I agree, the A7s is certainly not terrible. I don't think the DR your sensor can capture is worthless if it's crammed into 8bits either, but it's not nearly as workable in post.

If you pause the test I posted when it gets to 4:2:2 and 4:2:0 you can compare.. It also covers the reformatting of 4k 8bit to 2k 10bit myth (I'm taking it into float space actually but it's got all the overhead you need to test the theory). You won't see a noticeable difference in "workability" of the 4k after it has been reformatted to 2k HD in float space. I think a lot of the confusion is really coming from the camera companies that sometimes give you the pertinent info, and other times they leave it out while hyping things like pixel count. I've noticed they also are guilty of mixing terms, like "log" and "dynamic range" in their specs. i.e. You can't have "log" gamma in integer 8bit space, so calling your gamma profile sLog is a little confusing. Another good one, they seem to have propagated the belief that dynamic range is the "min" and "max" luma values, at least I've heard it described that way on blogs.. which would mean 0 black to 3 (which would be pure white) in a 2bit file is wide dynamic range. I got one more! They also have convinced people that low contrast is the "film look", whatever that means. This is the most confusing one. Movies don't look low contrast when projected (movies like Transformers, LOTR are over saturated and high contrast).. it's a pipeline issue only to preserve film gamma and it really requires 10bit log to equal film print density, everything else is just low contrast, so they should call it that.

It's probably not the colors in Premiere, more likely its the color space selected for the imported footage. You need to match the color space of the footage based on how it was recorded, then make sure it looks correct in Premiere before you export. See what you can do with interpreting footage preferences.. I don't have Premiere so I can't tell you exactly where it is. This is AE: You have to play around if you don't know what the source was to match your footage correctly. p.s. you just need one, there are 2 in the example.

Very cool, thanks. I wouldn't doubt that women see more colors than men, just compare our closets.

I've seen those clips too and I know what you mean.. it's just that I work in an environment where 8bit will get noticed, I tried sliding it in once before on a "D" title because I was slammed on time and drive space (and i thought the uncompressed looked fine).. a production person in QC flagged it by running a 4 or 5 stop curve on my job in a screening room. It's like synonymous with 1) not knowing what you're doing or 2) oops, a mistake. I went with #2. I wish the A7s did 10bit log though, not cine-like integer but actual log. Seems to be the ideal camera for that.

Okay, twist my arm. I'll take it for free!

You guys are making 2 false assumptions on either side of the argument, that either 1) the size of photo site is the primary factor in it's dynamic range, or 2) all sensors are created equally. I'm sure if it was that directly correlated, or that simple, there would be no need to hire engineers to work on novel cmos technology. Everyone has different proprietary sensor technology (and sony has the back illuminated exmor r) .. there are even rumors of multi-layered rgb sensors coming out that could bring back something akin to 3-chip technology, or color film, with it's crosstalk issues and all, removing the need for de-bayering (although this could be baloney), then there's variable pixels size sensors... basically, i'm trying to say, it's the technology, not just the size of the photo site... (please don't make a sex joke) although, it does seem that it's easier for manufacturers to make sensitive photo sites w/o the technical challenges associated with shrinking the individual receptor (for example the gap between photo sites). this is not different than shrinking transistors on a chip.. even though it's theoretically possible, there are manufacturing challenges.

This is from sony's specs pdf: Clean HDMI output: 1080 60i compatible device: 3840 × 2160(30p/24p) / 1920 × 1080(60p/24p) / 1920 × 1080(60i),YCbCr 4:2:2 8bit/ RGB 8bit 1080 50i compatible device: 3840 × 2160(25p) / 1920 × 1080(50p) / 1920 × 1080(50i),YCbCr 4:2:2 8bit/ RGB 8bit and this is from an article on "photopop" about the A7s: http://www.popphoto.com/gear/2014/04/atomos-announces-first-4k-recorder-sony-a7s "Like the Samurai and Ninja Blade, Shogun allows the recording of pristine, 422 10-bit images straight from the camera sensor captured using 4K/HD ProRes, Uncompressed RAW Cinema DNG or Avid DNxHD codecs." which is a little misleading to say the least, since they also don't mention in that same article that the A7s is outputting 4k 8bit 4:2:2. 4:2:2 is better than 4:2:0, but believe it or not, it's not a lot better. I'm curious about the RGB option though. I did this to show bit depth and sampling differences in grading a while back. Pause when you get to the 8bit 4:2:2 4k, in particular when it goes red, to see the artifacts. This also shows a "reformat" in full 32bit float space from 4k 8bit 4:2:0 to 2k float. but bottom line, if you get the shogun, you can then get actual 10bit 4k 4:2:2 from the GH4 and not from the A7s and that makes a huge difference in being able to actually utilize all that faux cine-like gamma in post.

That would be awesome, I need a camera for practical effects, your ungraded footage would be a really good test. I'd be willing to buy it too, with no intention to reproduce or use in any way, other than just testing in different grading tools, if that helps. Thanks.

This footage is looking pretty good to me... maybe it was just the internal settings on the other clips that were off. check it out. maybe a tad over-saturated in some spots, but really the best i've seen so far.

This is some of the best and most useful footage for evaluating the quality of this camera that I've seen so far. Any possibility of a download option? I'd love to test the source ungraded footage, but I certainly understand if you want to keep it private. Basically, if I got your 4k files that were recorded internally, converted them to HD float and could push them in lustre or Resolve w/o seeing heavy artifacts, especially in the shots that seem to have a full range of luma, from the long black shadows in those trees, to the white snow, I'm gonna put this camera back in the queue and you've erased my concerns. Thanks!

JCS -- Can you try to shoot some material for the low end or near blacks? I'm really curious to see transitions in the very dark range. And I asked someone to hold up some color charts but they had already returned the camera. I wanted to figure out a LUT for it. Or hey, more footage of chicks just standing around is cool too, have fun!

btw Andrew, by taking down music that is licensed to BMG or whomever, you aren't putting more royalties into the pockets of the artist, the corporations are just controlling the media outlets that your potential fans can hear you on, and ultimately that gives them more control over the artist. They get to decide who to promote and when to promote them. MIA for example has consistently used her website, and some fake websites like Vicky Leekx, to get her mixtapes out there w/o her labels consent. This brings her closer to her fans, and when her album is released, she makes up for it on her sales. It would be even better if they bought it on bandcamp so she got more like 90% of the profits. Anyway, record sales are not where the profits are now, it's in merch and live shows, so all promotion helps. A viral video could do more to promote a band than an army of promotion weasels. Kanye West wasn't complaining when that girl quit her job and his song "Gone" went viral with it.

That MIDI might sound a little more interesting going through this patch.. UVI's IRCAM Piano samples.. John Cage style "prepared" piano samples, replete with erasers, nails and coins in the strings. http://www.uvi.net/en/composer-tools/ircam-prepared-piano.html

yea, bandcamp is great. mark de clive is releasing his new album soon on bandcamp. i hope the site succeeds.. not to mention that i asked them if they had any merch and the person answered my email by asking for my address and size; a week later a free t-shirt arrived, very nice too. artists can set their own price or let you choose, I've found myself willing to go full price or above for anything I think is worth it. one guy was asking like $1,000 for his LP... musicians.

Renoise seems to have a rich community of users and demos that are really impressive considering it's a pretty inexpensive sequencer.. they aren't MIDI I don't think, but the sounds can be generated procedurally or with analog samples and you can grab the demo for free. http://www.renoise.com/songs For royalty free audio and effects, sounddogs has a clear license offering on most of their stuff.. no restrictions. Can be used for theatrical trailers or games etc, look for the no restrictions license. http://www.sounddogs.com/

How about MIA having to threaten to self release her album directly on her website because Interscope held it back over a year on concerns it wasn't "hardcore" enough. Artists are generally much smarter about this topic than most people realize. It's the accountants that just do not compute how to promote online.

Don't worry about the corporations, they're the ones that are really going to gain from controlling internet usage, the artists won't see a dime. I was in a meeting where the studio marketing people were talking about the decision to sell the final special edition movie of a popular franchise in a boxed set that would include all the movies already released, that was the only way to get the special disk. They knew it wold force the fans of the series to re-purchase all those disks.. this was a good idea to them. I really don't have much sympathy for the MPAA or the RIAA.

Well, if you guys are serious about supporting artists, don't shed a tear for the RIAA and the music industry, support websites like "bandcamp.com" where artists can sell directly to you. They aren't losing money because of Vimeo, that's just free promotion, they're losing money because of offshore file upload websites and the music blogs that link to them. If I can, I like to buy directly from the artists, I try to do that at live events for example (they sell cd's at live shows because they actually see the money). It's not just free downloads that are stealing from musicians, it's the labels and their managers too.

Just read their definition of fair use.. I started using Vimeo recently and I think one of my uploads that uses a known 70's disco track would be okay because the video is just an "educational purposes only" quality test.. the other though is a Video that highlights my favorite shots in Bernardo Bertolucci's "The Conformist".. They're just short edits, but probably not kosher. This might be only going forward though, I think the flag only happens when you upload. Should I take it down?? I'm going to live on the edge and let it stay. http://vimeo.com/93735999

No you aren't. Account delete. I totally agree. And don't get me started on the 8bit 4k 4:2:0 to 10bit 2k 4:4:4 myth. The A7s is priced a little high though for the fact that it's 8bit 4:2:0 with no higher export option. One raw hack and it seems like a winner to me.