Confused about speeding up a video

[nnamerz]

I have some CCTV recordings that I would like to speed up (like a timelapse) and I am confused about which method to use. I tested Avidemux and VSDC Video Editor and they both speed up videos in different ways:
•The "Change FPS" filter in Avidemux appears to speed up a video by cramming in more frames per second.
•The "Speed (%)" setting in VSDC appears to speed up the video by discarding extra frames, but keeps the final frame rate the same as the original file.

Is one method better or preferred over the other? Are there any major pros/cons for each method?

Here are the "MediaInfo" results for an MP4 test file I sped up by 10x using both methods:

ORIGINAL FILE
FILE SIZE/LENGTH: 37 MiB / 1 min 7 s 134 ms [00:01:07.134 (00:01:07:03)]
VIDEO BITRATE/FPS: 4 678 kb/s () @ 15 FPS (Constant) [Total Frames: 1008]
SIZE/ASPECT: 1920px * 1080px @ 16:9 (1.778) Aspect Ratio
VIDEO FORMAT: AVC (Advanced Video Codec) / High@L4 / CABAC / 1 Ref Frames
VIDEO CODEC: video/H264 / Advanced Video Coding / 8 bits / Progressive Scan

VSDC (Changed speed from 100% to 1000% to speed up by 10x)
FILE SIZE/LENGTH: 14 MiB / 6 s 734 ms [00:00:06.734 (00:00:06:11)]
VVIDEO BITRATE/FPS: 17.0 Mb/s () @ 15 FPS (Constant) [Total Frames: 101]
SIZE/ASPECT: 1920px * 1080px @ 16:9 (1.778) Aspect Ratio
VIDEO FORMAT: AVC (Advanced Video Codec) / High@L5.1 / CABAC / 3 Ref Frames
VIDEO CODEC: video/H264 / Advanced Video Coding / 8 bits / Progressive Scan

AVIDEMUX (Changed FPS from 15 to 150 to speed up by 10x)
FILE SIZE/LENGTH: 17 MiB / 6 s 727 ms [00:00:06.727 (00:00:06::9)]
VIDEO BITRATE/FPS: 20.0 Mb/s () @ 150 FPS (Constant) [Total Frames: 1009]
SIZE/ASPECT: 1920px * 1080px @ 16:9 (1.778) Aspect Ratio
VIDEO FORMAT: AVC (Advanced Video Codec) / High@L5.2 / CABAC / 4 Ref Frames
VIDEO CODEC: video/H264 / Advanced Video Coding / 8 bits / Progressive Scan

[DB83]

IMHO it is a no-brainer.

You would require to keep all the frames since critical info could be lost. So the avidemux method would be prefered.

[nnamerz]

IMHO it is a no-brainer.

You would require to keep all the frames since critical info could be lost. So the avidemux method would be prefered.

Would there be any potential drawbacks to having video files with incredibly high framerates? Or any issues with playing the video on certain devices?

[jagabo]

Would there be any potential drawbacks to having video files with incredibly high framerates? Or any issues with playing the video on certain devices?

Yes, there is a limit to how fast devices can decompress frames, and a limit to how fast those frames can be sent to the display. Many devices cannot decompress 1080p video at 150 fps. And even if you have a device that can decompress that fast you would likely only see 60 of the resulting frames per second as that's the limit of connection between the device and the display. Your better bet for this would be to use a combination of the two methods.

[nnamerz]

Would there be any potential drawbacks to having video files with incredibly high framerates? Or any issues with playing the video on certain devices?

Yes, there is a limit to how fast devices can decompress frames, and a limit to how fast displays can display them. Many devices cannot decompress 1080p video at 150 fps. And even if you have a device that can decompress that fast you would likely on see 60 of the resulting frames per second as that's the limit of the TV/monitor. Your better bet for this would be to use a combination of the two methods.

Thanks for explaining that.
Is there anyway you can elaborate a bit more on how I could use the combination of both methods?

[jagabo]

Is there anyway you can elaborate a bit more on how I could use the combination of both methods?

Discard some frames, and display the remaining frames faster. If you wanted to speed up 15 fps video 8 fold you could discard every other frame (2x) and display the remaining frames 4x faster, for a net increase of 8x and a 60 fps result that most players can handle. I believe you can do that with AviDemux and a combination of the Resample FSP filter and the Change FPS filter.

[nnamerz]

Is there anyway you can elaborate a bit more on how I could use the combination of both methods?

Discard some frames, and display the remaining frames faster. If you wanted to speed up 15 fps video 8 fold you could discard every other frame (2x) and display the remaining frames 4x faster, for a net increase of 8x and a 60 fps result that most players can handle. I believe you can do that with AviDemux and a combination of the Resample FSP filter and the Change FPS filter.

I apologize for all the newbie questions, but I just wanted to get a bit more clarification on the order of steps to take.
In the example you mentioned (the 15 fps file sped up 8 fold), would I need to use the "Resample FPS" filter first? And to discard every other frame, would I enter 7.5 fps for the "New Frame Rate"?
If so, would the next step be to use the "Change FPS" filter and enter 60 fps for the "Destination Frame Rate"?

[jagabo]

In the example you mentioned (the 15 fps file sped up 8 fold), would I need to use the "Resample FPS" filter first? And to discard every other frame, would I enter 7.5 fps for the "New Frame Rate"?
If so, would the next step be to use the "Change FPS" filter and enter 60 fps for the "Destination Frame Rate"?

That should work.

[nnamerz]

That should work.

Thank you so much for your help.

On a side note, would you happen to know if VSDC also allows something similar to be achieved? The reason I ask is because it seems like VSDC exports the videos much faster than Avidemux does on my system. Thanks again

[jagabo]

don't know VSDC at all. But look for similar filters. Resampling is often called decimation. Speed may be a matter of encoding settings. For example, there's can literally be a 100 fold difference in encoding speed between x264 at the Ultrafast (fastest) and Placebo (slowest) presets.

[nnamerz]

don't know VSDC at all. But look for similar filters. Resampling is often called decimation. Speed may be a matter of encoding settings. For example, there's can literally be a 100 fold difference in encoding speed between x264 at the Ultrafast (fastest) and Placebo (slowest) presets.

Thank you again for your help

[DB83]

Whilst it has been stated that whatever method you use drops frames (which could be vital) since practically all players allow you to scan through a video by 'speeding' up the up playback would it not be more prudent to leave the recording/capture 'as is' and let the player do the 'hard yards' ?

[nnamerz]

Whilst it has been stated that whatever method you use drops frames (which could be vital) since practically all players allow you to scan through a video by 'speeding' up the up playback would it not be more prudent to leave the recording/capture 'as is' and let the player do the 'hard yards' ?

That's exactly what I had done up until this point, but I'm now running low on hard drive space and a lot of these CCTV recordings are very large files (4 - 6 GB). Most of them are a continuous recording of my backyard when we had renovations done. And since they aren't vital recordings, I realized I could speed them up and have a timelapse version that results in a file sizes that are at least 60% smaller.

[nnamerz]

On a side note, would anyone familiar with VSDC happen to know how to speed up my 15fps footage the same way Avidemux does (as mentioned in the previous comments). I spent the last 2 hours searching forums and google and VSDC tutorials, but can't seem to find "Resample FPS" or "Decimation" anywhere in the settings. I found a "Resample" effect but that is just a visual video effect/transition.

The closest I came to getting an output file at 10x speed with 60fps was to set the "Speed" to 275% (Editor tab > Properties window) and then to set the output framerate to 60 fps (Export tab > Edit Profile window). Figuring out the speed percentage to get 10x was a bit of a pain since I'm not great with math.
To try and simply the math, I also tried setting the input framerate to 60fps (while creating a blank project) and then importing my 15fps video and adjusting the speed to 1000% and leaving the output framerate as "Original".

Both these methods appeared to create a 10x file with 60fps, but I'm not sure if tricking VSDC into thinking my input file is 60 fps will cause any issues with my output file. I'm also not sure why using these methods resulted in a significantly larger file size than Avidemux outputs since I used the same encoding settings.

[DB83]

To paraphrase your post, on a side note, if these recordings are not vital then why are your keeping them ?

I can understand recording for security purposes yet when it is determined that there is no 'breach' I can see no purpose in keeping them.


As for speed-up, since the source is 15 fps and a display should handle 60 fps without issue then surely a 75% saving is adequate. But then if there is no real purpose in keeping these.....

[nnamerz]

To paraphrase your post, on a side note, if these recordings are not vital then why are your keeping them ?

I can understand recording for security purposes yet when it is determined that there is no 'breach' I can see no purpose in keeping them.


As for speed-up, since the source is 15 fps and a display should handle 60 fps without issue then surely a 75% saving is adequate. But then if there is no real purpose in keeping these.....

I apologize for not being more clear. These recordings weren't for surveillance as much as they were for having a record of the "before & after" transformation. When I said they weren't vital, I meant in the sense that I didn't need to keep the original footage with all the frames and original quality. I have other recordings where every frame and maximum quality was vital (eg. the footage I gave police when my neighbour's car was broken into). But our backyard renovation videos don't need the same requirements....plus, some of them are several hours long and nobody is very likely to sit and watch through the full videos so I figured that speeding them up like a timelapse would not only save me hard drive space, but would also make the videos more manageable/interesting to watch.

In regards to the 2nd half your response, I totally agree that 75% savings is plenty, but would you happen to know how to achieve that in VSDC by just increasing the framerate instead of using the "Speed" setting which increases speed by dropping frames?

[jagabo]

Specify the bitrate that gives the size you want:

bitrate = size / running time

Use a bitrate calculator if you don't want to do the math yourself .

[nnamerz]

To paraphrase your post, on a side note, if these recordings are not vital then why are your keeping them ?

I can understand recording for security purposes yet when it is determined that there is no 'breach' I can see no purpose in keeping them.


As for speed-up, since the source is 15 fps and a display should handle 60 fps without issue then surely a 75% saving is adequate. But then if there is no real purpose in keeping these.....

UPDATE: So after converting a few of the files (using 23 CRF), I ended up only saving 15% in file size. It seems that the 75% file size reduction is only valid if you keep the bitrate the same as the original file. But since I was increasing the frame rate by 4x, I was also increasing the amount of data needing to be shown each second, so the bitrate also roughly increased 4x which resulted in much larger file sizes.

To reduce the file sizes, I wanted to ask if it's better to lower the frame rate and drop some files, or to lower the bitrate and lose a bit of quality?

[Cornucopia]

Bitrate has no direct relation to framerate, other than in uncompressed material.

Bitrate has a relationship to quality, but it isn't linear, and because quality is also affected by so many other factors - codec, codec settings, resolution, motion...- you cannot make a simplistic guess about quality other than very broad generalities and rules of thumb.

Framerate also has no direct relationship to quality other than the rule that higher framerates require higher bitrates in order to maintain relatively equivalent quality, but due to the fact that higher framerates means the delta change between frames is less, this means the growth of bitrate requirement is not linear, but more logarithmic. Again, since the motion in the scene plays a part in the complexity, one can't reduce this requirement to a simple formulaic expectation either.

Which things to trade off is totally dependent on the material and on your subjective preferences.


Scott