Hey!
I've read Harald's e-mail and I thought maybe I'll share my Kdenlvie
experience with you as well.
I've been a heavy Kdenlive user for about 18 months, when I switched
from Blender VSE, seeing it's not being worked on and I cannot expect
the bugs that cripple my workflow to be ever fixed.
About two months ago I've upgraded from 16 to 32 GB of RAM to help
myself with editing videos in Kdenlive.
A month later I immediately gave up using Kdeinlive, after I've
discovered Olive.
Olive is a very young project (about 20 months in development at this
point). And it made me realize how painful it is to do my work with
Kdenlive (or Blender VSE to be fair).
Olive uses OpenGL for all image processing. It uses GLSL shaders for
effects and compositing. I was even able to create a Despill effect
myself to get a better Green Screen compositing.
That being said I have a feeling that my trouble with Kdenlive will
not
lead to any progress, and I prefer to have issues with Olive that has
a
promising start and a community I can get in touch with easily.
Also - it gives me an order of magnitude better experience than
anything
else right away, so why shouldn't iI use it?
A minro problem was for the past year (or something?) I couldn't log
into my KDE account, because they enforce using your real name as
login,
and I have diacritics in my name, I also can't use a password that I'd
want so I had once to contact an admit to even log in, and I just
don't
have the tim deal with that.. That made me feel like the KDE community
is making it harder for me to communicate.
Why can't I use a login and password of my choosing? I find that a
strange decision and I'm just out of the KDE forums until that's
changed.
I've stopped reporting bugs, also knowing that the refactoring is a
priority.
I think Kdenlive has some great functions, but is also a huge mess.
There are three groups of effects based on keyframing (no keyframing,
a
table keyframing, and a timeline keyframing). The effects are
affecting
the image in ways they shouldn't - transform and wipe will change the
color balance, whic is visible on the RGB parade - I show that in my
rant video). The performance is abysmal if you want to do any
compositing (which I do a lot).
If you want to just "cut the tape" and don't need any effects or
compositing - Kdenlvie may be acceptable. But not if you want to do
anything more complex.
There's a UV Mapping compositor, that's useless, because the whole
pipeline is only 8-bit. Which gives you 256x256 pixels addressable.
GPU support is non-functional and crashes Kdenlive every time for me.
Multithrraded rendering or even playback is broken and gives me white
frames every time.
My recent two serious projects got random audio sync issues and audio
clicks (randomly different with each render). That was the last straw
for me, and I snapped.
Here are the two last videos I've made with Kdenlive:
https://www.youtube.com/watch?v=YpMP8uGGpzI
https://www.youtube.com/watch?v=Ulks8T6z6BU
The issues created by Kdenlive in my work has cost me too much time
and
beard hair. Even my wife is relived that I'm not using it any more.
I bless the Heavens for Olive, as I can retain my sanity and make
videos
that I love.
I think that as long as Kdenlive is coupled with MLT - it'll be
inhereting it's problems.
I don't know how that looks on the inside, but I guess Kdenlive is
probably doing a lot of work just to workaround MLT's limitations and
quirks.
Also - Kdenlvie has a long legacy, which is holding it back.
Olvie has non of that - and kowing it's developers and community - it
migth become a tool even better than some of the industry standards
like
Adobe Premiere (which also has a lot of legacy and quirks).
Anyway - I've expressed my issues and frustration in a video:
https://youtu.be/ym1brc2OcYQ
After I published this, I one of the developers of Kdelvie cam to
Olive
Disocrd and we got in touch talking about the issues.
I've send him my latest projects so he can investigate the audio
clicks/desync issue (I couldn't reproduce it in a simmle project).
I've decided to take that video down after a week. As I though maybe
it's sending a needlessly negative message, but since I've decidd to
make it public again, as there'struth there that has to be said, even
if
it's not nice to listen to.
I hope you can undesrtand that.
I hope Kdenlive can be reborn, but I am afraid a lot of things just
need
to be replaced entirely if it's going to ever become a truly modern
video editor.
In my honest opition (being completely unaware of the Kdenlive's
internal structure) it'd be better to just go back to the drawing
board
and design an entirely new achitecture taking into account all
experience of Kdenlive.
That experience is probably the most valuable asset you guys have.
I don't think you're gonna do that, as you've invested so much time
and
effort into it. I'm thinking about the "sunken cost fallacy".
I'm a really sorry to say that - but I don't elbie any more that time
spent on improving Kdenlvie is a time well spent.
The end product is no good for me, and Olive is already allowing me to
do the same work faster, better and suicidal-thoughts free.
Some more facts from my experience of the past months comparing
Kdenlive
and Olive:
Kdenlive is compositng in a single CPU thread (possibly in multiple if
you odn't use any effects - so let's ignore that), while my Ryzen 7
1700
CPU is at 10% load, and my Nvidia GTX 1060 GPU is completely idle.
When I render vidoes with Olive, my CPU is at 60-70% with all threads
active, and my GPU is at 30-40% load.
When I work with Olive - I can finally see in realtime what I'm doing,
with proper compositing, even without using Proxy.
In Kdenlive if I use 2-3 layers at a time and a few transporm effects
I'm down to 3 FPS and I'm guessing what the video will look like. And
I
use proxy in Kdenlive or it's down to 1FPS.
Olive (in alpha stage) has solid crash recovery (I never lost any work
from a crash yet) and crashes about 5-8 times less than Kdenlive 18 on
my system.
I've been begging for crash recover in Kdenlive to be implemented, and
I
had to give up my parallel rendering script when it was added, because
that broke.
Yes, I've coded an external tool to help me render Kdenlive projects
faster. but a new version of Kdnelive changed something and I couldn't
get it to work evr again.
Still - loosing 30 minutes of work is worse than rendering a video all
night, so I've picked my poison.
I hoped that maybe this will be implemented in Kdenlive itself at some
point:
https://github.com/unfa/kdenlive-multirender
Now I don't need to hack together stuff like that because Olive just
utilizes my hardware. It renders my projects in near-realitime.
I have nothing but love for you guys. The frustration is only for the
issues I experienced and how they made my work needlessly painful.
Experience working on my last Kdenlvie video made me consider using
some
proprietary software - for the first time in my life.
That's how bad that experience was. And I know I've checked all the
other open-source NLAs too already. They are all like this. Stuck in
the
90s.
Only Olive isn't.
I also felt I need to share my perspective publicly (in that rant
video), because many people regard Kdenlive as "open-source state of
the
art NLA" and that's just not true and (in my opinion) very harmful to
the image of what the open-source commnity can create.
I saw The Linux Gamer and Wendell benchmark Kdenlive once on some
insane
hardware with multiple GPUs and I was grinding my teeth - because
Kdenlive can't use that hardware at all and there's no point even
having
a GPU for Kdenlive.
Please forgive me if I have hurt you with this e-mail. That was not my
intention.
I wanted to share this, becasue I hope you can benefit from my story.
I wish you all the best,
- unfa
