Kdenlive: an open-source video editor
kdenlive.org
kdenlive.org
I only run Linux, so I'm not sure what I would've done without Kdenlive since I don't see what else I could have possibly used.
Huge thanks to everyone who contributes to the project!
It is simple, so am I.
I use Arch Linux as my OS.
The deb from Ubuntu 18.10 had unmet dependencies, which is not a surprise given how interwoven it is with gstreamer.
Finally, I installed the most recent version via flatpak and it rendered fine.
;)
[1]: https://shotcut.org/
They've done a fantastic job of fixing that.
Lightworks should run on Linux. It was used to edit Pulp Fiction, The Wolf of Wall Street and other blockbuster movies.
I used it on Windows a while ago and my basic requirements were met, despite the free version having some limitations (like export resolution and formats).
Davinci Resolve seriously might be the biggest Video Editor on all three platforms int he next three years. Highly recommended.
When I use Linux for Audio and Video I DON'T update the computer and I don't have anything else really on that image. I use a second image for non-Audio/Video production. I did the same thing with Windows for decades.
If you can get it working, the thing's a glorious beast.
Once I got opencl installed and rebooted the machine, I was able to start Resolve and get into the main interface. It couldn't find a GPU though ("clinfo" showed one), I'm guessing that it really,truly does need an nVidia card. I brought home a card from work to try, just to see if it would make a difference, but someone put a AMD card into the box of a nVidia card.
I tried Resolve on my wife's Win10 laptop with Intel GPU, and it seemed to install and work no problem. Since I'm basically going to be using the machine for just editing when I'm editing, I may just set it up with Windows and call it done.
From some of the introductions I've watched, it looks like Resolve is the software I want. Lightworks has worked well, but Resolve looks like a space age rocket ship. I guess we will see...
While rendering those videos I remember just being dumbstruck at how much you could really accomplish in Blender. You didn't really need much more even to be a professional video editor.
The trade-off that never seems to go away, though, with any software, is that in order to avoid all "bugs" [in common parlance, just things the software won't let you do], you need to know a tiny bit about video and audio codecs, the basics of how compression works, frame rates, and what kind of file at which dimensions your customer / delivery target needs. This eliminates like 99% of the quality problems that people will report with their exported videos: Doesn't play in my client's video software, etc.
I wish I had tried it earlier but I think the KDE dependencies put me off, fortunately it's also released as an AppImage and runs perfectly on my Ubuntu desktop.
It seems iMovie latest removed the ability to do that.
Do you guys have any recommendations for Mac? Free is better.
I create videos of product reviews for my followers
I need it to create video ads for products, mostly Facebook ads/Instagram
It's one of the reasons why I continue to stick with Windows. Kdenlive feels like it's 10 years behind Camtasia in terms of UI polish and being able to accomplish simple tasks like put a little tooltip overlay on the screen and make it look reasonable.
I think a lot of people who say kdenlive is good haven't tried Camtasia, or have created enough videos to really figure out kdenlive's shortcomings. I'm coming at this from the POV of creating 400 videos and have spent over 1,000 hours hardcore editing these videos.
I don't want to tear it down too much because it's one of the better screencast video editing tools on Linux but if your profession involves creating screencast videos, you're going to be majorly disappointed compared to Camtasia (or Screenflow). Enough to where you'll probably abort the idea of using Linux even if you're primarily a developer.
I wish I could throw money at the kdenlive devs and wake up tomorrow with a video editor that was as good as Camtasia.
OBS lets you combine multiple sources easily including adding static and dynamic graphical overlays and is basically the what all of the streamers are using on all of the three platforms Linux, macOS and Windows these days, and aside from being able to stream live to Twitch etc it is also perfect for offline recording.
However, OBS is not a video editor, so if your use-case is to first record your screen and to then record the commentary afterwards rather than commenting while you are screenrecording then OBS is probably not what you want.
But Camtasia is both a recorder and editor and it seamlessly integrates the 2. The main pros of Camtasia is that it's super optimized for editing screencast style videos (tooltip overlays, text, zooms / pans, really REALLY nice animations on any objects by just dragging a few sliders, etc.). The editing part is where kdenlive is way behind IMO, but that's where you spend most of your time.
A vMix Call equivalent that transparently wires into the audio board would also be really nice, but I don't wanna get greedy.
Fair enough but audio is a completely different animal that can be tamed in a number of ways:
1. You can buy a hardware USB audio interface + mixer for ~$200-400ish total which can do compression, basic EQs, noise gate, has multiple inputs, etc. and now that becomes your final source of audio so you don't need to process it with software. It's just a mic from your computer's point of view.
2. OBS has VST plugin support, so you can directly use hundreds of free VSTs to do various effects in real time without needing to mess around with audio redirection. No hardware needed except for your mic.
3. You can use REAPER (free but you should register it, sort of like Sublime Text) which is a DAW. From here you can use various VSTs and then use Jack (or comparable software) to redirect REAPER's output as input to another app (such as OBS). This takes a bit of initial set up but technically would work with any app, not just OBS and this is what I used to do before going with #1 because I didn't want to run all of these apps every time I wanted to start recording.
In the end, you can get fantastic sounding audio with any of the above approaches and with #1 and #3 you can do it with any application where as with #2 you're limited to OBS only (which isn't a bad thing if you spend your time streaming and primarily recording with it anyways).
#1: that USB mixer acts as a mixer only for inputs that you can plug into that mixer. That may sound obvious. It becomes concretely painful when you have video inputs that also have sound attached. For HDMI captures, you might have an audio takeoff or you might use an HDMI extractor, which is...fine, such as it is, but suboptimal. If you use a professional stack--my video stack is SDI-based--this is completely inadequate. If you use NDI for video transmission over your network (and my stack has a switch in it, 24 gigabit and 2 10G SFP+, in order to take between six and ten NDI streams into the video mixer), you're boned; the only option that OBS offers, and it's a bad one, is to replace your cue/monitor output (assuming you have an extra audio out in your video mixer) with the NDI output. This means you can't practically mix NDI stream output that isn't currently on-screen because you only get one cue.
2. This isn't a solution because the OBS mixer is deficient. Modifying levels on the fly is impractical because they're just local UI widgets that lack the usual affordances for mouse-driven audio control. And this could be fixed with hardware but far as I am aware (this one might have changed since I last used it, OBS is now just an NDI slave for me), you can't connect a MIDI slider controller to the OBS mixer in the first place. You definitely can't send signals back to a MIDI controller to enable the use of motorized faders, which are one of those "I can't go back, I won't go back!" things now that I have them in my stack.
3. Similar problem to #1, only in software. Redirection turns into a virtual patch bay and you end up having to figure out how to run things back and forth. And the tooling on both ends is very rarely automatable, so you'd better hope you don't lose your setup.
The best option, and it also gets you screwed over because of UI affordances, is to use obs-asio with bassasio (and yeah, this means you're using Windows, but you probably should be anyway because #3 is better with a more serious DAW option, like Ableton Live, as your audio backend). You can pull every channel off your interface and expose them in OBS and you can squint at Qt sliders to your heart's content.
Personally, I use vMix instead and have good iPad-based applications that I don't have to write myself (like I said elsewhere in this thread, I wrote this[1] to control OBS remotely, it's not nearly as good as the stuff already available for vMix!) for both audio and video switching and I can, and do, wire up MIDI surfaces for physical controls (motorized faders, etc.).
Don't get me wrong: OBS is fine if your needs are small. I got pretty far with it. When your needs are no longer small, it turns into a Jenga stack of hacks. I hang out in the OBS Discord, I think the team is very talented, but the software has trouble scaling up even to my relatively modest needs compared to commercial solutions (vMix and NewTek VT being the best options in that space, though VT is gonna cost you both arms and both legs).
Had I infinite time, I would go take a chainsaw to the entire audio stack in OBS, but...I don't, and vMix already does it capital-R Right: multiple ASIO devices, first-class mixer inputs for everything from audio interfaces to cameras to NDI feeds to WebRTC calls (seriously, vMix Call is great). And it comes with a pervasive automation suite built around multiple kinds of input devices--DirectInput joysticks/controllers, keyboard shortcuts, MIDI signals (and not just notes, but faders etc.)--with signal-returns to enable stuff like motorized fader boards (which have gotten super cheap, too). Given all that, and the lack of having to do any of it myself? $700 is cheap. ;)
[1] - https://bit.ly/buymyapp
I'm coming at from a single person trying to record screencast videos in a controlled environment.
And OBS is in my standard kit. It's just fine as an NDI satellite. But also, for what you describe, there's a $60 1080p vMix package...and it's got most of the bells and whistles I described, too. Automation and control are easy to overlook until you don't have them!
I open a recording program (either Camtasia or OBS depending on what I'm doing), hit record and talk. Then I hit stop.
That's it. At this point my audio is leveled and comes out sounding good since all of the processing happens in hardware in real time. It's also sync'd with the video.
Editing is where kdenlive falls apart which is really what my reply was about.
I want to see kdenlive overtake everything, but it's simply not there and given its progress over the years (I've been following pretty much every open source and closed source screencast oriented video editor for years) I don't see it ever getting there.
Not a single person replied positively even though the post wasn't hostile in the slightest.
life changing when I started using that for multitrack recordings..
Ardour works fine with an Audio Technica AT2100 USB (I tried it and can verify that it uses the standard USB host drivers without complaint) for both I/O and works out-of-the-box with most of my audio interfaces and an XLR mic.
My 2¢: Based on my experience with podcasters, I wouldn't recommend a Snowball for beginners doing spoken-word recordings. This is a "worse is better" scenario where a cheap dynamic will give you better out-of-the-box results than a condenser mic like the Snowball, in part because it forces speakers to get closer to the mic (increasing the ratio of signal to ambient noise), and in part because dynamics usually can't record anything over ~15 kHz (spoken word tops out around 8 kHz).
A Snowball is fine for folks savvy enough to use an EQ to roll off everything over, say, 10 kHz during production.
I alternate between an sE V7 dynamic, a Blue Encore 200 "active dynamic", and an AT2035 at my desk for recordings. The Blue is something I'm just testing but the other two microphones are great for different things. (Anyone who hasn't checked out a sE V7 should; for my money it's the best thing in the "SM58 range" of dynamics, with a really clear sound for a dynamic. And the damned thing won't roll off your desk if you have it handheld...)
You get around this be setting JACK to use the snowball device for capture only (and by doing so you are using its clock) and then attach whatever device you are using for monitoring as a JACK client using zita_j2a from the zita-ajbridge tools package.
I do know that programs like Adobe Audition on Windows offer that in a plug-and-play way.
if price is a problem there are several other environments on Linux for composing and producing. even something as simple as lmms can be made very powerful. you can also build drum patterns with hydrogen. you can use stuff like openmuse to score music, you can download some decent soundfonts (fatboy is good, timbres of heaven is amazing) to get more juice out of your midi, etc. then if you're into generative music there's stuff like supercollider and puredata, or dsl's like sonic pi for ruby and overtone for clojure that can drive the supercollider synth. or there's even languages like chuck, all sorts of interesting possibilities for audio.
anyway i guess my thesis point is there is always a way to get it done, as long as you keep at it (which it seems like you indeed have been doing so ^5)
But it's not an Ardour replacement. Some of their features overlap, but you can't do things like record long clips of audio in Renoise, while that's a core feature in Ardour.
Also, the workflow for Renoise is just not something most people are used to or will want to spend the time adjusting to -- though I certainly encourage them to give it a chance, as it's great alternative to the mouse-oriented piano roll workflow in traditional DAWs.
So I think for most Linux users, Ardour is going to probably be the DAW that they're most used to and the one most appropriate for recording and manipulating audio for screencasts and videos, with Renoise unfortunately relegated just to music production by tracker fans.
Kdenlive I haven't tried yet, but is on my short list to try.
I've been using LightWorks under Linux and it works pretty well. I mean, I have nothing else to compare it to, but I'm doing 4K30fps video, and a friend of mine tells me that his Macbook Pro with Adobe tools isn't up to the task. I'm running on a 5 year old PC I had in the closet (870K CPU, 16GB RAM, Radeon, SSD), and it works fairly well.
The only downside of Lightworks so far has been that to export to Youtube at 4K you have to "subscribe", you can't just buy the software. You can do your editing and play with it, and export 720p video.
But to export 4K, you have to either pay $25/mo or $170/year. I wasn't sure how much I'd like editing video, so I didn't want to pay $170. so I did the $25 for a month. I'm really enjoying it, but I kind of wish I could just pay $200-$300 and be done with it. Which I realize isn't rational, because that'll buy a lot of $25 months.
Other ones I want to try: Kdenlive, Davinci Resolve (tried it and it segfaulted, $170/year), Cinelerra (looks kind of crappy), Flowblade. Pitivi looks to be abandoned. Maybe I should try Blender, but I don't want to do any 3D stuff, don't want to get sidetracked.
There's also an online NL video editor that my daughter has used for school: wevideo. She showed it to me after seeing me running LightWorks and recognizing the workflow. I'm dealing with 75-200GB though, don't think that's going to work for online. Smaller video will probably work, my daughter uses it to voice-over a series of photos turned into video.
There are a few rough edges, but Lightworks has mostly let me, as a neophyte video guy, to make great videos. Took me an hour to import an MP3 this weekend though.
I think I'm sold on Davinci Resolve though. It's pretty sweet, extremely full featured, free for most uses and reasonably priced for the full kit ($300 IIRC. That adds the ability to do multi-GPU and have multiple users collaborating, plus 6K and 8K editing).
I had to upgrade my GPU though. My old one had 2GB and Resolve was like "Oh, that's ADORABLE!" It was a 5-6 year old computer though. I also put it under Windows, it just seems to run better there. I had battled for an hour to get to the point where it stopped segfaulting under Linux and started complaining about not enough video RAM. But now that Linux install is janked, something related to the AMD GPU packages. Probably not TOO hard a fix, since it was all .debs.
The problem I had when searching was that there was a ton of choice under Linux (my short list was 10), but researching them from that point is not easy. They pretty much all show the same screenshot (media, clip monitor, timeline, timeline monitor). But how do you tell: Is it reliable? Does it have the effects I'll need? Is it serious or a toy?
Normally, I'd have heard of some things and at least know a reputation for something in the field. But video editing is so foreign to my domain, that I've maybe heard one thing about the Linux options over the last decade. Mostly you hear about Final Cut vs. Adobe Premier.
I’ve also seen that black magic’s editor is available for Linux, but it is an enormous package...
Not once did I have to resort to Google or documentation for help -- it was all fairly intuitive. I want to reiterate that I wasn't doing anything fancy, just trimming bits and pieces off clips here and there.
I'm actually not aware of another one that uses proxy files on Linux, and with even cell phones shooting at 4K any video editor that doesn't is unusable in most cases.
DaVinci Resolve is also extremely good (probably even better than Kdenlive), but it's a commercial software although they do have a free unlimited version.
FYI, Blender can user proxy files, although I don't think there is an option to automatically create the proxies when you add a video clip.
They aren't in the same league. Resolve is used in Hollywood productions. The license is cheap, 300$, If you wanted to buy it, the free version works great. The big hardware console for Resolve is 30k.
The thing is, I really want to like Kdenlive because it is really close to being able to do what I want, but I just felt that every step of the way I had to be very creative just to figure out a workflow that would work well. I eventually moved to the video editor in Blender which has a lot fewer features, but the workflow is really exceptional. I figured that what time I saved struggling with Kdenlive I could spend using other tools for features that aren't in Blender.
I don't do any video editing any more (I decided video isn't really a medium that I enjoy being expressive in), so take my comments with a grain of salt.
Second that. Had great and worse times with Kdenlive.
My general feeling: KDE software tends to be more unstable (i.e. crash or freeze more frequently) then GNOME software. But they have in general more features. Just compare the KDE office suite, Calligra, against OpenOffice (which is however not related to GNOME). It is amazing at what speed KDE software moves, in general.
The presentation and spreadsheet do need some work, though.
> Second that. Had great and worse times with Kdenlive.
Agree also. Two elements of hope regarding stability : - The timeline refactoring is nearly completed - A nice fuzzing system has been added, which should really help : see https://kdenlive.org/en/2019/03/inside-kdenlive-how-to-fuzz-...
Freedom of updating and installing newer versions of software without waiting for maintainer/distro half of year to build new package (outdated next week because it's STABLE package).
This is main problem of any linux desktop initiative - linux users get newer gimp after windows users.
Hope snaps/flatpack/appimage solve this problem in future (snap versions of apps look ugly now).
Then use Arch, or some other rolling release distro.
> Hope snaps/flatpack/appimage solve this problem in future
These are terrible solutions, not least due to the theming issue. Rolling release is the real solution.
Windows + chocolatey or macos + brew (cask) is better than any repository with dependencies on linux.
Also, do not want Arch or any <distro-name> - I just want LINUX (or Desktop Linux).
One predictable system, instead of many kinds/variants of basically the same programs - one system to hate, one system to deal with. (Arch wiki is good though)
What do you mean ? How can you have Linux without using a distro ? Are you saying there should be one official distro instead of many ?
And this is why I think snap/flatpack/etc model ("system" in system) is real solution for linux desktop.
Valve's Steam client already achieved same thing: provided one "sdk" for apps, one runtime (that developer can rely on) - because you can't have one main linux organisation like microsoft and apple.
Linux is kernel - I talk about good alternative for macOS/Windows, alternative desktop system, that can do much better than other two.
[Awaits inevitable flood of comments along the lines of "I tried x on Ubuntu and it didn't work but it worked fine on y"]
Using "stable" distros makes little sense for other than complete newbies and enterprise-like critical usage and such imo. If you don't need bleeding edge stuff they are a good choice though, but for average users modern rolling releases are definitely stable enough.
The freedom is there.
I just replaced my GPU today. I ran `sudo pacman -S nvidia`, rebooted, and I was good to go.
As an aside: Arch isn't as "bleeding edge" as some people would have you believe. They have a testing repository, so packages get vetted before making it into the main package repositories. They often wait for a point-release of some software (like Gnome Shell) to help ensure that only stable versions make it in.
Plus, the Arch wiki is the most useful Linux resource I've found. I reference it even when I'm using an Ubuntu machine.
New version of Gimp available for all windows users at day one after release is not bleeding edge - it's just normal new release.
Meaning of "bleeding edge" and "stable" became strangely twisted in linux community.
New version of browser is not bleeding edge - it just normal life cycle of modern software (year 2019).
You have this freedom. It is known as "compiling from sources". In fact GNU/Linux is the only big OS/ecosystem where you have this freedom.
But like all other freedoms, you have to work hard for it.
On windows you just can download the same program and use it without any compiling. And you can freely downgrade it if you want to. No compiling, no messing with repos, makeinstall, and randomly corrupted system.
For a free (as in free beer) software, it is really good. It support a wide variety of feature that can be extended through plug-ins, has a decent UX (even though some stuff required me to look in the documentation) and has a decent documentation (although some of it is outdated).
Overall, if you are looking for a free video editor, my experience is that KDenLive is the best you will get (as of 2019). It has some obvious drawback (last time I used it, it was crashing fairly frequently, but a lot has been fixed), but if you are looking to do some amateur/semi-pro video-editing, KDenLive should do it for you.
I don't know on Linux, but for Windows I also had to search around and manually install some codecs because otherwise it wouldn't render to a given format (I believe it was mp4).
Then I've found OpenShot, that was enough for my requirements (https://www.openshot.org/). It has a Linux version. Source code here: https://github.com/OpenShot/openshot-qt
I tackled various realm of creative production, namely motion graphics and electronic music. This all started when I was using Windows and pirated program of famous tools like AE, but I was able to get around in Linux with FOSS tools. I use LMMS like a pro. I even used GIMP on Windows.
This was NOT the case with video production. I was making a simple lyrics video on kdenlive, but I had to scrap the project because every time I tried to add a little bit of effect, it was half-baked or simply impossible. I cannot change size of the video without having a major headache on anchor points. I cannot "crop" a content right on kdenlive. The problem? Kdenlive is actually the most feature-rich video editor I ever tried on Linux.
If you only do light editing it may work for you, but ONLY light editing. You can't have custom stuff going on as video creation tool is basically not there. I also use it as a subtitle tool... which basically sums up what your expectation should be.
It just seems like Linux cannot currently guarantee anything about the stability or even existence of GPU-acceleration on an arbitrary machine. So to start with that and then smear the software across multiple LTS versions of multiple distros seems like a perfect recipe for our current reality.
Could maybe even build a funding model off of that. Compare a fundraiser trying to gain general stability for a buggy video editor to a fundraiser for getting "Paradise Video Editor" to extend its rock solid UX to a 2nd piece of hardware. I bet the maker of that hardware would find value in such support. :)
If you mean specialized GPU video encoding / decoding, there is VAAPI, which is also pretty much supported on all GPUs and distros. But selection of particular codecs depends on the hardware.
For example, let's say Ubuntu 18.04 LTS on x86_64 with GPU... (dunno, choose a decent one that has a respectable open source driver).
Here's my question-- how much more stable would such a video editor be than the extant ones that ship with general Linux support?
I'm asking because I see lots of reports that so-and-so crashes because the user installed from the ancient Debian package, or the Nvidia driver is garbage, or some library doesn't play well on some distro/arch/distro-arch-combo. I'm just curious how much of the eternal bugginess of Linux video editors can be ascribed to those problems vs. the more narrow problem of designing fast-and-stable video editing software.
Since this will be a dealbreaker for many users, they should know this upfront.
All respect to open source software, but for many projects using Kdenlive is just not an option as of yet.
I've definitely grown cynical but I will try out kdenlive and see if it has matured. Shotcut has been my Linux go-to for several years now.
Multiple timelines, not multiple tracks?
There's are so many other large, complex applications on Linux, but I've never in my life experienced or seen as many reports of crashes as I have with kdenlive.
Is stability just not a priority for its developers?
Its not viable if you ignore all the ways it is viable.
>open source is not that visible in the graphics/animation space
OpenEXR, Alembic, USD, OpenImageIO, OCIO, Open Shader Language, OpenSubDiv, etc, etc.
I wonder if they deal with closed captioning, titling, synching with a VTR, multicamera, drivers for video I/O cards, proper file parsing (depending on things like FFMPEG is not such a great idea), etc.
Thanks again for such great software!
Many in the free software movement shun commercial software. To make things worse, GitHub recently got aquired by one of free softwares old enemies.
Personally I've used and even recommended GitHub before Gitlab became available and I don't shun them or even Microsoft, but I also see where they might be coming from.
If you could introduce a single point of failure merely by using git with a proprietary remote service, then you are using git wrong.
That said: Kdenlive is, by far, the best option OSS available on Linux. openshot is trying, but is pretty unstable and not reliable for big videos.
kdenlive is replacement for windows movie maker, not professional video editing.
KDEnlive, as I said in another post, feels like 1995's software. But it works for what it's suppose to do.