[FFmpeg-user] Is there an equivalent of Drobox Lepton but for h.264 videos?

[Andrew Randrianasulu]

сб, 30 сент. 2023 г., 19:31 Stéphane Archer <archerstephane at gmail.com>:

> On Sat, Sep 30, 2023 at 11:00 AM David Bernat <david.bernat at gmail.com>
> wrote:
>
> > Been very seriously considering design scope. Dropbox is a no-go. Want to
> > connect offline to describe your use case?
> >
> >
> > On Sat, Sep 30, 2023 at 10:10 AM Stéphane Archer <
> archerstephane at gmail.com
> > >
> > wrote:
> >
> > > On Sat, Sep 30, 2023 at 10:00 AM Andrew Randrianasulu <
> > > randrianasulu at gmail.com> wrote:
> > >
> > > > сб, 30 сент. 2023 г., 16:58 Stéphane Archer <
> archerstephane at gmail.com
> > >:
> > > >
> > > > > Is there an equivalent of Drobox Lepton but for h.264 videos?
> > > > > Lepton is a lossless compression algorithm to compress jpeg at
> around
> > > 20%
> > > > > there initial file size without any data loss. This is ideal to
> > > > > achieve jpeg.
> > > > > What about h.264 videos?
> > > > >
> > > >
> > > >
> > > > well, h264 videos already heavy and usually lossy compressed, any
> > further
> > > > compression will be not lossless
> > > >
> > >
> > > lepton works on jpeg that are lossy, yet they manage to have a smaller
> > file
> > > size adding a lossless compression on top of it.
> > > I'm not sure why it won't be possible to do the same on a h.264 file
> and
> > > get a smaller file size. I know h.265 and av1 have lossless video
> > > compression. And I recently heard about FFV1.
> > >
> >
>
> Dear David, I don't understand what you mean by "design scope".
>
> I would like to try a different analogy
>
> If the data was text and I wanted lossy compression then I would replace
> all the vowels with an "a" and then compress the resulting data with gzip.
>
> now if I switch gzip by using lzma instead, I would end up with a file
> smaller or the same size as the gzip version. Why does lossless h.265 or
> lossless av1 give me a larger file size than the original h.264? Why is
> this different for video?
>

As far as I understand all those h26x series of codes use

https://en.m.wikipedia.org/wiki/Motion_estimation

so  only small difference between non-keyframes encoded, making file
smaller (unless you try to compress say analog tv noise ...)

DCT part is usually lossy by dropping some numbers ...

so, if I understand correctly you can repack jpeg slightly differently, or
recompress it completely into something like jpeg-xl, but video usually get
a lot more info already removed, due to way only parts of image usually
changes from frame to frame.



> I hope my text compression analogy was clear.
>
> --
> Best Regards,
>
> Stephane Archer
> _______________________________________________
> ffmpeg-user mailing list
> ffmpeg-user at ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-user
>
> To unsubscribe, visit link above, or email
> ffmpeg-user-request at ffmpeg.org with subject "unsubscribe".
>