Wednesday, July 26, 2006

HDV Post Production

Post production: HD for Indies is a fantastic blog by a fellow in Texas, Mike Curtis, who shares a vast amount of information on how HD can be used for low budget films. But by "low" he means the low millions or low hundreds of thousands. In this blog I want to concentrate on Low as in I'm doing this on the computer I also use to surf the net on and write emails, etc etc. More like "HD for Amateurs and Starving Artists". However I want something more flexible than iMovie and Final Cut Express.


I'll share my set up.


I still on FCP HD (4.5) so native HDV is out. Also, I'm on a G5 iMac (2Ghz, 20" screen, so I'm not THAT close to starving). I captured with Virtual-DVHS which worked correctly, but with no batch control, no preview while capturing, no cueing, and losing all timecode data on capture. I ended up with .m2t transport streams.


I then pulled these files into MPEG Streamclip to transcode to Quicktime. Streamclip now has a batch list which makes things easier, and two hours of footage took about all night to convert to DVCproHD 1080i.


We began editing the project in DVCproHD as it seemed the most suitable intermediate codec for use. We got nice realtime playback (but not effects, as FCP4.5 does not support the iMac's PCI-e graphics card) but the footage was noticeably softer, and there had been quite a colour shift from the original MPEG streams. I found the Apple intermediate Codec didn't play well in my FCP 4.5 (this may again because of FCP expecting an AGP card). Graeme Nattress (of Nattress filters for FCP/Final Touch fame) recommended PhotoJPEG 75%. I've been a fan of PhotoJPEG for a while when I found that for offline quality work it offered the best balance of quality, file size and playback. Even on fairly low power kit (and this was back in 2001 when I was editing on a G3 ibook) you could get good looking playback on offline files. So it's been nice to find out that it scales up well to being a suitable offline intermediate for HD. While there was a slight shift from the MPEG to PhotoJPEG, it was in a more pleasing direction (colours were more saturated, contrast was a little stronger) and anyway I assumed I'd be going back to the MPEG files anyway.


Editing with PhotoJPEG was pretty good. Effects rendered fast, and we did the titles in JPEG too so they rendered REAL fast. However I found JPEG's Achilles heel, it doesn't work well for multi generation work. Some basic reframes (to cut out mics in the shot, visible lights, etc) and rig removal for some reflections and cables in shot threw up some fairly obvious shifts in tones and a much courser, grainier image quality. to combat this, I re-did the resizes in FCP rather than after effects and the wire removal I re-rendered, this time with effects in 8 bit uncompressed all the way through. It was much better.


Going back to the MPEG streams once the edit was done was a pain, as FCP wanted to find clips with .mov extensions, which meant I had to individually select every similarly named .m2v file (demuxed from the m2t files with MPEG streamclip).


They loaded slowly but accurately, however any attempt to render anything or export at all just crashed the program. I REALLY need to upgrade FCP, but that's not possible right now. So I decided to do everything at SD at this point, and I'll just have to keep the M2T files online until I can upgrade to FCP 6 (might as well skip 5 at this point.) I tried FCP5 on one of the computers at work, which was a 1.4Ghz G4 eMac, and while playback wasn't flawless, it was pretty good. It seemed to drop parts of the image rather than whole frames, but when the playback window was resized to fit the eMac screen it was barely noticeable. There was even some very basic realtime playback (frame droppy dissolve, one layer title or a bit of simple colour correction). In order to get back to HDV tape I rendered out a Apple Intermediate Codec Quicktime file which I dropped into an iMovie project, which then rendered out a HDV master to go back to tape. All in all it looks pretty good, but close examination reveals that there are have built up quite significant compression artefacts. With FCP5 this wouldn't be such a problem as you would go directly from HDV originals to your output codec of choice, though for me this would mean I would not be able to use FFmpegX to create DVD and H.264 files. FFmpegX seems to have issues with Quicktime movies, it really prefers to have MPEG files in some form. It created very high quality h.264 files at 1280x720 25p but naturally had to retain some of the compression artefacts from the MPEG2-AID-MPEG2 chain, before adding it's own effects. However it looks extremely good considering for a 9 minute movie the final file size is only 190MB.

One thing I'm having difficulty with at the moment is colour correction. SD uses the 601 colour space, HD the 702 colour space. I think there are problems in going from one to the other. When I'm transferring to DVD or DV tape, and playing back on TV, the image is far too bright, which causes sync problems when played back on TV.


More on colour correction later...

Labels: , , , ,

Sunday, July 23, 2006

HDV Short: Production

well, the short film is finished and here are some points. I'll split this into two posts with this one concentrating on the actual shooting.

The shoot went quite well as both actors were very professional and worked well with the director. While this post will deal with the technical side of the production I should obviously emphasise that without a good story and decent performers, then great technical proficiency is irrelevant. we had a small crew, of four, with Karen and myself swapping between camera and lighting and sound, and a friend Grace acting as continuity supervisor and general PA. Four really this wasn't enough, especially as one crew member at a time was required to baby-sit my daughter (Ayşegül, my wife, was directing, and little Leyla was actually one of the actors, very good she was too)

Lighting. Karen Savage who is co-credited with lighting takes the credit for the success of the lighting in this project. We used 1000w open face lights ("redheads"), which worked well for general overall lighting but were hard to use for lighting close ups or detail. Really even for low budget productions, it seems that some low power Fresnel units would be required. The Z1 is not a particularly light sensitive camera, and in even reasonable light interiors, we found ourselves always on the edge of under exposure. We tended to be around the 1.6 - 2.2 stop area. This is fine as really we wanted an open iris to achieve shallow depth of field.

We tended to bounce the light a lot (direct light with the redheads would have been to harsh in the confined space. This softened the light nicely but made modelling tricky, especially backlighting. One of the advantages of the location was the way from the dining room and living room, and to the kitchen was semi open plan which meant we could get backlights a healthy distance so the backlighting effect was subtle. In my opinion, backlighting is REALLY important in video, it helps alleviate that flat video look, but needs to be subtle.

Gels on the light made a big difference. Even just a little blue made a redhead resemble daylight, while a 1/8 CTO on a key-light would punch up skin tones nicely. Again, it's GOT to be subtle or it looks like bad mixed source lighting.

Focus. Well, this is high definition after all, and what's the point of shooting HD if it's soft? Focus is MUCH more critical in HD. Some sort of monitoring system is essential. We used my ibook, taking a firewire line from the camera, downscaling to DV, and then playing the results live through a programme called BTV pro. It's just about acceptable for judging SD focus but not even close for HD. We have several shots that are just a little bit soft. the Z1 has peaking in the viewfinder( peaking finds sharp edges in the image, and highlights them in the 'finder - as sharp edges must be in focus) and expanded focus (zooms in the middle of the image but only in the 'finder), though CF25 softens the image somewhat so peaking isn't as effective. The expanded focus feature is OK, but still, the little monitor isn't sharp enough for HD focusing. You can try the old zoom in/focus/zoom out method, but then you won't be able to use the cameras rather nifty shot transition feature for pull focusing. What's the point of pull focuses (pull foci?). Also, faces tend not to have sharp edges, so focussing on them isn't as easy. One solution might be a target card of some sort, that could be focused on when setting up shots.

Additionally I have noticed on some of my other footage that there seem to be better areas of the lens. Fully wide open (1.6) or at either the widest or longest ends of the lens, the image loses a bit of sharpness, but the image really pops out as being high definition in the middle range of the lens. Now this is not based on systematic testing, but one looking at rather off the cuff shooting. Maybe some later tests will confirm this, but I need to get my hands on a proper testing chart AND access to a proper HD broadcast monitor.

Highlights: HDV (at least the Z1) does not handle highlights any better than DV. Really, for a truly film like image you need to keep large areas below 100IRE. White walls, windows, skies etc. need to be controlled! (How do you control skies? I'll get to that...). In the past when shooting mixed light I would have recommended gelling lights to match daylight coming through windows, now I could recommend going the other way. To get good results, either invest in plenty of CTO and ND gels, big, 1.5 metre wide sheets of the stuff, on 100 metre rolls! Expose for highlights, and then add fill in order to balance your contrast ratio. Sometimes you might get good results shooting after dark and simulating light coming through windows with a couple of 800w or a 2K light, maybe gelling them with ½ or ¼ CTB to give a subtle mixed light effect. We did that for one scene and it worked well, though you can see if you look in the final shot that we cheated and placed the light indoors.

As for skies, which look rather ugly when blown out to white on video, i think the best bet is either to plan exterior scenes where you can see the sky, or ALWAY shoot with the sun at least 45° behind the camera. I've got some nice blue skies and cloud patterns to expose well when following this, but as soon as I pointed the camera at all towards the sun's direction, even when the sun had dropped below buildings, the sky was an ugly blocked out white. There might be something that can be done with Sky replacement in after effects, but that of course limits your freedom with camera movement, and requires extra time with After Effects or something similar in post.

I would also recommend shooting as much as possible towards the even, or heading as far north (or south if you're in the southern hemisphere) as you can. I remember seeing Richard Attenborough on an BBC show about British film in the 60s waxing lyrical about "that northern quality of light". What that refers to is that at ANY given time of the year, in the north, the sun will be lower in sky (the same is true the further south you go in the southern Hemisphere) and at lower angles you get a better quality of light because the sunlight has to travel through more atmosphere, and is therefore slightly softer. Anyway, long and the short of tit is, try and use the sun as a key light in all exterior scenes, and not as a backlight unless you have a truckload of HMIs for fill/key lighting, or want a specific cameo light effect.

I'm trying to figure out exactly which Picture Profile settings really get the best out of the Z1 - I'm leaning towards sharpness set to 7-8, Black stretch on, Cinematone setting 1, though the added contrast of Cinematone is a little scary (lots of fill required) it does strengthen the blacks, while black stretch also gives you a bit more leeway in the shadow details. We dialled sharpness back to 2 for our film and it really does look soft in general.

One of the areas I think Low budget filmmakers using HDV must think about is using filters (polarisers, grads, NDs, diffusion, etc.) on the cameras to compensate for the formats shortcomings.

Cineframe Initially I was suspicious of this mode, and I still won't recommend it to others wholeheartedly, but I must admit I'm coming around. It does look very film like (especially when transferred to SD) and it does get rid of some of the MPEG compression artefacts. When converting 50i to 25p in post, the de-interlacing seems to leave behind artefacts from both the 4:2:0 colour sampling issues AND the extra detail from the interlace jaggies. Going CF25 seems to avoid this, though it IS noticeably softer. I notice that while CF was poo-pooed initially in many blogs and discussion boards, a few are coming around to it now, particularly David Newman, of Cineform. However I've not used any really highly rated de-interlacers (magic Bullet, Nattress Film effects et al) so I've not really have the chance to see what the best that can be achieve. I've used some free FCP de-interlacers, or the ones in Virtual Dub or MPEG Streamclip, and after all you get what you pay for.

Conclusions With HD, Everything shows up. The odd cables in the background, lights reflected in shiny surfaces, crew reflected in glass, creases and stains in clothes. On Standard Definition, these things could be got away with, but NO MORE. You MUST double check constantly whether crew members' coats, make up kits, cables bits of light stands are in shot or not. you must take much greater care when shooting HDV. It just shows mistakes up much more clearly, even when down-converted to SD. When the image is sharp clean and well exposed, it does sort of pop out at you, and other less precise shots will suffer by comparison. The much ballyhooed compression artefacts aren't really an issue unless you're REALLY going the Saving Private Ryan path, but de-interlacing does throw up the odd hidden things. You really have to pay attention to set design. Us DV filmmakers have been dealing with the deep DOF for a while and to be honest it's not the deal-breaker for me that it is for some. I'll just back away and zoom in, keep the iris open and maybe go for a slightly cleaner set and background than normal. For outdoor shooting think about where you're pointing it in terms of how the background looks. - I can see polarising filters making a big comeback. Should I ever get a Z1 of my own I can see myself wanting to invest in a set of filters (polariser, a variety of NDs, a few grads, some tints) in order to make the image as pretty as possible before it goes to tape.

Next entry I'll deal with Post Production.

Labels: , , , , ,