This is as much for my own record of my process so I can refer back to it from time to time. This includes the steps that I use to capture, edit and post-process my videos. This is somewhat a tortuous process because things just don’t work perfectly well – especically between PC and Mac.
Filming. Filming is done with a Sony DCR-H30 – mini DV camcorder. I bought it because it was small (fits in a fanny pack). Originally I bought this because it did MPEG-1 to a stick. I briefly toyed with the idea of switching to MP-1 as my capture media but it just was lousy for editing. Pinnacle, Premier, iMovie, all choke on MP1 after while. So back to DV.
Initial upload of clips is to my Mac PowerBook Laptop into FinalCut HD. This works pretty well. All editing is done in FCHD. I make open and close sequences in LiveType – I like the moving backgrounds. I am getting used to FCHD but I still wish for Adobe style transitions which are shown “between” the tracks. You will notice thay my style is cuts only on FCHD.
I render everything within FCHD and then output back to DV Tape using th e Play-To-Tape option for transfer to PC. I found that trying to move MOV or AVI between the Mac and PC seemed to lose quality – I doin’t know why – so the simple solution is to use the DV Tape as interchange format – it also gives me a nice final copy for archive. Most clips are < 4 minutes so this does not use much DV tape.
On the PC, I pull the video in using Pinnacle DV – this is a great great product. I trim off the excess and then render to a DV AVI (I do not use the Pinnacle’s built in WMV encoder because I don’t like its settings).
On the PC I use Windows Media Encoder to go from the AVI/DV output from Pinnacle to WMV – I do 250Kbps VBR video and CD-Quality VBR Audio resulting in about 323kbps 320×240 WMV files. Make sure to de-interlace. This is a nice two-pass encode that gives pretty good bang for the buck looks-wise.
On the web, I use the 323kbps WMV as the default format. This plays well on PC’s and Macs (even on slower CPUs). This means that most desktops can play the video with a single click (i.e. the right software is already there). This does sometimes hang on Macs – the solution is to download the whole file and play from the desktop. Also I now provide a high quality QT (below).
While WMV is best for the one-click approach, my instinct is that for online good quality archival, H.264 is better. If a few yaers from now, I lose the original media and need to do a derivative work, I would rather have a QT file than an AVI or WMV.
Back on the Mac, I export the FCHD project via QuickTime Movie (not QuicktimeConversion). Somehow the Quicktime Player cannot read Quicktime Movie export from FCHD but JES can. Then I use JES DeInterlacer http://www.xs4all.nl/~jeschot/home.html to deinterlace to yuv2 MOV file. This is a large file a quick conversion – and I will recompress anyways. Then I open this large raw file and use Quicktime Pro and re-encode the de-interlaced DV into H.264 using the LAN/Intranet setting. This is my “high quality” copy. The ideal would be to compress from H.264 directly from the timeline but there is no way to select de-interlace in the H.264 settings. Too bad it is not just a filter applicable to any compression operation. The QT API supports deinterlace. Deinterlacing NTSC is so critical when encoding NTSC materials. On my Mac systems the process is over 2 hours to encode 4 minutes of video. But it looks pretty when done.
Then the whole thing goes up to the web.
My great frustrations in this is how (still) ineffective the transfer is from PC to Mac. it seems as though both camps do all they can to discriminate against the other. For example, MOV playback on the PC is lousy and uses a lot of CPU. Another example is that encoding from a MOV/AVI using Windows Media Encoder results is worse quality than a caputre of the *same* media from a DV tape. The fact that Windows Media Player hangs on the Mac when playing from the web is just idiotic.
Some images
This is the video interlaced – it has better colors, crisper lines, but if you look at Steven’s glasses you see the hashing because he is moving. The hashing is clear in this image but in general it leads to really ugly artifacts in vertical lines as they move from left to right.
This is the same video deinterlaced. The colors are muddied (look at the detail in Steve’s Shirt and the angled lines in the Sakai web page in the background). But the hashing is gone from the glasses. When it plays back, it is much smoother and while it all looks somewhat muddier, you are not distracted by the artifacts.
Sadly this is a no win situation. The compromise is to deinterlace. The problem is that if you *don’t* deinterlace, then the atrifacts are baked into the comressed video. In DV – the structure knows about both fields the interlacing so it can be removed – but in H.264 the jaggies are “part of the video” and not trivially filterable out. So if sometime in the future we want to use the video, it is better to have it a little muddy and witrhout nasty artifacts. The cleaner the video (this image is a little dark) the less the muddiness will appear because the colors will be more consistent from frame to frame. Also, not using handheld will keep the field to field differences lower.
Of course the real solution is to store the DV, and then the real answer is just buy an HD cam – but not today :)