Erkki Halkka

Yep. Film and Digital Video are different beasts. It's going to take a few dozen Siths and Sin Citys for people to start appreciating digital cinema as it its - a new format. Right now the trend is to make the Video footage look like film, as closely as possible. This is natural, as more or less all high profile work has been done in film so far. We've been trained to watch film for the last century. So, we do all sorts of stuff to make video look like film - reduce the frame rate, add grain and scratches, adjust the color curve etc. In practice, we try to mimic film's articfacts. If it was the other way around (All high profile work had been done on video, and all reality shows were shot on film for the last 100 years), we'd probably be discussing about the poor motion reproduction of 24fps film, the ugly graininess of it, dust and scratches etc. And people would shoot 60 fps (video cadence), very slow speed (no film grain) small format film (Bigger DOF) to get it to look like video... and the video dudes would say "it'll never look as good" ;-) *** Anyway, one thing worth taking into consideration is the quality of prints we actually see at theaters. Here in Finland, we probably get at least one generation poorer copies of the blockbusters than you do in the US, because of adding subtitles. So, when i said that to me i.e. Harry Potter's technical quality (sharpness etc.) didn't look that much higher than my HDV to 35mm print's, it may very well be because of that generational loss - i'm pretty sure my spot was first generation positive from the printout. When it comes to films i've seen in theaters, "Revenge of the Sith" as a digital projection was totally in it's own class, when it came to apparent technical quality. Very sharp, very low noise etc. So, one could argue it was BETTER than film. It'll be interesting to see the future developement of digital cinema - there are already cameras that shoot at 4K resolution, and really high def, really high dynamic range stuff is waiting around the next corner... *** BTW, we also did a "Guess if this is film or video or 3D" thingy at a company used to work for - http://www.matheus.fi/Quiz/QuizNav01.html The examples are in PAL D1 resolution. All the film examples in there are from 35mm scans, all the video examples were shot by me on a DVCPro 50, 50i / de-interlaced in post. All 3D is LightWave - by me too. It'd be interesting to hear if anyone got them all correctly at the first go. ;-)

HDCAM 1080i is actually really close to HDV as a format - same bit depth, same resolution, just slightly better color sampling. Data rate is much higher than with HDV, but the real life difference shouldn't be big, because of the differeces in compression algorithms. Reading the above experience with HDCAM and HDV, downconverted to SD PAL, it sure sounds to me like the problem lies in how the footage was treated in post. Personally, i've found that to get really good results, one should digitize the raw .m2t HDV stream to the computer, and make an uncompressed intermediate directly from it. Compressing it to another format is a bad idea, especially if you're going to do some complex post production work to it. Cineform intermediate is rather OK, if you're not going to do much FX work. I work in a PAL country, so the frame rate i'm aiming for is 25P. When 50i HDV footage is de-interpolated correctly, then downconverted to SD, the result should be perfect - because the SD information is combined from 3,75 pixels, one gets almost 4:4:4 color sampling too. When using green screen, this footage can be keyed without any serious problems even at HD resolution. Even blue isn't a problem, if the end result is SD resolution. I've had no big problems with color correction either. Slomo is not a problem at SD resolution - you can slow down 50i HDV by 50%, and get more or less full resolution slomo 25P. This only works at 50% though. When the final master is SD resolution, HDV footage should look more or less perfect, wher following this route, and very good at 720P resolution. Even at full 1080 rez, the result should be acceptable. Here's a pop video we did using the above principles, we mastered at 720 25P, and made DigiBeta PAL copies from that master. The result was as sharp as you'll ever get at PAL resolution. If some of you happened to be at HDFEST screening in LA some time back, you may have seen this there (uprezzed to 1080 / HDCAM for the festival). http://www.poetsofthefall.com/ (Lift video)

Well, that was civilized ;-) I guess everyone's entitled to an opinion.

Uh, yes.

No i didn't. 3 am and writing technical stuff don't go together ;-) I used HVR-Z1, which is switchable between 50i and 60i, at 50i settings. This is essentially the same as using FX1E though. Only real difference is that HVR-Z1 has black stretch, which helps keeping more of the information when shooting - i always try to set the camera as low contrast / flat as i can.

Yep, that's what i used. 50i. De-interlaced to give me 25 progressive frames / sec. On theaters, the film is played back at 24 fps, so everything gets slown down by 4%. Here's (roughly) the difference between doing De-interpolation (film look) in camera, and doing it in post: FX1's cineframe option doubles the fields by interpolating them (simulated at the top), where motion-aware de-interlacing interpolates them only where there's motion in the frame (center) - this example uses After Effect's built in de-interpolator at best settings, which is usually good enough. You can see some reduction in resolution at the top and the bottom of the stationary ball, but otherwise most of the detail is preserved very well. In the bottom, the original interpolated (field rendered) 3D image for comparision. In real life footage, the difference to original is usually even smaller than here, because the original video imagery isn't razor sharp like my 3D ballz. Edit: Compression techniques really do make a world of difference. I haven't got slightest idea how "open" the HDV specification is, will there be improvements in the future etc. Also, i only have enough experience from one camera, in one mode (50i), so... dunno. Interesting topic, should try to dig some info out.

I'm VERY aware of these limitations. If you took a look at that pop vid i linked to, you probably noticed that it has both chroma keying and strong color correction. All 200 shots of it. I'm not saying it was a walk in the park to do it... and the result wasn't perfect at HD resolution. This was not a surprise, knowing the limitations. But i'd still say it it was *passable*. Did i suggest it was easy, or that HDV is *just as good* as film?? I don't think so... The format has limitations, but i think they are a smaller issue than you think - they can be mostly overcome. I was talking about it probably being passable as 35mm in SOME scenes. You know, scenes that do not need that much post and don't have that much DOF. Cut within the film, not in a controlled A / B test. I'm aware that what you can do with HDV in post colorwise is more limited than what you can do with a good film scan. That's a non-issue. What i'm trying to say, is that if you take these limitations into account, and know what you are doing, you CAN get very good stuff out of it. Maybe not perfect, but passable. To get good stuff outta it, you need to light, shoot and post knowing your medium. People who are used on working in film may not always be the best choice when working on a video format. It's a different beast. BTW, do you think that in some scenes, HDCAM could be mistaken for film?? I haven't worked that much with HDCAM footage lately, but the shots i HAVE done post on, didn't really look much different from HDV. The ones i worked on last time were actually slightly more video-like than my own HDV shots i compared them with, but that was mainly because the DOP had used too much sharpening in the camera. Anyway, HDV is a bit lower in technical quality compared to HDCAM, but not THAT much. Resolution is the same. Bit depth is the same. Color spaces are HDCAM's 17:6:6 (which is roughly 4:1,3:1,3) VS HDV's 4:2:0. HDCAM is only slightly better in this regard, HDCAM saves about 1/3, and HDV 1/4 of the color information. Even though HDV's compressed much more, the compression algorithm is better/more efficient. HDCAM probably has the edge here though - it should be roughly the same difference as between MiniDV and DVD. The HDV results should be visually very close to what you get from HDCAM - the technical difference is much smaller than i.e. between MiniDV and DigiBeta - and much, much smaller than the difference between a good DOP and a bad DOP... Well, one could always shoot in the fake progressive mode, if afraid of the motion artifacts. It still has double the resolution of MiniDv, possibly more, depending on how it's done in the cam. But, as i mentioned, with good algorithms, one can get really good results when de-interlacing. If you wish, i can post examples or something. That's why i said "not exactly there, but closer".

Okay, here's an example: http://eki.pp.fi/temp/HDV_VS_2K_Film.png I know this isn't the best example as far as shots go, but the best i could do with footage i happened to have on my machine. The 2K scan is a reference plate from a commercial shot in 35mm (i didn't shoot it, only directed), and the HDV shot is from a music video (this on i did shoot, but didn't direct): http://poetsofthefall.com/ (The "Lift" video) I cropped the footage so that the faces aren't seen, as i don't want to post images with recognizable talent without asking permission first - sorry ,-) Edit: Look closely at the flower and red stripes on the tie on the guy on the right, as well as the lips of the woman. You can see the HDV color sampling errors in the original HDV, but also how they've been reduced to acceptable level in the processed image

Well... i'll go through the main reasons why you don't think it would work, and why i think it might: 1. Compression... Yep, HDV is heavily compressed, but if there are no visible compression artifacts, does compression matter?? It's quite rare to actually get the HDV compression to show visible artifacting when watching the actual moving footage. And even then, you'd usually have to really be paying attention to the technical quality, instead of "just watching a movie" to notice it. 2. Limited chroma sampling... the same thing. The chroma sampling artifacts rear their ugly head very badly in some cases (i.e. car's red taillights on a night scene), but on most real life footage, it's hard to notice the difference, especially after proper processing. 3. Interlaced footage... with proper motion detective processing, one can get rather good results when creating progressive footage from interlaced. The vertical resolution does get reduced, but only in the moving parts of each frame. In practice, motion blur hides most of the artifacts. 4. DOF... this depends on the shot. 35mm has much more narrow DOF than HDV at a given F-stop / "apparent" lens length (the same framing). But if you watch the footage in actual films, not all shots have shallow a DOF. First of all, photographers tend to shoot at stops around 5.6 - 8 or so. If you shoot HDV at full open aperture (using ND filters), you get - not exactly there, but closer anyway. Another thing is wide lenses - they have a wide DOF, even on 35mm. I wasn't trying to say all shots would look perfect, or the same, or as good. But I really do think some HDV footage scene could probably have been intercut with 35mm scenes without the audience noticing, if the work was done properly all the way through. A lot of this means treating the footage just right in the post production. I've done quite a bit of real life tests on the subject, and i still say you can get surprisingly close to a 2K film scan with HDV in some cases, when you treat the footage correctly.

Went to see Harry Potter last night, on a big multiplex, and saw the commercial... It still looked good. About in the same ballpark as other "HD quality" commercials there. There were also some other spots that were clearly mastered on standard def - compared to those, the HDV footage looked very crispy and nice. Compared to the actual feature... hard to say how it compared technically, 'cause Potter was done with so much more effort in lighting, color correction etc. Could some HDV footage have been cut within the main film, without anyone noticing the difference?? In my opinion, probably yes. Would the main film have looked as good if shot entirely in HDV?? In my opinion, probably not. Would it have looked good enough for "Joe Average"?? Probably yes. Oh, BTW, Potter was a good ride, very entertaining IMO - liked it.

Well, we did one commercial, shot on FX1, that will run in theaters on 35mm print. So far, i've only seen it on the test theater in the film lab. I need to go to movies to see it on a REALLY big screen... good excuse ;-) The result was very good, in my opinion - especially considering it was a greenscreen shoot. We asked the film lab dude if he could tell what it was shot on, and he said it "looks like 35mm, but i guess you wouldn't have asked if it was". The spot is really simple (just a hand holding a card), but it should be enough to tell how things would look in bigger projects too. We did some post production sharpening etc. for the footage. There were some graphic (text) elements in the commercial, and in the digital, HDTV resolution master they looked much sharper than the FX1 footage. The FX1 footage looked just as sharp as graphics did on the 35mm print though - the film transfer softens things so much i couldn't see the difference anymore. Unfortunately, you won't be seeing this one, unless you're planning a trip to Finland ;-) Edit: fixed a typo