Scaling 1080 DSLR footage down to 720, but screening in 1080.
I've seen this topic addressed on other threads but havent found an answer to my question...
I'm editing a largely talking-heads documentary. We shot most of the interviews on a DSLR in 1080 24fps. I've been editing in a 720p ProResLT sequence, so that I can take advantage of the "extra" resolution by zooming in to 100% at times. (Also, the DSLR shots aren't the sharpest, so this way the softness is not as obvious.) I plan to master to 720p ProResLT. (There are some photos and graphics, but the B Roll is all from youtube so it is already low quality and doesnt gain anything by being in 1080 vs 720)
The issue however, is that when creating a DCP for theatrical screenings, this exported 720 master is going to be scaled back up to 1080. (Usually the film will be shown in 720 on Blueray or on the web, but for some theaters that require a DCP, it will be scaled up to 1080)
So my question is, which is preferable with 1080 footage when I
a) want to zoom in by 130% for closeups pretty often and
b) when in some cases it will need to be delivered/projected at 1080:
1) Edit in a 720 sequence so that the zoomed-in shots will not be scaled larger than 100%, export a 720p master, but then for the DCP scale that master back up to 1080. So in the DCP, shots that were originally 1080 at full quality, would now be 1080 up-rezed from 720 which was downrezed from the original 1080 footage.
2) Edit in a 1080 sequence, the zoomed in closeup shots will be at 130% and thus not full quality, export the file for the DCP at 1080 from this sequence - and at least the non-zoomed-in shots would be at full quality, ie 100% and not larger. And then for the 720p Master (if there is a point in having one) I would either nest that 1080 sequence into a 720 timeline, or I would import the exported 1080 master file into a 720p timeline. But in the 720 export, the shots that I scaled up to 130% in the 1080 timeline will now have been scaled back down to 100%. I'm not sure if quality is lost going up and down like that, in either option 1 or 2.
Any advice is much appreciated
The results of such tests are subjective - what's important is which way do YOU think looks better? Would be very easy to run a few test exports and compare them, using the intended final codec settings.
Simple test. Put 1080 clip in 1080 sequence, zoom in, export as 1080. Put same 1080 clip in 720 sequence, zoom in, export as 1080. Import both clips into 1080 sequence. Put one clip above the other, so they are synced. Apply a CROP to upper clip, to create a split-screen effect. What can be telling is to adjust the crop slider back and forth while viewing the Program monitor as big as possible, provides a very direct comparison. Set Program to actual pixels so there is no scaling applied. Does the image get fuzzier, noisier switching between the two clips? Or go full screen and toggle the upper layer on and off repeatedly. Decide which version looks better.
Safe Harbor Computers
I would start with a transcode to a lossless codec and a proxy in AME (using same file, with multiple simultaneous jobs). Using this, I start with 4 outputs to test.
the first one is a 1080 at LT prores (slight compression, but excellent quality and small files). Second is a proxy at 1080. Third is an LT prores at 720 and a proxy.
Start both in the project (both LT prores) and substitute the proxies. Take your 720 sequence and blow it up by dropping in a 1080 and upscaling it...
Then adjust your 1080. Mark up both and export the zoom areas. Notice anything?
Are you zooming\panning in Premiere or in camera?
If you do this in premiere, you will notice quality drops either way. Since you've already scaled the video to 720 (you will never see the original quality again), you are not blowing up to original quality. However, you shouldn't see too much drop if you start with a lossless file format, as all of the quality is retained. Export to a lossless, and it will be faster on the export. Then compress that (which will be faster and better quality compression with smaller files). I fire off 1080p at 60fps files 2 to 3 times a week on a 2008 macbook pro with 4gb ram and 256mb graphics, and 5400rpm hdd's. I've moved to Compressor and I fire off more than that now, by sending the batches to other machines when I rent some studio for use in the project.
If you work on compressed file formats straight from cameras, they are transport streams with buffering. When outputting, you have to wait for that buffering both when reading the original, and when compressing, and in between you have to wait for it to rebuild the frames from the original so it can recalculate and compress a new set. If you use a PROXY version of the file format, it will be a middle-area between the two extremes, allowing for less choppy playback in edits.
Typically, when zooming in, I don't just use the zoom keying in premiere. I'll use After Effects to actually upscale the entire clip to a larger resolution, clean it up (Photoshop style cleanup) and denoise it, then place the comp into my sequence. Then I start the clip with a key ZOOMED OUT, instead of IN, and let the zoom pass inward from there. Better quality.
Depending on how you shoot, you can limit the zooming you have to do in POST. For me, I use cameras and remotes, but I use HDMI wireless transmitters like the MYWIRELESSTV2 to send the remote infrareds. I only need one transmitter for each camera, and I can set up a viewing post where I can view the input, zoom, pan etc. It's a great way to set up a small, low cost studio environment. Just make sure you have the proper automatic PANHEAD for your camera, and that it responds to the remote. You can tie the infrared sensor to a small dowel just far enough in front of the cam so it can react. I've done this several times. It works great, and takes the zooming out of post. Plan the zooms, and start in while you are on another camera. Return to normal afterward or do a slow zoom out. I also place two cameras in almost the same position, one for zoom one for full shots to change to. You can then use input cards (straight into the computer) or just the camera memory cards and edit the video without losing quality in zooms.
Thanks for all of your feedback.
I just want to clarify that when you write "Since you've already scaled the video to 720 (you will never see the original quality again), you are not blowing up to original quality." that you are talking about taking a nested 720 sequence and dropping it into a 1080 sequence. As opposed to scaling up 1080 footage in a 720 sequence to its full 1080 size, to create a closeup.
BTW, when I refer to zoom, I really meant a punch-in, not a camera zoom.
Why do you transcode at 60fps rather than 30?
OK, I ran a test yesterday. I exported the following (all in Prores422 unless otherwise noted):
1) 1080 sequence exported at 1080 (where punched-in shots are scaled up to 150%)
2) 720 sequence exported at 720. (where punched-in shots are at 100% ie 1080 and the original shots are in 66.67%) Enlarged in quicktime player to compare to the 1080 export.
3) 720 sequence (where punched-in shots are at 100% ie 1080) exported at 1080
4) 720 export (where punched-in shots were at 100% ie 1080 before exporting) re-transcoded to 1080
5) 720p sequence (where punched-in shots are at 100% ie 1080) nested into a 1080 sequence
6) 1080p sequence (where punched-in shots are scaled up to 150%) nested into a 720 sequence
I then compared them in quicktime viewer, at 150% view. Keeping in mind that my footage wasn't crystal sharp to begin with, and I added a sharpening effect-- I hardly noticed a difference in quality/sharpness between them all. There may have been a very very subtle difference but I had to look back and forth a few times to tell which one was better.
I also exported the 1080 sequence (#1 above) in ProResLT and didn't notice a difference between that and ProRes422.
Did you test areas of motion?
By the way, if you take a 720 sequence to a 1080 encode, you are scaling original footage down if interpreted, or clipping 1080 if not, then upping back to 1080. Some quality difference either way, but a lot less with your Punch In. With a punch in style drop in, you are not scaling down first, so no quality loss at that stage, only some effect quality adjustments if done gradually (with motion zoom; ken burns). So your quality will look very close going back up to 1080 on a full export. I suggest using LT for pro-res. Best file size for quality balance (along with compression speed when dropping a h.264 for youtube or an mpeg2dvd\x264 blu-ray in avchd raw stream). I shoot 60fps a lot of the time and transcode that.
However, I've come to appreciate a good 24p drop from that on occasion. Since I shoot singers and they are kids, parents love the sharp motion and clarity of the 60p, and for solos I usually drop down to 24 and telecine it with twixtor in a nested clip. The results are dashing. I keep the 60 and produce everything from disc to F4V-usb and YouTube screenings for output.