When and how to change the framerate

[nonoitall]

I've recorded a movie with an analog capture device via a satellite receiver, but instead of capturing at 29.97 fps (720x480) like I thought it would, one of the files I recorded ended up recording at 30.031 fps. I'm using AviSynth to sync audio and clip the parts of the movie properly, but I can't join them together when one of the clips is of a different framerate than the others. The movie is also telecined and I will be IVTC'ing it when I encode the movie into its final format. My question is when and how I should go about reducing the framerate in the problem file. Should I have AviSynth remove a frame every so often (ChangeFPS), or should I have it convert them smoothly (ConvertFPS)? In either case, should this be done before or after I apply the IVTC? (I'm assuming that using ConvertFPS on telecined material would probably create some nasty artifacts, but I'm new to this.) Thanks in advance for your suggestions.

[manono]

Hi-

It's hard to know what happened without examining the frames. Is it just speeded up, with all frames intact? Were frames dropped? If you could post a small section somewhere for anyone interested to have a look, that might help. And what is the final format, DVD? And in what format did you cap it, AVI, MPEG-2?

Personally, I don't think I'd use either ChangeFPS or ConvertFPS on it.

[FulciLives]

I had this FPS problem when I used a PCI capture card (BT chipset) ...

What I would do is use AssumeFPS(29.970,true) but that only seemed to work (for video and audio) when I was going UP in speed. For instance I would normally get 29.967 or 29.968 or 29.969 and using AssumeFPS(29.970,true) would work.

I would run the AVS through TMPGEnc Plus but only process the audio using LPCM WAV audio and using SSRC and it would spit out a corrected-for-length audio file.

I would then process the video through CCE SP and in the end the video and audio would match.

However once in a blue moon I would get 29.971 or 29.972 and the AssumeFPS(29.970, true) would work for the video of course but the audio would always crash TMPGEnc Plus and SSRC (actually it was SSRC that would crash).

So I would try AssumeFPS(29.970,false) and process the audio in a different way ... if it were me I would try the TIME WARP function in GOLDWAVE. I would shoot for the specifed time it should be (adjusting the video to 29.970fps) rather than try to adjust it by a percentage etc.

- John "FulciLives" Coleman

P.S.
The method(s) I outlined should work assuming the original capture (with the odd ball framerate) is in sync now as it is (i.e., before trying to adjust to 29.970fps).

I should also point out that when I used this "trick" I was always doing my captures in AVI format (usually with the PICVideo MJPEG codec).

[jagabo]

IVTC first, then everything will be at 23.976 fps.

[nonoitall]

I did try AssumeFPS, but this particular clip is about half an hour long and even that slight adjustment puts the audio noticeably out of sync after a few minutes, unless I alter the audio as you suggested. (I'd prefer not to alter the audio, other than changing its delay it so that it syncs up with video. If there's no other way though, I could use that as a last resort.) It was captured losslessly in an AVI container with software encoding (regular PCM audio and Huffyuv video), and I've kept all the original files. As far as the final format goes, I will be putting it on DVD and also possibly making a CD-sized XviD AVI from it as well. There don't appear to be any dropped frames, and it seems to follow the P-P-I-I-P telecined pattern, except on scene changes, where the pattern may reset, which I guess is normal(?). (Then again, I haven't checked through the whole clip.) If a clip would be helpful, let me know how long of one you need (in frames or seconds) and if you need audio with it.

[manono]

Hi-

IVTC first, then everything will be at 23.976 fps.

Not if the initial framerate is really something other than 29.97fps, it won't. You'll wind up with a framerate 80% of the capture framerate.

Since you don't like AssumeFPS (which is also what I would have suggested) because of also having to redo the audio, then IVTC normally and encode. Use DGPulldown on the final framerate. In the example you gave, the 30.031fps source will become 24.0248fps after IVTC. Encode that and then tick the Custom box in DGPulldown and set it for 24.0248->29.97fps. That way you can use the unchanged audio.

That will mean encoding the different parts separately, and then rejoining them at the authoring stage, which will be the only time they'll have the same framerate. I think it's either that or do it FulciLives' way, or use ChangeFPS on it, which will drop unique frames and make it play jerky, or use ConvertFPS on it, which will blend the heck out of it.

If what you said is correct, there's no need for a sample.

[nonoitall]

Well, I think I'll just go for stretching the audio. Hopefully the 0.2% decrease in speed won't be too noticeable. Are there any really good free ways to stretch the audio or is letting AviSynth do it in AssumeFPS just as good as any?

EDIT: I kinda answered my own question I think. Am I correct in understanding that AssumeFPS only changes the sample rate of the audio, without actually changing sample data (in which case I can resample the audio using SSRC to achieve consistent sampling rates)?

[guns1inger]

audacity might be able to do it

[kschang]

You may be able to much with the sampling rate. AVISynth have something that can let you much with that, I think, though it's probably best to demux the audio, muck with it, and mux it back when you're sure it matches.

[jagabo]

Hi-

IVTC first, then everything will be at 23.976 fps.

Not if the initial framerate is really something other than 29.97fps, it won't. You'll wind up with a framerate 80% of the capture framerate.

Not if he uses TDecimate() in mode 2 and specifies 23.976 fps.

[manono]

1. If you had in mind an unusual IVTC, I think you should have mentioned it in the first place. If I didn't understand what you were thinking, what makes you think that nonoitall would understand?
2. For many people on this site Decomb is the default and only IVTC, if they use AViSynth at all, and many, if not most, have never even heard of TIVTC.
3. TDecimate's Mode=2 will get it down to 23.976, but it does it by removing unique frames, something like 1 every 20 seconds or so (I think). It keeps it the same length, but if a "normal" IVTC brings the framerate is 24.0248, it starts dropping unique frames to achieve the desired final framerate. The result will play jerky, although better than using ChangeFPS on it. He said it was telecined normally, but with a framerate of 30.031fps. That's all I have to go on.
4. I don't have any beef with you. I really don't. You're just as knowledgeable (more in some areas) and much more patient than I. You are very helpful in many, many threads. But in this case, I think I stand by my earlier recommendations of either slowing it to achieve 23.976fps and also slowing the audio, or using DGPulldown on it after a normal IVTC.

[jagabo]

TDecimate's Mode=2 will get it down to 23.976, but it does it by removing unique frames, something like 1 every 20 seconds or so (I think). It keeps it the same length, but if a "normal" IVTC brings the framerate is 24.0248, it starts dropping unique frames to achieve the desired final framerate. The result will play jerky

His capture must have duplicate frames because the only way one could capture an NTSC signal at 30.031 fps is by duplicating occasional frames. TDecimate will presumably drop the duplicate frames preferentially. I admit that I have never tried using it that way though.

Since he was already using AVISynth I just gave another option that used it. If he wanted more information he could have asked.

[manono]

It could have dupe frames, but all he said about it was that the pattern changed sometimes at scene changes. Nothing about dupe or extra frames. I agree that it would explain the strange framerate, and also agree that TDecimate's Mode=2 might be useful in such a situation. But there could be other explanations. For example, maybe you've heard of networks speeding up the film so they could squeeze in more commercials, or dropping fields in a regular pattern for the same reason. I agree that this might not explain one of the files having the strange framerates, and the others not having it. Or maybe there's something screwy with the capture. FulciLives said he's seen similar things. Maybe I should have been more insistent about asking to see a sample. Actually, I'm getting pretty curious about this thing now.

nonoitall, if you're still around, and if you're still working on this thing, and if you're still willing, please post a sample. Audio isn't necessary. 30 seconds of a sequence with movement will be plenty. If you need to know how or where to upload, just ask.

[jagabo]

I wouldn't mind seeing a sample too.

[FulciLives]

My old BT chipset capture card almost always captured at 29.967fps or 29.968fps or 29.969fps instead of 29.970fps

I finally settled on using TheFlyDS as my capture program and I set the AUDIO up as the MASTER STREAM. I got the same results though using iuVCR and using AUDIO as the MASTER STREAM. I did try using VirutalVCR and the dynamic resample method but that never really worked on my old computer (which was very slow but fast enough for Full D1 captuer using PICVideo MJPEG on the 19 setting). I always figured it was the lack of speed and the computer's inability to keep up that resulted in no joy with VirtualVCR.

Anyways setting AUDIO as the MASTER STREAM was the only way I could keep A/V sync.

I normally detest attempting to do a IVTC unless I have an ultra clean source so normally I would just adjust to 29.970fps and encode as interlaced. As I said before I used AssumeFPS for the video and audio which seemed to work with no ill effects.

Anyways I just wanted to say the how and the why of it happening to me ... I'm sure setting AUDIO as the MASTER STREAM is what does it but again that is the only way to keep things in sync with some systems.

I should point out though that I never got less than 29.967fps nor any higher than 29.971fps (maybe 29.972fps) so the 30.031fps seems a bit excessive to me.

Also I had audio issues if I got above 29.970fps which usually resulted in me just doing the capture over (when I was using my BT chipset capture card I was mostly doing VHS and LaserDisc capture so it was easy to repeat a capture that had gone bad).

- John "FulciLives" Coleman