looking for s/w that can recreate an MP4 file at slower playback speed

[ginahoy]

I downloaded an MP4 video shot from International Space Station (http://vimeo.com/32001208). It's a time-lapse of the upper atmosphere, showing aurora and lighting, taken over several weeks. It's an incredible video but would be much more enjoyable if it were slowed down a bit. I would like to recreate the file at a slower speed (not just slow it down during playback). I would also like to replace the audio (music) background with something more to my liking.

Can someone recommend a 'freeware' converter or editor that can do either or both of these tasks?
Thanks!

[Baldrick]

You can try avidemux, video->frame rate or filters->resample fps and lower it. And you can change audio source under audio->main track->audio source->external mp3/wav.

[poisondeathray]

Either you mux with a lower fps (it just plays back slower, you can do this with mp4box / yamb for example) , or you attempt to generate slow motion in post . You lose no video quality with muxing , but the other methods require re-encoding

Muxing at a lower fps means exactly that: If the original file is 25fps, a 4x slowdown would play it at 25/4 = 6.25 FPS. It will look like a slideshow (because a timelapse is a slideshow)

There are 3 basic approaches to produce slow motion after the footage has been shot. They all attempt increase the framecount so it was as if you shot at a higher FPS. 1) Duplicate frames, 2) Blended Frames, and 3) optical flow or motion interpolation . So a 4x slowdown would need 4x the number of frames added per unit time.

Duplicates produce choppy motion like the muxing ; blends produce smoother motion but are blurry ; motion interpolation attempts to generate new "in-between" frames from motion vectors of the surrounding frames, but can suffer from morphing edge artifacts, and computationally more intensive . Also if you have frame errors, it can propogate those errors into "good" frames, so you need to fix the "bad" frames first. If you're familiar with twixtor , kronos , or after effects' pixel motion or timewarp, those are examples of the 3rd option. The free way is mflowfps in avisynth, or msu frc .

They all have drawbacks, but method 3 usually produces the smoothest motion if you can tolerate the artifacts. Some source types are interpolated very well with few artifacts, others can produce disasterous results. The "best" way is to shoot a higher FPS, this way every frame is real and unique.

Audio is simple, you can just mux it in. If you need specific timing, it helps to use a video editor, but you can shift the audio with most programs

Here is a small comparison of the 3 methods

[Gavino]

So a 4x slowdown would need 4x the number of frames added per unit time.

You mean 3x of course.
Three frames added to each original one giving a total of 4x.

[poisondeathray]

Yep, thx for the correction Gavino.

[ginahoy]

THANKS PDR for such a detailed explanation of the options. As a novice to all of this, the clip you prepared was extremely helpful and appreciated. It may be a few weeks before I have time to mess with this, but I'm so glad I found this forum, as I am likely to have further questions.

[Edit] removed question on meaning of 'muxing' as I found answer at program links provided above. Thx.

[ConverterCrazy]

Either you mux with a lower fps (it just plays back slower, you can do this with mp4box / yamb for example) , or you attempt to generate slow motion in post . You lose no video quality with muxing , but the other methods require re-encoding

Muxing at a lower fps means exactly that: If the original file is 25fps, a 4x slowdown would play it at 25/4 = 6.25 FPS. It will look like a slideshow (because a timelapse is a slideshow)

There are 3 basic approaches to produce slow motion after the footage has been shot. They all attempt increase the framecount so it was as if you shot at a higher FPS. 1) Duplicate frames, 2) Blended Frames, and 3) optical flow or motion interpolation . So a 4x slowdown would need 4x the number of frames added per unit time.

Duplicates produce choppy motion like the muxing ; blends produce smoother motion but are blurry ; motion interpolation attempts to generate new "in-between" frames from motion vectors of the surrounding frames, but can suffer from morphing edge artifacts, and computationally more intensive . Also if you have frame errors, it can propogate those errors into "good" frames, so you need to fix the "bad" frames first. If you're familiar with twixtor , kronos , or after effects' pixel motion or timewarp, those are examples of the 3rd option. The free way is mflowfps in avisynth, or msu frc .

They all have drawbacks, but method 3 usually produces the smoothest motion if you can tolerate the artifacts. Some source types are interpolated very well with few artifacts, others can produce disasterous results. The "best" way is to shoot a higher FPS, this way every frame is real and unique.

Audio is simple, you can just mux it in. If you need specific timing, it helps to use a video editor, but you can shift the audio with most programs

Here is a small comparison of the 3 methods

I know this is an old thread (to those who read it before). But the sample video clinched the point.
I wish the full Avisynth code for Motion Interpolation would be great for people like me who depend on this forum for help.
I have always been helped before.

[poisondeathray]

I know this is an old thread (to those who read it before). But the sample video clinched the point.
I wish the full Avisynth code for Motion Interpolation would be great for people like me who depend on this forum for help.
I have always been helped before.

I don't remember what exactly was used (it's an old thread), I think the default mflowfps from mvtools2. There are many variations you can use based on mvtools2, such as svpflow, interframe, many wrapper functions based on mvtools2 such as smoothfps, smoothfps2. - and you can play with various settings like blocksize to get slightly better (or slightly worse) results. A quick search will show you example scripts of all of those. If you still need help just ask

The newest avisynth interpolation function is FrameRateConverter , which does a slightly better job in general of hiding the occlusion artifacts. Many settings to tweak too.
https://forum.doom9.org/showthread.php?t=174793
https://github.com/mysteryx93/FrameRateConverter

[ConverterCrazy]

I know this is an old thread (to those who read it before). But the sample video clinched the point.
I wish the full Avisynth code for Motion Interpolation would be great for people like me who depend on this forum for help.
I have always been helped before.

I don't remember what exactly was used (it's an old thread), I think the default mflowfps from mvtools2. There are many variations you can use based on mvtools2, such as svpflow, interframe, many wrapper functions based on mvtools2 such as smoothfps, smoothfps2. - and you can play with various settings like blocksize to get slightly better (or slightly worse) results. A quick search will show you example scripts of all of those. If you still need help just ask

The newest avisynth interpolation function is FrameRateConverter , which does a slightly better job in general of hiding the occlusion artifacts. Many settings to tweak too.
https://forum.doom9.org/showthread.php?t=174793
https://github.com/mysteryx93/FrameRateConverter

Thanks for taking time to reply.
I will proceed as suggested.

[ConverterCrazy]

Sorry.