Video editing in open source is a tumultuous
space; projects come and go, and when they stick around they have a
tendency to embark on sudden ground-up rewrites and media-framework
transplants. But the dirty little secret is that that situation is just as
true in the "professional" proprietary market as it is for those of us who
stick with free software. Fortunately, while ambitious efforts like Lumiera remain experimental, the formerly
low-end nonlinear video editors (NLEs) are closing the feature gap by
adding steady, small improvements. Case in point is the KDE-based NLE Kdenlive, which released its most recent
update in October 2011. Version 0.8.2 adds new audio/video inspection and
monitoring tools, as well as effects that offer real editing functionality
— not mere eye-candy. The high end of the market is not as safe as
it used to be.
The 0.8.x series introduced a slew of new
features — mostly in the "special effects" vein, but some that
contribute to better project management and usability, too. The project
wiki maintains a near-exhaustive list
of build and installation options for users on every platform supported by
KDE, including multiple Linux distributions. I opted for the Launchpad
personal package archive (PPA) maintained for Ubuntu by Olivier Banus — it was "strongly recommended" over the official repositories, although exactly why is not explained. The primary package dependency to worry about is the low-level video editing framework MLT, which is also used by the GTK+-based OpenShot NLE (and is also provided by most of the project's distribution-specific package repositories).
On the whole, Kdenlive conforms to the same basic user-interface conventions shared by every major NLE. Each editing project you start consists of a set of video, audio, and other media "clips" that you select, and a timeline on which you drop, rearrange, and align those clips. Your experience in the editor is defined by manipulating items on the timeline, and any changes or effects you make to individual clips on the "clip monitor." When you are satisfied with your results, you render the final output into the format you want.
OpenShot, PiTiVi, and Kdenlive all offer largely comparable approaches
to this workflow, and have roughly the same level of support for the video
input and output formats most people need. Where they differ is in the
details: the ability to create special effects, the smoothness of
manipulating clips and settings, and support for esoteric non-video content
that you might want to generate for use in your production.
The canonical example of this is title cards; most NLEs allow you to
assemble text and images into title elements directly within the editor
(albeit with varying degrees of sophistication), rather than forcing you to
create them in an external application. Unlike some of the others,
Kdenlive can also generate
countdown sequences, clips of solid-color backgrounds, still-image
slideshows, and static "noise."
Via MLT, Kdenlive uses a plug-in API called Frei0r for its audio and video
effects. The 0.8.2 release includes several hundred effects, ranging from
simple audio/video corrections, to image transformations, to decorative
effects. Over the years, Kdenlive has also added a series of inspection
and monitoring tools that help you get a better handle on the contents of
your media files: you can monitor audio waveforms and color scopes as you
skip through your video, which makes editing far more precise than relying
on pointing your own eyes and ears at the preview window.
I am not a video professional, but, based on discussions with friends and relatives who are, it is these sorts of effects and monitoring options that make up some of the biggest differences between "consumer" and "professional" NLEs. For example, a lightweight editor will allow you to cut and rearrange clips to your heart's content, but without a way to adjust the colors so that two clips shot in different locations have the same skin tones, the lightweight editor will never be useful for any kind of professional-looking output.
On that front, I am pretty pleased with what Kdenlive currently offers. For video, you have control over hue, contrast, and saturation, complete with adjustment curves (which, as with photographs, allow you to vary the amount of adjustment in different parts of the image) and automatic balancing. The various "scopes" for monitoring a clip include histograms, waveforms, and a vectorscope. For audio, there are standard VU meters, spectrum and spectrogram monitors, plus a dizzying assortment of audio filters.
Among the new special effects are some novelty techniques such as a
stop-motion animator and a light-painting
tool, but there are others with more practical aims. The first is a
perspective image placement tool, which allows you to put a rectangle onto a
clip and distort its sides and corners until it appears to match the
perspective of other in-frame surfaces. You can then composite a still image (or
another video clip) into the properly-aligned rectangle and have it render
as if it were physically present in the scene. The second is named
rotoscoping, but it essentially just allows you to mask off areas
of the screen by drawing Bezier curves right onto the monitor. You can
draw such masks on several keyframes, and the effect plug-in will
interpolate on the intermediary frames. With a little feathering (i.e. an alpha-transparent border) added to the edges of your mask, you can replace objects or backgrounds. Both of these are effects that proprietary NLEs have offered (on other operating systems) for some time, but their addition to Kdenlive is another sign that the program is growing up.
Also in the "growing up" category are some changes to the user interface
and project-management features. The biggest is that the individual
components of the UI are all detachable and can be rearranged at will. For
starters, this allows you to work around older annoyances like the
inability to see the "clip monitor" and "project monitor" at the same time
(as noted by editor Jonathan Corbet in 2008). But it also enables you to assemble separate layouts for specific tasks and save them for later reuse, so you can activate all of the audio monitoring options for an "audio editing layout" or set up video monitors for a "color correction layout." There are a lot of monitors in 0.8.2, and fitting them all on-screen at once is not feasible.
The project-management features I found welcome include a "notes" pane
which allows you to keep text notes on your work directly within the
project (a method that definitely beats maintaining them in a separate file
that you periodically misplace), and something called "proxy clips" in the
main editors. Proxy clips are CPU-friendly transcoded versions of your
original clip which are shown in the effects and preview windows instead
of the original, thus saving CPU time and speeding up the responsiveness of
the entire UI. When a proxy clip is rendered, Kdenlive uses the original
clip, but working
on the proxy clip makes life easier at every step up to the final one. In
a take-a-moment-to-think-about-it twist, the proxy clips are usually larger
in size than the original — it is the highly-compressed HD video
formats that consume lots of CPU, after all. Proxy clips decoded to something simpler like vanilla MPEG-2 use more bytes, but don't require overhead for every frame.
The final cut
Kdenlive has evolved considerably since I last took it for a drive
(which if I recall correctly, must have been sometime during the 0.6.x
series). There are, however, several areas where I still find the editing
experience painful. The first is that many of the mission-critical UI
widgets are tiny. Although the toolbar buttons for generic items like
Open, Save, and Undo get a full 32 pixels each, the timeline's playback
point (which you must grab and slide in order to change positions in your
movie) is barely 8 pixels high. That is better than the playback point in
the clip monitor, however, which is 5 pixels high. Most of the tool buttons on the monitors and timeline are larger than that, but they are still microscopic by GUI standards, which makes them harder to hit and harder to tell apart.
I also found the effects interface inconsistent. It uses bottom-tabs (which is atypical to begin with), but activating an effect is the real issue. A checkbox next to each available filter would be simple and unambiguous, but to activate an effect, you must select it from the "Effect List" tab, then drag it onto the clip on the timeline that you wish to apply it to — after which, it will appear in the "Effect Stack" tab back in the effects window. There is a separate interface for activating and editing transition effects, by means of pulsating icons that appear in the timeline only when you pause the cursor over the right spot. The transient pulsating buttons appear in a few other places when working with the timeline, such as when resizing a clip. That does not make sense to me either; in all of these cases there is nothing to be gained by hiding the button in some circumstances, nor does the animation make the button easier to see or hit.
Ultimately, as is the case with the wildly-varying button sizes, the
transient pulsating buttons just contribute to an inconsistent look across
the application as a whole. Some of the icons are flat in style, some are
3D and rounded, some have boxes drawn around them and some have circles.
All of the new monitors use a black canvas as a background, while the rest
of the interface uses the widget library's default color. It may be unfair, but little details like that can turn off potential new users, particularly artsy ones like video editors.
The good news is that Kdenlive is making steady progress in adding the
mid-to-high-end features that it will need to eventually make a play for
the "pro-sumer" NLE user. That pro-sumer user is not just the Linux fanatic
patiently waiting on Lumiera, either — there is a great opportunity
in the NLE market right now on every platform. Apple alienated quite a
of its own professional users when it released Final Cut Pro X in June
2011, which many customers considered a downgrade. Certainly lots of them
simply jumped ship to Adobe Premiere, but the competition got a thorough
examination. From the other direction, EditShare announced in 2010 that it
would release a cross-platform version of its flagship Lightworks NLE as open source, but it
has subsequently pushed back the release date to "to be announced."
Meanwhile, the existing open source projects — including Kdenlive
— keep on plugging away, release by release, feature by feature.
Comments (6 posted)
Not so long ago, a programmer was someone who programs, but that
seems to be the last thing programmers do nowadays. Today, the
definition of a programmer is someone who complains unless the
problem being solved has already been solved and whose solution can
be expressed in a single line of code.
So, you might ask, why all of the new complexity if we get no
performance advantage? Because simple approaches don't produce
enough new intellectual property to keep the usual suspects in
business. If you just take the GSM/GPRS/EDGE specification and
change some parameters (like symbol rate and number of timeslots)
you can get a very efficient, flexible system, but you won't have a
much fodder for IEEE papers. Worse yet, you don't generate a lot of
new patents, and with a lot of the patents on 2.xG systems expiring
early in this decade, the NEPs needed a new gravy train. This
complexity also deters small players from building their own UMTS
(Thanks to Paul Wise)
Comments (3 posted)
Keith Packard announces
launch of a new calendar server project. "I started hacking at
Radicale to see how far I could get. I changed the storage code to store
one event per file, then added hooks to use git for change
management. Then, I found a full vcalendar/vcard parsing library in python,
vobject, which I used to replace the ad-hoc parsing code. Finally, I added
support for VCARD entries as well, allowing the system to store both
calendar and contact information.
Comments (20 posted)
For all of you who like to complain that *Office is too big and bloated:
version 1.6 of the venerable GNU "ed" editor has been released. There's
not much in the way of new features, but a number of fixes have been made
and it is now possible to use regular expressions containing NUL
Full Story (comments: 30)
The Apache Software Foundation has announced
the release of Hadoop 1.0. "A foundation of Cloud computing and at
the epicenter of "big data" solutions, Apache Hadoop enables data-intensive
distributed applications to work with thousands of nodes and exabytes of
data. Hadoop enables organizations to more efficiently and cost-effectively
store, process, manage and analyze the growing volumes of data being
created and collected every day. Apache Hadoop connects thousands of
servers to process and analyze data at supercomputing speed.
Comments (2 posted)
Version 0.3 of the Jato Java virtual machine is out. This release includes
a number of performance and portability improvements; it is now claimed to
be able to run JRuby, Clojure, and Eclipse on 32-bit x86 systems; 64-bit
x86, PowerPC, and ARM support are in the works but not yet ready in this
Full Story (comments: none)
Version 1.4 of the Scribus desktop publish system - the first major stable
release in four years - has been announced
1.4 is based on Qt4 and offers much more extensive undo/redo operations,
lots of new import filters, better color management, and a lot more.
"Now that Scribus 1.4.0 has been released, the Scribus Team will
focus on stabilizing the 1.5 development branch, which will comprise
amazing new features like support for PDF/X-1a, PDF/X-4 and PDF/E, Mesh
Gradients, native PDF import, XAR import, a completely rewritten table
implementation, a rewritten text system, and much more.
Comments (5 posted)
Team OpenOffice.org has muddied the office suite waters further with its announcement
longer version [PDF]
) of "White Label Office," a release candidate
consisting of OpenOffice.org 3.3.0 with a number of fixes applied.
"By publishing White Label Office 3.3.1, Team OpenOffice.org is
taking the first step towards a maintenance release for OpenOffice.org
3.3.0. The release candidate is intended to serve as a starting point for
creating the best possible version in collaboration with the users.
Needless to say, folks in the Apache OpenOffice project are not amused
Comments (36 posted)
Newsletters and articles
Comments (none posted)
Jean-Baptiste Queru reflects
on the ups and downs of the Android Open Source Project in 2011.
"Not releasing the Honeycomb source code was catastrophic for the
AOSP community. I had never before received so many angry emails, so many
threats, to the point where I had to take several weeks off at some point
to get away from it. Even today, there's a lot of bitterness left on all
sides. From start to finish, Honeycomb probably cost AOSP anywhere from 6
to 12 months.
Comments (14 posted)
The H covers
of version 1.0 of the Clementine music player. "The
major update adds support for the Spotify and Grooveshark music streaming
services. A Global Search feature has been added that allows users to find
music on their local system or on the internet. Other changes include audio
CD support, and improvements to the settings dialog and album cover
searches, as well as the addition of more transcoder options. A number of
bugs found in the previous versions have also been fixed.
" A brief
found on the project's website. See the changelog
for more detailed information.
Comments (11 posted)
Page editor: Jonathan Corbet
Next page: Announcements>>