-
Compressor, “Optical Flow Technology,” 24p to 29.97 – horrid artifacts – semi-rant
Primarily a rant, as this appears to be a bug without a solution – found several similar posts on here, some even recent, and none ever have a resolving answer.
Footage shot on Sony F3, 1920×1080 24p. Edited @ 24p in FCP 7, graded in Color, sent back to FCP. Export self contained files and import to compressor to convert to 29.97 for broadcast (30 sec spots), same codec, Prores HQ.
I have tried many, many methods, but of course the two that look the best, setting frame rate conversion to “Better” or “Best,” the ones that employ Compressor’s “Optical Flow Technology,” inevitably are encoded with subtle – but quite noticeable and completely unacceptable – pulsating digital artifacts in random areas of the frame. It can happen on moving shots, or even still areas smack dab in the middle of still shots. It is MADDENING.
Of course I also tried “Fast” (every 5th frame is a duplicate), “Good” (the same crappy frame blending that I can get from FCP or After Effects, makes image look soft) and also tried After Effects with Pixel Motion (even crazier artifacts) and Adobe Media Encoder with all render options set to highest levels (comes out nearly identical to Compressor set to “Good,” soft frame blending.)
I’ve tried it with interlacing and without. I’ve tried exporting to different codecs. I’ve tried “sending” straight from Final Cut rather than exporting the self contained file. I’ve even thought about rendering everything at “Best” and at “Fast and bringing them all BACK into FCP to try to cut around the artifacts, but they exist in several places in each spot, not even sure if it would be worth it.
With delivery on Monday and still SD versions of the spots to create, I am at wit’s end, and have determined Compressor set to “Fast,” believe it or not, is really my only option, even with the dupe frames. Thoughts???
Guys, please – Do I have anywhere else to turn to get a decent 24-29.97 conversion?
-
17 Replies
-
Hi Tom,
Don’t use optical flow for that operation.
Go to the easiest and faster option “Fast (nearest frame)”.
That just will repeat the needed frames without re-processing and mudding all the picture.
rafael -
Thanks, that’s what I was resorting to, but if it’s the correct workflow than that makes me feel better.
Follow-up question:
Several of the stations that I am submitting to said that they would accept a progressive frame file for delivery, but if I introduce interlacing myself at this point, is there less chance that something they do, beynd my control, would make the image quality not look ideal for air? Or progressive/interlaced delivery is irrelevant?
-
Another option is to convert using proper 3:2 pulldown. You can do this with compressor or after effects. I forget the exact settings, but if you search the cow, you can find instructions on how to do this with compressor. I would imagine that would look better then straight up frame duplication.
Matt Lyon
Editor
Toronto -
Hi Tom,
Matt points the other option which in the end is similar to the “Fast”: repeating part of the image you already have without degrading the original picture.
“Fast” will repeat some full frames, “pull-Down” will make the new frames with half of the lines of the adjacent frames. The interpolated frames would be less noticeable.
This is how p24 stuff is played in NTSC TVs.
rafael -
Thanks, trying pulldown now.
To confirm – I am frame rate to 29.97 in the PRHQ dialog box, set duration to 100% in Compressor, setting rate conversion to fast, deinterlace to fast/upper field first (staying 1920×1080), and also setting “interlaced” to “top field field” in the ProRes HQ-specific dialog box. This will provide correctly interlaced 29.97 footage, yes?
What should I see different than the repeated fourth frame – a blended 5th one (field of each image?) or a split frame if I view fields in a 59.94 comp in After Effects for instance? I would like to be able to proof that it is done correctly…
Thanks!
-
If you take your processed clip back into FCP, you can inspect the item properties and confirm that it is 29.97 fps.
Then you should view your clip in the viewer, while it is set to %100, so you’ll be able to see the fields. The playback pattern should be:
AA BB BC CD DD
In other words, whole frame, whole frame, mixed frame, mixed frame, whole frame.
Find a section with a camera move or fast action, as it makes it easier to see the pattern.
HTH,
Matt Lyon
Editor
Toronto -
Hi Tom,
Based on the conversations here I’m assuming you mean 23.98p to 29.97i and not 24p to 29.97i though in a 30sec spot it’s not a big issue either way.
Here’s the settings for Compressor:
– import 23.976p footage to Compressor
– apply Quicktime encoding profile (usually ProRes 422 or Uncompressed)
– in Encoder tab of inspector, click Settings button and under Motion, change frame rate to 29.97
– in Frame Controls tab of inspector, turn frame controls ON, set output fields to lower, select “Fast (Nearest Frame)” for rate conversion, duration is 100% of source
– renderThe term “pulldown” is being used incorrectly in this thread. The process of adding 3:2 pulldown is really two processes. Pulldown is specifically slowing down aka:pulling down 24fps to 23.976fps. 3:2, or more accurately 2:3, is the process of creating and repeating fields to convert 23.976 to 29.97. If you are shooting 23.976 then all you need to do is add 3:2.
If you are in fact shooting 24p then ideally you do need to do both the pulldown and 3:2 processes. In which case change the the Set duration from 100% to 100.100% via the menu setting 24@23.976 in Compressor. Make sure you are encoding the audio and not doing a pass through so the audio slowdown is rendered as well.
30p… the new 24P!
John Heagy
-
[John Heagy] “If you are in fact shooting 24p then ideally you do need to do both the pulldown and 3:2 processes. In which case change the the Set duration from 100% to 100.100% via the menu setting 24@23.976 in Compressor. Make sure you are encoding the audio and not doing a pass through so the audio slowdown is rendered as well.”
I think for p24 the best way is to conform the footage to 28,98.
The audio slow down is physically unnoticeable.
rafael -
Thanks, guys, all very helpful info. The spots were encoded and delivered successfully. Just wanted to point out that the last 3-4 few replies in particular were instrumental in proper encoding for this particular workflow for those stumbling upon this post 3 years from now…a lot of the search results that I came up with before posting were not clear in their success, wanted to make sure that others with this issue know that this worked.
And now, the best part – the final invoice!
-
[Rafael Amador] “I think for p24 the best way is to conform the footage to 28,98.
The audio slow down is physically unnoticeable.”For a 30 sec spot in this case yes. For a 5 min or longer interview, absolutely not as the audio will drift out of sync.
Reply to this Discussion! Login or Sign Up