Video el Cheapo

July 7, 2007 10:08 AM

How can I build a high-def editing system on the cheap?

I want to build a high-def editing system on the cheap (excluding monitors, keyboard etc...) for a round £500.

I'd like to avoid expensive software -- I'm considering using Cinelerra CV as my main package. Is it any good? What (free/cheap) Linux distro would be best?

Now what kind of PC do I need? The source video will be made up of individual JPGs -- 3200x2400, around 22 meg in size a piece -- and is intended to be printed back to film after the edit is complete. I'd like to stay away from any further compression if possible, and especially avoid the awful HDV.

Where do I start? What's important when considering building a PC for these needs? Do I just buy a CPU and motherboard and chock it full of memory, or are there other things to consider?

posted by popcassady to computers & internet (9 comments total)

You can't build an HD-capable editing system for that cheap.

You say the source "video" will be individual jpgs. How are these to be acquired? How do you intend to output the final product? Have you considered not editing at HD resolution? You are aware that the jpgs you are mentioning are higher res than hi-def video (1920x1080) anyway, right?
posted by mzurer at 10:59 AM on July 7, 2007

I'm fully aware that they are high res than 'hi-def'.

I want to know if it's possible to edit at this resolution without having to resort to a lo-res offline edit.
posted by popcassady at 12:05 PM on July 7, 2007

...and even if I have to resort to editing a proxy video and conforming later, I'd still like to know what I should be looking for when building a system for this.
posted by popcassady at 12:25 PM on July 7, 2007

I would agree that it seems difficult to build a full HD editing system at that price, however if you're only editing image sequences then I think it's doable, especially if you're using free software. All you'd need is software capable of working with and outputting an HD-size frame.

It's possible to edit at that resolution, but probably unwise if you're goign to be outputting to any type of standard HD format. I would pick the HD standard, then work on that frame size.

The advantage of working at the smaller frame size is that you would also cut down on your render times. The JPGs you're talking about almost certainly contain much more information than you need for the screen, so you can scale them down before editing and make them more manageable to work with.
posted by bcnarc at 12:35 PM on July 7, 2007

The 2k standards for 1.85 are 1806x976 and 2048 X 1106 (forget that they are not in the same proportion as aforementioned resolution). Let's say I'll be working at the higher resolution.
posted by popcassady at 12:54 PM on July 7, 2007

Buy as much Mac as you can afford for five hundred quid and borrow a copy of Final Cut Pro HD.
posted by jtron at 2:03 PM on July 7, 2007

A used Mac and Quicktime Pro?
posted by rocco at 4:18 PM on July 7, 2007

I want to know if it's possible to edit at this resolution without having to resort to a lo-res offline edit.

In a word, no.

Compression and proxy editing is de rigeur when working with high definition and film formats, especially if you want to do it "on the cheap". You simply can't avoid it, unless you're willing to spend big bucks on high datarate RAIDs and the requisite capture cards and/or Fibre Channel connections. And £500 won't even pay for the software required to edit video with the specs you stated.

Doing the math based on your specs, 22*24fps = 528 MB/sec, or 4.2 Gbps. You'd need a serious RAID to sustain data rates that high. You can theoretically build a cheaper, Port Multiplied eSATA RAID that may get you in the ballpark, but it's still going to cost you a LOT of money.

You need to ask yourself the question, "what is my time worth?". If you're going to be editing this for hobbyist purposes, with no critical deadline or need to work within accepted industry standard workflows, it may be possible to hack together a budget, OSS based solution to your problem. The caveat, of course, is the word "may". Cinelerra is an interesting project, but in no way is it ready for any level of real world production work. You'll save yourself a lot of time and headaches if you go with a more supported off the shelf solution, such as a Mac running Final Cut Pro

What you need to be looking for when attempting to build a system for what you're looking for (which I really don't recommend, btw) are several things:

1. Hard drive space. Assuming that you'll want to load in your original JPG sequences before transcoding to a more usable (e.g., DPX, especially if you're going to film out at the end), more efficiently compressed (e.g., Apple ProRes 422) format, consider that you'll need at least 31 GB of diskspace per minute of footage (that's not including audio). You should then multiply that figure by at least a factor of two in order to get an idea of how much disk space you need.

2. Hard drive bandwidth. If you're intent on trying to work with your original gigantic JPG sequences, then you really dont have much of an option other than the aforementioned Fibrechannel or even eSATA RAID solutions. They aren't cheap.

But if you transocde those sequences to a smaller frame size and a more efficient codec, you can more than adequately edit the footage even on the internal harddrive of most any modern laptop computer. Apple's new ProRes 422 codec, in particular, achieves incredible, visually lossless picture quality at datarates 10 times less than the uncompressed HD source. The disadvantage to ProRes, however, is that you're locked into using Final Cut Pro 6 on Macintosh computers (it's not a cross platform codec, yet).

3. Monitoring: If you're intending this edit to go back out to film, you need to monitor the playback on a properly calibrated monitor, using a lookup table (LUT) that matches the intended output film stock. Otherwise, any color correction or adjustments you make to the picture will not match when it goes to filmout. Monitoring alone is one of the most expensive aspects of a film editing pipeline, and it's quite shocking to realize that the average LCD monitor for film editing/color correction purposes costs at the very least, $10,000.

There are, however, products such as the Matrox MXO, which will allow you to get pretty damn good monitoring, using a cheaper, dedicated Apple Cinema Display.

If you're not worried about preserving color/picture fidelity when it hits the big screen however, you can probably ignore this.

4. Pipeline: This is actually the most important thing of all. All the previous hardware suggestions are meaningless if your editing workflow makes it a pain in the ass to perform an offline to online conform of the footage. My biggest concern is that it sounds like your footage has already been scanned in as JPG sequences, which is bummer in and of itself, because you really should have been supplied Cineon or DPX sequences if your intended working format is film. Those formats preserve the metadata of the original film scans (i.e. timecode, keycode, reel number, etc etc), which makes it a snap to downres them and then later reconform them easily. The JPG sequences most likely have no metadata in them, so there's no easy way to reconform all that data when it comes time for the online, short of doing it manually (making sure that the offline JPGs are named exactly the same as their high-res counterparts, and manually relinking them).

Generally speaking, most all nonlinear editing applications have very poor support for loading and playing back image sequences in the same manner as a self-contained Quicktime movie. This alone can be a show-stopper for you, unless you come up with a plan to encapsulate the sequences as Quicktime movies, using a tool such as GlueTools which can "virtually" convert a DPX/Cineon sequence as a Quicktime movie, complete with metadata.

This is probably too much information, but hopefully it at least nudges your way towards the right direction.
posted by melorama at 6:55 PM on July 7, 2007 [1 favorite]

I know you've marked a 'best' answer. Just to add.

Do you have any idea how expensive a film out it? Each group that does it, does it a bit differently; figure you're going to speend at least $1k/minute.

If you wanted to actually do all of the above, what you'd have to do is create a proxy (a low res copy.) It's a problem due to data size. It's a problem due to input (Yes, I'd like to import 14400 frames (that's 10 minutes @ 24 fps) is going to require 312 gigs (at your sizing) not including the proxy. It's going to cost you more than $1k for drives that can just hold this (forget the cheap USB drives. You'll have to invest in a serious raid).

For $100k I can point out a system than can actually edit film frames. For $1k or so, you can edit @ HD sizes, and if it's compressed with actual store bought hardware (at any store with very little 'custom' configuration.)

But, I did mention that if you intend a film out that it's going to cost about $1K a minute right?
posted by filmgeek at 4:25 AM on July 8, 2007