Film -> PAL (change fps and a bit of restoring)

[poisondeathray]

1) "florida3.mpg" has blends and isn't very fluid

2) "florida4.mpg" has bad jerks when you look at it in avisynth with a bob deinterlacer (e.g. Tdeint(1,1) , or yadif(1,1) ) . If you examine fields (assumetff.separatefields) , it confirms wrong fields. But it's not as bad when you use VLC for some reason, I had VLC on loop playback, and it seems to studder the 1st time, but not the next loops. Maybe it's "learning" or adaptive?

I don't know what is different, but I wouldn't use it from the avisynth analysis. You can burn it to a test dvd and see if it plays ok. I personally wouldn't use this one.

3) yes the original option which manono gave you , keep it progressive, but the problems are the PAL audio speedup , and integration with your interlaced PAL project

4) you could try one of the motion interpolation options (e.g. mflowfps), that generate in-between frames , but it tends to generate spurious edge artifacts

The problem is you start with 23.976 and you want to go to 50 samples/sec . You can't make something from nothing. You need to generate those in-between frames somehow. The usual solutions are 1) duplicate 2) blend 3) frame interpoloate . They all stink to some degree. No solution is idea

[elmuz]

Thank you... I'm wondering how TV companies handle this situations... like MTV in PAL countries..

[poisondeathray]

I don't have that much experience with PAL, but I think they keep it progressive and do the PAL speedup . I suspect they use progressive , encoded as interlaced . But they are not mixing it with a true interlaced PAL video or project. You can broadcast it as 50i (with content as 25p), that is what the BBC does. So the temporal resolution is still 25, not 50 . You're trying for too much, you have to more than double the number of frames and then interlace it. That's a lot of information from nothing, you have to choose either choppy playback (dupes), blended frames, or motion prediciton artifacts if you want to upsample the temporal resolution.

I think the smoothest would be using convertfps. Motion interpolation using mflowfps is the best in some scenes (because they are new distinct frames) like the opening scene, but others it fails miserably because the prediction fails (e.g. camera flashes in the nightclub).

If this were my project, I would use convertfps for this. You get blends, but they are evenly spaced, without dupes, so playback is smooth . There are no field issues because it's done on frames, then interlaced after.

FFMpegSource2("video.mp4",atrack=-1)
TDecimate()
Convertfps(50)
LanczosResize(720,576)
AssumeTFF()
SeparateFields()
SelectEvery(4, 0, 3)
Weave()
SSRC(48000)

[Alex_ander]

The clear difference between the 2 ways of field insertion is that in one case (changefps-separatefields-selectevery-weave) both top/bottom inserts within the 25-frame period (24->25 for simplicity) are repeats of 'first' field* (so that top/bottom fields alternate), and in another case (separatefields-changefps-weave) they are repeats of last field and both are top or bottom (+ one of the weaves has 2 top or bottom fields! - depending on assumed field order).

*note: in standard soft pulldown methods for mpeg repeat_first_field and top_field_first flags are used

So those versions are still worth comparison in encoding.

TBTB...TBTBtBTB...TBTBTb (first version, 1st period, TFF assumed, small letters are field repeats)

TBTB...TBTBbTBT...BTBTBb (second version, 1 period, TFF)

[elmuz]

I haven't understood properly what you're saying...
If I have TB TB TB TB ... Bb BT BT ... do I have problems in any playback like DVD player?

Anyway, consider my situation (starting with a 24p and going to an interlaced PAL project) which of them would you suggest to convert that movie?
Suppose speedup isn't allowed, because here music matters...

[Alex_ander]

I haven't understood properly what you're saying...
If I have TB TB TB TB ... Bb BT BT ... do I have problems in any playback like DVD player?

Yes, I meant there might be some problems with this version:
.................
SeparateFields()
ChangeFPS(50)
Weave()

(it will not insert fields in proper manner)

...and suggested to test another one (I gave full script in the previous page):
.................
ChangeFPS(50)
AssumeTFF()
SeparateFields()
SelectEvery(4,0,3)
Weave()#TFF

poisondeathray recently suggested a similar version (ConvertFPS in place of ChangeFPS) which is potentially more fluent but at the cost of field blending (e.g. in future you will not be able to restore original frames if you ever need it for some reason). Try both.

[elmuz]

Thank you both... I have enough elements to go on...
We can say topic solved.