Indie feature workflows

A fried wrote me an e-mailing asking about HDV and other workflows for doing a low budget indie feature. This is what I wrote in response. I heartily accept corrections and clarifications from my dear and few readers. I've already clarified a few things and add links as appropriate.:

Hey Matt,

Also, I know it's not a full-on professional camera or anything like the other two, but what do you know/think about the HDR-FX1?

I'm going to a seminar next Thursday being run by the Australian distributors, and I'll have some more comments then I'm sure. However, what I saw at a trade expo last year wasn't that mind blowing at all. It really just looked like sharp DV. Which isn't all that much of a surprise. Sony tend to produce cameras that are very 'flat' - they're accurate in terms of colour and gamma. But accurate doesn't mean appealing. Thats why I prefer the Panasonic 'cine' line (DVX100/SDX900/Varicam) because it produces an appealing image. The skin tones, the highlight handling, the overall tonality.

Now, the FX1 is a HDV camera. This is a format which puts a HD sized frame (1080i) into the same data rate as normal DV. So you're compressing the image at a greater rate than DV.. I don't know if you have much experience working with DV in terms of colour correction, grading or compositing... but its a format that breaksup under pressure. My feeling is that HDV will be worse in this regard - and it seems some of the preliminary reports are that this is the case. It just doesn't hold chroma very well. I assume many of the tricks that have worked to 'smooth' DV will work well for HDV, but until I actually get some footage it will be hard to know.

As you're planning to shoot your feature next year, you will also be in a position to use the new 'consumer' HD cameras from Pansasonic. These will be DVCPro cameras rather than HDV, and I think they will offer the option to record in DVCProHD which is the format used by the Varicam. This has double the data rate per frame (50megabits vs 25 megabits) as HDV. Based on my experiences with the Varicam, this still isn't ideal, but its much better than HDV I would imagine. However, I don't know if those cameras shoot progressive...

Which brings me around to progressive vs interlace. Not sure about how much you know about this stuff, but its importantif you plan to go to film out. Film is progressive, standard standard def video is interlace. Going from interlace to progressive looks like shit, cause you get all these artifacts known as 'interlace tearing'. It looks like shit. Or, rather, it just makes bad video-to-film transfers look even worse.

Its less of a problem in the NTSC world because they can use a technique known as pull down to extract a 24P (24 frames per second, progressive) image from a 30i (30 frames per second, interlaced) sequence. In the PAL world, where we are at 25fps, then that process is redundant. Basically, the easiest and best method to get to 24P is to convert all your footage to progressive (using various field blending techniques) then you slow it down by 4%. Now, because the process of field-blending isn't great, its MUCH MUCH MUCH better to shoot in a progressive format in Australia.

But there's a difference between 'true' progressive and 'fake' progressive. The high end HD cameras and the Panasonic cine-line all shoot true progressive (the DVX-100 does it by using 'true' pull down cadences). True progressive is a single image per frame - exactly like film. Fake progressive is a camera which actually records interlaced but does some field blending (or pull down) on the video stream and records that to tape.

Unfortunately, the HDR-FX1 shoots *fake* progressive at 24FP with a very fake simulated jitter of 24P. At the other frame rates, it just does field DOUBLING, (throws away half the vertical resolution, then doubles the remainder), which is actually worse than the XL1s 6-year old approach (which is a blend). However, cause the HDR-FX1 is a high def camera, the field doubling ends up being more or less the same resolution as a progressive-scan standard-def image.

Most of the advice seems to be to shoot in interlaced mode with the FX1 and let the experts do the field-to-frame conversion. They've had plenty of experience doing this with DV until the appearance of progressive DV in late 2003/early 2004.

However, there is also post to consider.

HDV, from the sounds of it (no personal experience yet), is exactly like the early days of DV via firewire. DV via firewire back in 97-99 was a pain in the ass. There were all kinds of weird problems, the cameras and interfaces were moody, and in general it was kind of a black magic to get it working properly. e.g. there used to be a 'static burst' problem in the early firewire drivers. At the end of every clip there would be like 2 seconds of white noise produced by the interface.

The problem HDV is actually HDV pe rse. It records in mpeg-2, the codec used in DVDs. Mpeg-2 uses a technique known as 'GOP' or groups of pictures. Its an inter-frame codec. It examines at sequence of frames (a GoP), analyses them, and determines how to best compress that sequence. Information which doesn't change is more highly compressed that stuff that moves. From an editing perspective, the problem is that you can only cut on the beginning of every GoP. Kinda fucked. There are a number of solutions on the market for this bullshit workflow. Most of them seem to basically transcode from mpeg-2 into another format that you can actually, like, edit properly in. Problem with this is that you're taking a highly compressed codec (HDV) and re-rendering the whole thing into another compressed codec... so you lose a whole generation before you even begin. Its shit. So, the only viable workflow for high quality work imnsho is to take it to a good intermediate codec, like DVCProHD or, if you're totally nuts, to Uncompressed 8bit 4:2:2 HD frames (which is about a 100meg/s). Actually, if I used a converter that keeps the timecode information... I'd do an offline in photo-jpeg or some other similarly low data rate format then online it into uncompressed HD. Still, its a fuckaround.

Where does this leave you?

Well, if you're doing a zero budget feature, the chances of you going to film out are pretty low. Pretty much everyone I know, with a few exceptions, is finishing to SD or HD and saying they have film prints. They then ship the screener to all the festivals and if they're accepted, they scramble to raise the capital via the FTO/FFC/AFC/Bank to get a film print. At over $1000 a minute for a proper 35mm transfer, its something thats very hard to actually afford. You should plan your post path for eventual film out, but don't make all your choices based on it.

The reason that a Varicam and a CineAlta look fantastic for video isn't that they only shoot big HD frames. They also have better colour depth, better CCDs, better gamma-handling... and, most of all, they use big fuck off lenses. A HD lens is worth between say, $30-100K. There is no way in hell a camera which costs $10K including lenses can approach the sharpness, colour rendition, depth of field, speed etc, of a real lens.

Given the teething problems of HDV, I'd be more inclined to go SD DV. Either the Panasonic DVX-100 or the Canon XL2. The DVX-100 It shoots with a gamma appropriate for film out, it shoots proper progressive scan, and its actually pretty smartly designed for a camera operator - mechanical focus and mechanical zoom .There are a number of accessories out for the DVX-100 which make it a good indie feature camera including matteboxes and follow focus units.

The XL2 is also worth considering. Its got a very sophisticated engineering side (you can tweak the chips) and the glass you can get for it is better than the DVX-100 if you get the manual lens. You can also use the P&S adapator and use 35mm lenses on it, although you lose about 2 stops so you gotta have lighting kit. It is shoulder mount, which is helpful, but if its balanced like the original XL1 then its really not properly balanced at all.

If you can afford it, the SDX-900 is a fucking amazing camera. Beautiful pictures, well designed, easy post flow (firewire via the 1200A deck). A few indie features have been shot on it. You can use it as a single operator. I did an entire video clip with it in a no light situation and it came out nicely.

There is also the 'reel stream' approach to DV. It extracts uncompressed 720P (interploated) raw images from the CCD blocks ot the DVX100 and the XL1. Its workflow is unproven and it'll be cumbersome for a one man band, but the potential is pretty fucking hot if you're a post nazi like me. I considered importing them to Australia, but I'm lazy.

There are a lot of articles on HDV appearing now, so you can continue following its progress. In a year, everything may be sorted and you'll have a great solution.

Stu.