I’m currently working on the post workflow for a year-long documentary about following wolf packs. The scientists are shooting on the RED One and various small cameras including the Canon HV20. The decision was made right away to do everything at 24p (actually 23.976p), which means every camera needs to be in this mode, and every camera or source that cannot be must be converted at some point.
The HV20 is a very popular HDV camera that offers a 24p mode that is not really 24p. HDV is an interlaced tape format that shoots 60 fields per second and interlaces them together into 30 frames. To create a 24 fps cadence the camera adds what’s commonly referred to as 3:2 pulldown (but more accurately it’s 2:3 pulldown). This method was invented oh so many years ago to allow movies shot at 24fps to be shown on television which is 60i (30fps interlaced) here in the US. Cameras like the Canon HV20 and Panasonic HVX200 have this mode to create a “film look.” The HVX200 also has a 24pA mode that uses 2:3:3:2 pulldown, but more on that later. If you want to get into the technical details about pulldown, you can read the wiki article about telecine here
Original Shot at 1920x1080
I have worked with pulldown for years editing 16mm film. I used HDV for a few years, but back then the cameras I used didn’t have a 24p setting. The HV20 has been out for quite awhile, so I thought a quick Google search would bring me the answers I needed. Basically I wanted to capture the footage into Final Cut and make it 24p. I found quite a few forum posts that suggested workflows using JES Deinterlacer (among other tools) and a few that used only the HDV codec. First of all, I wanted to keep as much of this within the Final Cut Studio suite as I could and second, I want to get out of the HDV codec as fast as possible. Any workflow I decide on will be based on 1920×1080 ProRes 422. The last thing I want to do is take a highly compressed format like HDV, run it through a bunch of conversions, then recompress it back to HDV. Very noticeable generational loss will creep in. I don’t really mind if the conversions result in big files that take up a bunch of space. Quality is the goal here. I found this knowledge base article at Apple. There are problems with the Apple suggested workflow though. When using Compressor, the default action for any ProRes preset is to maintain the number of pixels of the source footage. HD is 1920×1080 pixels. This is referred to as full-raster. There are actually very few cameras and systems that shoot full-raster because it’s expensive both for the imager, and it takes up a lot of space on the storage medium. HDV is only 1440×1080. DVCPRO HD is even more compressed at 1280×1080. It’s not just the “lower” formats, HDCAM is 1440×1080 also. Only recently has full-raster become an option on cameras such as Sony’s F23 and Panasonic’s P2-based HPX3700 to name a few. So if you apply a ProRes conversion to HDV footage in Compressor, you will end up with a ProRes file that is only 1440×1080 instead of 1920×1080.
Squeezed to 1440x1080
What’s the big deal? The issues actually start to crop up later down the line. If you have a Final Cut sequence that is 1920×1080 and you drop a 1440×1080 clip into it, Final Cut will automatically resize the clip so it shows up as 1920×1080. If you open the viewer, you will see that there is both a zoom and a distort on the clip. Final Cut is terrible to say the least, at doing any sort of resizing. It will never look as good as a third-party application such as Compressor, or a hardware converter such as a Teranex. If you apply any text on top of this clip, then move that text to a shot that is true 1920×1080, the text will look different because its no longer on top of a clip thats 1440×1080. It will become distorted. When you mix formats, its even more critical that you keep them as similar as possible (or ideally, convert everything to the same format). If you’re only working with footage that’s 1440×1080, then this isn’t as big a deal. But since 95% of the projects I work on are mixed format, this is a big deal for me.
As you get further down in your edit, you may start working with people that use After Effects or you may want to color correct using Color. When the After Effects artist opens your shot on their system, it will only be 1440×1080 – it will look horizontally squished because After Effects doesnt know that it’s supposed to be shown as 1920×1080. If you’re working with Color, it can either render out the original resolution, or it can convert to 1920×1080 – if you have it convert to 1920, then it will look as bad as if Final Cut had resized it itself.
The Solution? I said above that Final Cut will never be able to resize as well as Compressor, but didn’t Compressor make the ProRes file in the first place? True, but this is another limitation in Compressor. As per the Apple method, you’re supposed to select “Reverse Telecine” in the Frame Controls tab. This causes the other options to ghost out, including the Resize Filter. The default sticks to Better when doing the Reverse Telecine. If you then go down to the Geometry tab and change the resolution to 1920×1080, you will get this picture on the left:
Shown actual size - result from Apple's KB article (1440 stretched to 1920 by Final Cut)
This is pretty similar to what a Final Cut render would look like. Yes the file is now 23.976p, but any diagonal lines have pixelization from the stretch that Compressor (or Final Cut if rendered in a Final Cut sequence) is applying to the footage. The solution is actually to run the file though Compressor a second time:
Step 1. Follow Apple’s steps in the knowledge base article to create 23.976fps ProRes files at 1440×1080.
Step 2. In Compressor, create a new Progressive ProRes preset with the following changes: 1. Turn Frame Controls On, change Resize Filter to best. 2. In Geometry, change the frame size to 1920×1080. Drop this preset onto the clip you created in Step 1. Then you get this picture on the right (I should mention that the image is soft because the camera is shaking pretty severely during this shot. Thus the soft image is due to motion blur, but it helps to illustrate the pixelization):
Shown actual size - when run through Compressor a second time
The final issue with 29.97 to 23.976 conversions is that the time code changes. Yes you are going from 30 frames per second to 24, but you’re also changing from Drop Frame time code to Non-Drop. Therefore your new 23.976 clips will have a slightly different time code than the source tapes and this discrepancy will increase as your time code increases. The simple solution is to just treat your new 23.976 shots as the new source shots. If you have to do any reconnections, just go back to those.
The title of this article is “Pulldown Removal for HDV and P2” but where’s the P2? When you shoot with the HVX200 in 1080 mode, you will notice that there’s no option to shoot 1080p24. That’s because 24p is not part of the DVCPRO HD 1080 spec. It will only record interlaced. If you want 1080p24, you have 2 options: 1080i/24p and 1080i/24pA. The 24p setting will create a 2:3 cadence like the HV20 and any film to video transfer as described in the Telecine Wiki article. You can then edit this footage in a 60i sequence with no issues. The 24pA setting will create a 2:3:3:2 cadence that is meant to be removed to create a 24p clip. If played back in a 60i sequence, it will look “studdery”. When you use Log and Transfer in Final Cut, there is a preference for “Remove Advanced Pulldown and Duplicate Frames” which then imports the 1080i/24pA footage as 1080p24. But this won’t work with 1080i/24p footage. What if you set the camera on 1080i/24p but really wanted 24pA so you could remove the pulldown? Use the method described in the paragraphs above.
Edit: This method also works for telecined film. Most producers now receive their film transfers on HDCAM (or better) at 23.98psf. But this method could come in handy on older footage that is either HD at 29.97, or on SD upconversions because all SD NTSC sources must be interlaced. Examples would be masters that are on Beta SP or Digibeta, or “offline” versions on miniDV. The original Cinema Tools workflow was designed for editors to offline edit using miniDV dubs of the film (which are 29.97 interlaced versions of the original 24fps film), and then reconform to the original film later on during the “online” edit.
Workflow recommendation:
HDV capture in Final Cut includes a cool feature called Start/Stop Detect. As you capture through firewire, Final Cut can sense when the recording starts and stops and creates new clips. This is very similar to shooting to clips on cards like the P2 system from Panasonic, or the SxS system from Sony. When your 1 hour tape is done, you will have 50 or so clips sitting in your bin (depending on how many shots you have of course). Half of your logging work is now already done for you.
This is my first How To/Workflow article. Let me know what you think in the comments section below. Thanks -E