Can mpdecimate_trim create smaller sized videos that are the same length of time as the original video in order to retain all of the original audio? about mpdecimate_trim

[Yekutiel]

I did not.

[KenjiTakahashi]

You're right that this tool is not for you. It is explicitly meant to remove frames completely, along with the audio parts.
It sounds that you're looking to reduce the file size for videos that have "still portions" in them.
It is not possible the way you described, but it is actually what "frame dropping" mentioned in the README does. It finds identical (or "similar enough", if a decimation algorithm is used) frames and replaces duplicates with pointers to the original one.
Pretty sure there is a ffmpeg incantation that will do it for you, but I'm afraid you'll have to research the details for yourself.

[Yekutiel]

Thanks for taking the time to carefully respond to me.

It sounds that you're looking to reduce the file size for videos that have "still portions" in them.

That is correct.

To be clear, let's say I were to use OBS studio to create a screencast for 60 minutes while I was coding. Furthermore, let's say that for 10 to 15 minutes during that screencast I were to stop using my mouse and keyboard, yet continue talking as I, for example, thought out loud. For example, I might stop using my mouse and keyboard, yet say, "Hmmm. I'm a little confused here. I wonder if I should..."

During such times I would want my voice recorded; however, because I would neither be using my mouse nor my keyboard, software (computer code) that, for example, replaced video frames that were identical (or "similar enough", if a decimation algorithm were used) with pointers to the original video frame could significantly reduce the file size.

Pretty sure there is a ffmpeg incantation that will do it for you, but I'm afraid you'll have to research the details for yourself.

Given the extreme power and popularity of ffmpeg along with the usefulness of such a feature to many users, I presumed that ffmpeg must have such a feature.

[aaronse]

@Yekutiel, curious, did you find a solution that keeps synchronized visual frames and audio samples for segments where audio activity is detected?