The quest for sustainable free/libre non-linear video editors
If you follow conversations around the voting for an Adobe Premiere port to Linux, you might have felt some frustration. Indeed, so far, very little focus has been given to supporting free/libre projects instead. But that bears another question: how can we actually support them?
There seems to be a popular notion that you can throw money at a free/libre software project, and that will do miracles. In the recent case of Natron, it’s supposed to do miracles even in the total absence of developers (irony intended).
So maybe what we are looking for in a project to support is sustainability? Here are some questions that can help us identify a sustainable project.
Does it have a financial plan? Would it use community’s recurring donations or fund itself through education or sales in Windows Store? Does it even accept money through any channel?
Does it scale in terms of human resources by growing a contributive community, or does it have a bus factor of one?
Does it a have an established practice of helping people to do non-coding contributions?
This is something we set out to find out about free/libre non-linear video editors. Please note that we are intentionally leaving out quite a few projects, because no, hell, no.
The history of Cinelerra and its forks is so convoluted that, recently, it took Igor Vladimirsky 4,742 words to summarize it. (Igor, being a Cinelerra-CV contributor, was also most helpful when answering extra questions for this article.)
A much shorter version looks like this.
Cinelerra-HV is the original project by Adam Williams, started in 2000 on top of an earlier project called Broadcast 2000. He works on it in his spare time. Although Michael Collins, co-founder of the project, tried to get this work sponsored, there is little-to-no information if he ever managed to do that. In fact, back in 2008, Adam stated:
Cinelerra is a lousy program to commercialize. It takes so many codecs depending on so many GPL libraries & it’s so heavily dependent on really flaky niche hardware, the support & licensing would be impossible. Good commercial software has few hardware dependencies & can interact with the outside world without requiring massive 3rd party libraries.
When asked, whether the development of Cinelerra was ever funded by any entity, Michael responded:
The credit for the financing of the development of Cinelerra should go entirely to Adam Williams. Trust me, Adam is the hero and everyone else is just hanging on hoping for a better tomorrow, including me. He could have shut it down years ago and no one would be talking about it.
Adam, in return, provided this statement:
Nothing has ever been funded. Video editing on a computer would have to be as big a breakthrough again as it was in 1998. Adobe themselves conceded their last major PC program was Creative Suite in 2010. There’s a lot of demand to add cloud services, mobile apps, and new standards, but it’s not the enabling technology those of us working on the problem 20 years ago were creating.
Adam is pretty much the only developer of Cinelerra-HV, who only merges patches for features/improvements that he personally likes. He also doesn’t have an open source code repository and only publishes changes once a year, as part of a new release. Thus sustainability of Cinelerra-HV remains largely limited.
Cinelerra-GG is a fork by Phyllis Smith and William Morrow aka GoodGuy who initially collaborated with Michael Collins on a Cinelerra-HV based commercial project, but left to work in a more community-focused environment. While earlier Cinelerra-CV project aimed to sync their version to the upstream HV project and thus maintain compatibility, the GG version contains changes that are incompatible with the original HV version. Basically, William rewrites any original code by Adam that prevents new features from being added.
Update (15.12.2018): Phyllis Smith provided this insight:
We are not interested in paid development, but we can assist new developers, and there are programming tasks in our bug tracker which could be done by C programmers.
We think Cinelerra-GG is sustainable, although it is difficult to predict the future. Bill has been working consistently on Cinelerra for 12 years now — the first 8 years for his own personal use, the last 4 trying to get all the mods merged in to a single place with the place being the best version available.
Things could be improved with more help and we have never turned any person or any good mods/patches away. With the advent of the cinelerra-gg.org website, I believe there is now more interest in helping.
The Cinelerra-CVE branch of Cinelerra-CV (where most developments happen) by Einar Rünkaru takes this approach even further by only merging features from HV that he considers interesting:
There shold not be promises about porting something from HV — we port something feature by feature. If feature is interesting, we port. Codebases of the forks are developed too far away from each other — the formal copying from HV does not work any more.
One might say, the three versions of the whole thing are mutually incompatible forks by now.
The Cinelerra-GG website doesn't mention donations at all, and the Cinelerra-CV website only mentions donations in terms of supporting the expenses of running a private server hosting the website and the git repository.
Neither William, nor Einar have expressed interest in sponsored development, they are more interested in patches and new contributors and seem to be willing to help newly arrived developers. Einar, in particular, stated in a private conversation:
Other people who contributed some code to Cinelerra have probably found other tasks. I am left alone maintaining 2 forks [stable Cinelerra-CV branch and unstable Cinelerra-CVE branch]. As there are no new developers or contributors, we can't have any established practice [of mentoring] for them. In the short term, I will continue experimenting with CVE and hope that some new developer who has interest to work on Cinelerra appears.
This semi-forgotten NLE was started as a rewrite of Cinelerra-HV with an intention to remove various limitations that prevented Cinelerra-CV developers from extending the original feature set beyond what Adam Williams deemed useful.
Developers ended up designing everything from scratch, so now there is probably not a single line of original Cinelerra code in Lumiera. Sadly, they also got focused on writing specs and the back-end so much that 11 years later, there is still no usable version of Lumiera.
Much like with Cinelerra, developers are not publicly looking for sponsorship, but encourage developers and non-coding contributors to join.
Blender's Video Sequencer (VSE for short) has a long track record of being the non-linear video editor of choice even for people who have no use for other Blender's features like 3D modeling, sculpting, texturing, and animation. This is mostly because VSE is extremely dependable in terms of stability — so much that people learn to adapt to its rather unusual UX (slightly less unusual with 2.80, thanks to left-click selection and Spacebar for playback).
There have been multiple projects to improve VSE outside of Blender Foundation, including Blender Velvets by Flavio Soares, Blender Power Sequencer by Nathan Lovato et al., all recent work by tintowtin et al. However, none of them deal with how VSE works internally, and therein lies the problem.
In terms of internal design, VSE is substantially outdated and out of sync with what's going on with the rest of Blender. There was an attempt to modernize it in 2013 as part of a Google Summer of Code project, but it was never completed to a point where merging results would be a sensible idea, and by now one would probably have to start it all over again.
We asked Ton Roosendaal, what it would take, financially and organizationally, to get someone to work on VSE fulltime. This is his response:
We calculate average costs of a developer to be 4-5K [euro] per month, that includes overhead (work space, computer, travel, insurances, taxes, etc).
While at that, Ton also suggests that there are a couple of independent open source video editor projects that would deserve support from video editing enthusiasts much more than Blender.
Update (14.12.2018): Richard Antalík becomes VSE maintainer. This part of Blender hasn't had one in years.
This isn't an actively developed project (latest source code changes are from 2016), but it's worth mentioning because the history of Novacut provides an insight into the complexity of creating an actually useful NLE.
Jason Gerard DeRose, Tara Oldfield, and Jeffrey Ballagh started this project around 2009-2010 and used ca. $100K of their personal savings to put into development, then raised another 28K+ via Kickstarter from 802 backers (their first campaign in September 2010 wasn't successful).
Most of the development was happening in dmedia, a distributed object store, because the project was built with the idea that developing a visual story would happen collaboratively over the network, in the real time.
Eventually, the Novacut team ran out of money, significantly cut down on their development time, then ceased development. The front-end never got much further than a prototype with very basic video cutting (without audio support).
It would be easy to call this a death by the back-end overdesign, but the reality is that the back-end does matter and it does take away a lot of time and effort. So much, that today, both Kdenlive, Flowblade, and Shotcut use the same framework, MLT. (So did OpenShot before moving to its own back-end between versions 1.5 and 2.0).
Let's be realistic: $128K is peanuts for a team of several people working full-time, it could only last for a year or two at best. Most NLE projects on Linux are at least a decade old, and even now they are considered not stable or sophisticated enough by quite a few users.
So Novacut team arguably delivered what they promised in the campaign:
If we meet our funding goal, we're confident we can deliver ground breaking collaborative editing, although likely with just a few of our planned editing features implemented.
They just didn't manage to do the rest.
Kdenlive is often called #1 alternative to proprietary options like Adobe Premiere, Sony Vegas Pro etc. (that is, among people who are not die-hard fans of other free/libre NLEs).
In 2012, the team attempted to do 2 months of paid development and raised $7,419 from 230 backers on Indiegogo to fund a major refactoring work by Till Theato. Merging the changes didn't exactly go as planned, and changes started sipping in only in 2015.
Currently, the project is actively maintained, and upcoming 19.04 release is going to be a major milestone, bringing more stability and new editing features. Since moving to KDE's infrastructure a few years ago, Kdenlive follows KDE's release schedule, which means three updates a year with bugfix follow-ups.
The team participates at KDE's annual hacking sprints in Randa (Switzerland) which are sponsored by KDE e.V. Beyond that, there is no direct sponsorship for the project presently, and no one in the team is ready to do paid development. Here is what Vincent Pinon stated last year :
Getting money is one problem, but spending it is another ;) We don't have anyone who already masters Kdenlive code, and would have available time to work on it for a paid job. This is a thing to think of and discuss with other KDE projects who do so...
Flowblade is one of the more recent projects (if 9 years can be considered recent). It's a rather capable tool, however, Janne Liljeblad is pretty much the only developer. Most other contributions are translations and fixes to accommodate for translations.
When asked about accepting donations in the bug tracker earlier this year , Janne replied:
This has been a task on my private todo list forever, but I have not been interested in pursuing it so far, basically it has been my view that the money received will not be enough to justify the effort. The one time campaigns by OpenShot, Pitivi and that one third thing were all disappointments and have probably shut down that road, but the PayPal button could produce some revenue.
Dan Dennedy (re)started the Shotcut project in 2011 as a complete rewrite of an earlier project by Charlie Yates, having already created Kino (with Arne Schirmacher), MLT (with Charlie), as well as having contributed to Kdenlive. Back when we interviewed him in 2012, he said:
Basically, I missed the type of work I did with Kino (GUI app, not just libs), and I want to give the world an open-source, cross-platform video editor.
Shotcut looked a lot like somewhat modernized Kino in the beginning but gained multitrack support, video and audio effects, more editing features etc. The software is actively maintained and gets monthly releases. We spoke to Dan about the sustainability of the project, and here is what he had to say:
I now make enough money from advertisement on the web site, forum, and YouTube to work on MLT and Shotcut full time. It is not making me rich, but it is close enough to my day job salary. My day job had become a lot of work, and Shotcut was facing some neglect a few years prior as a result. Unfortunately, it does not get much contribution even though I publish SDKs and instructions on the web site. There is one other person who contributes quite a bit, and he receives donations made using the donate link on the Shotcut downloads page.
I believe the revenue is sustainable for the near future. However, I am not confident long term. Since going full time May 2018, I have given Shotcut my full attention to resurrect it from neglect, to keep it thriving while only casually exploring future options. Maybe I can hire someone in 2020 to help with Shotcut while I try to expand in new directions.
OpenShot was the third NLE project to use Kickstarter for funding its development. In March 2013, Jonathan Thomas raised $45K from 1,463 backers to do a complete back-end and UI rewrite. Being the sole developer in the project, he managed to release the first beta only in early 2016, as opposed to estimated time of arrival in Nov/Dec 2013.
Jonathan runs a small business, OpenShot Studios LLC, where he creates custom websites and Android apps for clients (OpenShot is advertized in the portfolio). He also created OpenShot Video Editing Cloud API and put it on AWS Marketplace (we don't know how much revenue it generates for him yet).
Update (15.12.2018): Jonathan clarified on his future plans for the project:
We have a few consistent developers and contributors, but I am going to spend a lot of energy in 2019 around formalizing our process and team workflow, with the goal of making it easier to contribute. 2019 will also be the first year where I am 100% full time working on OpenShot, so hold on tight! I have a ton of big things planned!
Pitivi is one of the oldest NLE projects on Linux. Started in 2004, it's older than both Kdenlive and OpenShot, both of which feel like they've been around since the dawn of times.
The team is currently working towards v1.0 that's likely to be out early next year. Most of the paid development in the project happened thanks to participation at the Google Summer of Code project (the current most active developer is a former student).
In 2014, the team tried to raise €100K euro to sponsor completion of Pitivi 1.0 and development of new features. They only managed to collect a little over €25K despite providing estimation for various goals. They were supported by 889 users.
Today, Pitivi developers confirm that they could use donations to organize hackfest and even hire interns, but they don't seem to have a plan for paid development.
Update (15.12.2018): Alexandru Băluț and Thibault Sanier responded that the team is looking into improving the sustainability of the project:
Besides wrapping up Pitivi 1.0, sustainability is the next important thing for us, as it turned out of the discussions at the recent GStreamer conference. Medium term we're doing things that work but will probably not make a huge difference, to keep the buzz going. We'll try our luck with donations, recurrent donations, small fundraisers for specific features. GSoC mentor stipends also add up, but they are a side-effect. Long term is of course what matters. We are working on finding business leads through an investment. We can't give more details for now
So what can you do with all this information? It really depends on where you stand with regards to taking action.
If your view is that only a business-like entity can survive, then we are talking about OpenShot, Shotcut, and Blender being the fittest in the game of evolution. All of them are likely to continue existing and evolving one way or another.
If you are of an opinion that throwing money into development on a regular basis is the way to ensure the future of a project, OpenShot, Pitivi, and Blender are where your dimes will be used most efficiently.
Do note, please, that in both scenarios above you should keep in mind that VSE will always come second as a development priority in Blender.
Finally, if you think you should ensure the prosperity of free/libre NLEs on Linux by rolling up your sleeves and contributing by submitting patches or being helpful in any other way, pretty much any video editor project out there will greet you with arms wide open — just pick your preferred NLE.
Of course, there's the easy way too: just pay for either DaVinci Resolve or Lightworks for Linux and be done with free/libre NLEs on a free/libre platform.
This leaves a few conversations open. One of them is how long other projects can exist and evolve at a perceivably fast pace without donations or a solid plan for paid/sponsored development. So far, it looks like once you've been in this game for at least a few years, you stay there and even manage to get work done. But nothing is ever guaranteed.
And the other conversation is whether funding development of any free/libre NLE is realistic at all. Consider this. The most any crowdfunding effort got in terms of the number of supporters is OpenShot and its 1,463 backers. However, at the minute of publishing this article, 6,148 people voted for the Adobe Premiere port to Linux (voted, not paid for, mind you).
Essentially, we are looking at a tiny fraction of the overall market. But it's up to us to take action and figure out how far we can actually go.