Unique Frame Rate / Speed Problem

[Croweyes1121]

Guys, I have a VERY strange problem.

I'm trying to complete a fanedit, but I've run into a snag. I have a broadcast recording that is 29.97fps (23.976fps once detelecined properly). The broadcast was evidently played too slow, meaning the pitch is too low on the music, and the scenes take slightly too long to play out. And yet, the 23.976 frame rate produces smooth motion. How can this possibly be? Either the broadcasted frame rate is incorrect for the film (which should introduce motion issues, which it does not), or the scenes should play at the proper speed/pitch when at the broadcasted frame rate. I have never seen anything like this.

So I'm left with two options:
Option 1: raise the audio pitch to match the BD pitch, and leave the motion smooth on the broadcast. This has its own set of problems - most importantly, this screws up all of my audio transitions where I'm letting the TV footage audio play out over several shots, including BD footage, because when I transition back into BD footage, the music is in a different place than it was due to the speed change.

Option 2: change the speed / pitch / duration of the broadcast footage. This fixes everything, but produces choppy motion due to dropped frames inherent in speeding up footage.

Please help!

[johns0]

Use audacity to change pitch without changing tempo for the audio.

[Croweyes1121]

Use audacity to change pitch without changing tempo for the audio.

See "Option 1" in my original post. The problem with this method is that the broadcast is still too slow in relation to the Blu-ray, so even if I change the pitch of the audio AND maintain its tempo, the footage on the broadcast is running too slowly in relation to the BD footage.

Is there any way the broadcast can be 23.976fps and still be too slow compared to the BD? Shouldn't 23.976fps match the BD, because that's the frame rate of the film? I don't see how the broadcast and the BD can both play back smoothly at 23.976fps, yet be playing the film itself at different underlying speeds.

[manono]

Option 2: change the speed / pitch / duration of the broadcast footage. This fixes everything, but produces choppy motion due to dropped frames inherent in speeding up footage.

The video can just be speeded up. No dropped frames. If, for example, your video was came from a PAL 25fps source originally, it can be speeded up to 25fps again, with the audio also speeded up to match and also pitch-adjusted if necessary.

[Croweyes1121]

Option 2: change the speed / pitch / duration of the broadcast footage. This fixes everything, but produces choppy motion due to dropped frames inherent in speeding up footage.

The video can just be speeded up. No dropped frames. If, for example, your video was came from a PAL 25fps source originally, it can be speeded up to 25fps again, with the audio also speeded up to match and also pitch-adjusted if necessary.

But the movie itself is 24fps (or 23.976fps, whichever). The source file I have is the 29.97fps broadcast, but it absolutely plays too slowly. How would you go about introducing a 4% speed-up to that source without dropping frames in the process? Either way, the end result MUST be 23.976fps for my project, because I am combining this TV footage with BD footage which is already 23.976fps.

[ProWo]

Cut an 3min sample (ge with avidemux) and post it here

[Croweyes1121]

Here you go.

It's 29.97 at the source. It needs to be sped up to roughly 104% of its current speed (pitch AND duration), while maintaining the smooth motion of this original clip. The end result must be 23.976fps.

[poisondeathray]

Did you compare both ? Are they the same version, same edit, same frames ?

[Croweyes1121]

Did you compare both ? Are they the same version, same edit, same frames ?

They are from the same film. I'm trying to incorporate TV footage into the BD of the same movie. The only thing that makes any sense to me is that the film footage for the TV version was transferred onto a 25fps medium prior to editing, then slowed down (improperly) to 23.976fps for US broadcast. If that's so, then the original source for the TV version is 25fps, and I'm just out of luck altogether.

[poisondeathray]

And yet, the 23.976 frame rate produces smooth motion.

No.

It's not "perfectly smooth" there are motion issues; after IVTC there are duplicates at 23-24, 49-50, 68-69, 97-98... ie. a pause or glitch almost every second

(Or if you want, it's not perfect 3:2 telecine, there are breaks in the pattern with some 4 duplicate fields)

[Croweyes1121]

And yet, the 23.976 frame rate produces smooth motion.

No.

It's not "perfectly smooth" there are motion issues; after IVTC there are duplicates at 23-24, 49-50, 68-69, 97-98... ie. a pause or glitch almost every second

(Or if you want, it's not perfect 3:2 telecine, there are breaks in the pattern with some 4 duplicate fields)

Any way to properly fix it?

[manono]

Any way to properly fix it?

I haven't looked at the sample. Why didn't you discover the dupe frames that it took pdr to point out to you?

Anyway, if the net effect of the dupe frames is to add one per second (or thereabouts), then speed it up to 25fps as I suggested earlier followed by removing the dupes to bring it back to 23.976fps. Perhaps combine that with the audio adjustment originally suggested by johns0.

[poisondeathray]

Bascially the same thing as manono said, but looking at it bit differently - If you IVTC to 23.976 and there are duplicates, this means the effective frame rate is <23.976 when you remove the duplicates. This is consistent with "scenes take to long to play out" and "pitch too low" observations . So if you speed up that decimated (<23.976) framerate version to 23.976, hopefully it should be "right"

The problems I see are the pattern is not consistent, and some scenes have low motion, this can make "automatic" decimation filters prone to errors. Also, broadcast versions can be different - ie. broadcast edits. You might have to do some fiddling in a NLE to make it match better

[Croweyes1121]

Any way to properly fix it?

I haven't looked at the sample. Why didn't you discover the dupe frames that it took pdr to point out to you?

Anyway, if the net effect of the dupe frames is to add one per second (or thereabouts), then speed it up to 25fps as I suggested earlier followed by removing the dupes to bring it back to 23.976fps. Perhaps combine that with the audio adjustment originally suggested by johns0.

Hey, nobody’s perfect. That’s why we have a forum for questions.

What’s the best program/method for removing dupe frames? I did a quick test with 30 seconds worth of the original 29.97fps footage, removed every duplicate frame manually, exported the remaining frames at 29.97fps, changed the rate to 23.976fps with tsmuxer, reimported the result into a 23.976 project in Premiere, then time-stretched the original audio to match the new shorter length of the video, and EUREKA! Correct speed, pitch, and no wonky motion. But 30 seconds of manual deletion took a few hours, so there must be a quicker way.

[poisondeathray]

There are "automated" ways, such as duplicate decimation filters in avisynth

http://avisynth.nl/index.php/External_filters#Duplicate_Frame_Detectors


So I would fieldmatch first such as with TFM(), then apply one of those dupe removal filters, then AssumeFPS(24000,1001, sync_audio=true) (or if you want use a dedicated audio editor for the audio)

But some potential issues, e.g. the 2nd scene, the distant shot, there is low motion compared to scenes with close up galloping horses. This scene is prone to errors. All these decimators use metrics and thresholds to compare similarity between frames. Other scenes are easy, but that low motion or similar one could be problematic, and if you remove "wanted" similar frames, you're going to introduce new sync issues.

If the pattern was a bit more consistent, you could use pattern based decimation (e.g. n in m decimation; ie. in a group of m frames, decimate n frames). You could still probably approach it with this, but you need to examine it more closely to see the rough pattern first

[jagabo]

I tried DeDup() after TFM(). On a sequence of 1000 frames with constant motion (990 to 1899) I got 775 unique frames, and I'm pretty sure no unique frames were dropped. That works out to a very unusual 23.22677 fps (30000/1001*0.775). TDecimate(mode=2, rate=23.22677) gave 775 frames.

[Croweyes1121]

I tried DeDup() after TFM(). On a sequence of 1000 frames with constant motion (990 to 1899) I got 775 unique frames, and I'm pretty sure no unique frames were dropped. That works out to a very unusual 23.22677 fps (30000/1001*0.775). TDecimate(mode=2, rate=23.22677) gave 775 frames.

I've sent you a PM.