Here's few random comments from someone who has used Cinelerra for numerous DVD projects over
the past few years.
Firstly, I'm not sure why there's an emphasis on deinterlacing in the context of making a DVD.
While deinterlacing will make the clips and the resulting video look better when played back
on a non-interlaced monitor (which usually means on a computer via software which does not do
decent realtime software deinterlacing) it will actually make the resulting DVD look *worse*
when played back on a normal interlaced TV. The only exception to this is if the field order
gets changed between the raw video and the final DVD MPEG2 stream - then the DVD will look
absolutely terrible. One could fill a book with the details of field order, interlacing, when
to use it, how to get it right, etc etc - it's a complex subject which only gets worse in NTSC
land. The upshot though is that you only want to deinterlace material which will be viewed on
a computer; for DVDs destined for a normal TV you want to preserve interlacing (and field
As to cinelerra being a memory hog, I can confirm that this is the case. However, the extent
that this was encountered by the author is not consistent with my experience. Admittedly I
use the HV (Heroine Virtual) cinelerra, not the CV version seemingly used by the author, so
perhaps there are memory issues in the CV version which are not in the HV version. Even so, I
can confidently say that in 5+ years of using cinelerra for projects up to 6 hours long with
individual clips lasting 90 minutes I have never once had Cinelerra HV cause the OOM killer to
act. This is on a very lowly 866 MHz Pentium3 with only 512 MB of RAM and minimal use of swap
(in the past it's been even slower with less RAM).
Cinelerra does have a propensity to crash for no apparent reason though. "Save often" is
definitely good advice. It's interface is also somewhat unusual and takes some getting used
to. Once you know how to drive it you can do incredible things, but the learning curve is
unnecessarily steep which is unfortunate. In this day and age there are certainly more
intuitive GUI interface concepts/nomenclatures for achieving the same things.
This is certainly not to bash cinelerra. I think the authors (both HV and CV) have done an
incredible job on getting this complex tool to the point where it's at today. It's a tool
which does allow me to complete the projects I'm working on to a standard I simply couldn't do
with any of the other FOSS NLEs in existance today and for this I am very grateful. Having
said that, the developers behind the CV version are aware of the need to work on stability and
the GUI (there has been recent discussings about this on the CV mailing list). As stated in
the article though it does seem that the developer pool is rather thin and as a result any
significant work on cinelerra-cv (and the HV version for that matter) does take an inordinate
length of time.
I should also note that I use cinelerra only for the video portion of my projects. For the
audio side of things (which invariably involves multitrack sources and mixdown) I utilise
As a parting comment, I agree totally with the general conclusions reached in this article -
we have some promising FOSS video tools but they all need varying amounts of work before they
are ready for the masses. They can be used now to do "real work", but only if one is willing
to put up with the often considerable rough edges.