It's great to see that Kdenlive is getting back on its feet and development is picking up a little (thanks for all your work on the bug fixes & new releases guys!).

I thought I would dig up my number one feature request one more time, to explain why I think it is important and to see what others think of it, or if there are current workarounds/workflow improvements available right now to make things easier. I even made a video to clarify things!

Relevant feature request in bugtracker.

Currently when working on event videos, documentary, or multi-camera interviews, I spend hours just chopping down multiple synched tracks to develop a concise story.
In Kdenlive I need to go through 11 separate steps to make a single multitrack edit. Back when I used Final Cut, I could do this in 3 steps.
I've shown the difference here in a basic animation:

Improving Kdenlive Workflow: Ripple Edits

This feature has different names in different programs, but the concept is simple:

Mark In and Out, then hit a single shortcut, which deletes all footage In-to-Out in unlocked tracks, slides everything else back to fill the gap, and puts the playhead on the In point.

I'm not suggesting Kdenlive follow the proprietary programs on everything, but fixing this issue would carve hours off a large-scale project.
When I tell other pro editors that I use F/LOSS, they're intrigued, one or two have been almost ready to give it a go, but when I show them the editing workflow in Kdenlive they laugh.
This arduous process is simply not compatible with industry needs (which are: fast, fast, and fast).
I can deal with it for now, because I usually work alone and use Kdenlive mostly for ideological reasons, but it's incredibly annoying.

Long-term solutions?

What is required to get something like this implemented in Kdenlive (short of me learning to code to a high level?)
What about a bounty or similar?
I know that in the past year it has just been an issue of there not being enough developers and the difficulty of untangling the existing codebase, so I completely understand that I can't just say "hey devs, fix this" and expect results. But is there anything that I or others can do to help it along?


Short-term solutions?

does anyone have workaround suggestions or workflow improvements that could reduce my 11-step cut by a couple of mouseclicks?

I'm finding that using the keyboard and mouse alone is quite tiring when editing long hours with Kdenlive. Anyone have better results using a graphics tablet or ShuttlePro?

I had suggested to Vincent that the developers focus on bug fixes, easy minor features and a few select, complex features. This workflow request may be a high enough value item to qualify as a worthwhile feature.

If I were to try to do this, here's what I'd need.

A set of video and audio tracks sufficient to build a test project.
A video, w/ audio narrative, of your actions, in Kdenlive, that demonstrate the current workflow.
A video, w/ audio narrative, of your actions, in a professional NLE, that shows the 'Ripple' effect. (Being a committed fan of the Grateful Dead I like that effect name but I digress... )

Placing a bounty on this feature would certainly show that the user community is behind the developer community's efforts. But, that something that you'd have to take the lead on.

P.S. The codebase isn't that bad. It's actually pretty good in comparison to other F/LOSS software that I've worked on. Blender is atrocious.

Ok, thanks for your suggestions Steve!
if you click through to the video you can see the two different processes animated, but I can do screencasts if that's more useful - I don't have Final Cut anymore but Premiere is on my partner's computer, I can't remember if it's the same process or slightly different.
I can set up a small test project tomorrow and a small bounty once one of my invoices gets paid, I'll keep you posted!

I don't understand why you are cutting both video tracks at the same point. Do you want to show the same scene from different angles side by side?

hftom wrote:I don't understand why you are cutting both video tracks at the same point. Do you want to show the same scene from different angles side by side?

Another possibility he's referring to is when compositing effects are used and his suggestion would make trimming a section of clips in the project that use some some kind of overlay go quicker.

Hi hftom, I'll describe a typical scenario: imagine I have an hour of this synchronized footage, which I need to cut down to a 5 min summary.

Before any thought can be given to "which angle should I use" or "are both shots properly composed/focused/exposed" etc, I need to work out how to get across the feeling and subject matter of a 1 hour event in a 5 min story. There's a lot to cut down, but I want to keep my clips aligned so I can choose the right angle later.

So that means a fast '1st pass' going through the footage from beginning to end, cutting out things that I know I won't use, and making notes/ adding markers at any points of interest so I have a good reference and overview to make clear, fast decisions.
To save time in final cut/avid I would scrub through and edit at 1.5x speed (1.25x speed for jobs in German, as my comprehension isn't as fast) but Kdenlive doesn't allow me to listen to the audio at 1.5x speed, so I have to do it at 1x speed. In this case, that would mean 15min unnecessarily wasted right there - audio scrubbing is my #3 feature request...

Then comes a 2nd pass - with a better understanding of the footage and the story that I can build from it, I can now be more ruthless in my cutting.
Again, in final cut/ avid I would always duplicate my 'first pass' sequence which is now perhaps 45min long, and rename it '2nd pass' and edit on that - this means I don't lose any time if I decide to reintroduce something I had removed in a previous pass - I just go to the old sequence, copy the sections required, and paste them into my current sequence.

My workaround in Kdenlive, as suggested by ttguy (thanks for the tip!) is to copy and paste the entire 1st pass and shove it down the end of my timeline, which gives me a ridiculously long timeline - I usually hang on to the previous couple of passes in the timeline for quick access. I also incrementally save the project (project 01.kdenlive, project 02.kdenlive) so that I can import old projects as clips in an emergency. But this is not ideal, because then I don't have the ability to put different effects on the individual clips, or move fine edits around.
Due to this inflexibility I'm a little more careful in Kdenlive, so it takes me a couple more passes overall than when I could 'delete with confidence' in other NLEs.
For example, in this case, my 2nd pass might end up as 30min long rather 20min, requiring a full 3rd pass to get to 20min. (another 20min+ wasted here - duplicating editable sequences is my #2 feature request!)

I might cut the footage down to 8-10min over 4 or 5 passes, and these first few passes are entirely based on the spoken content in the footage - not aesthetics or anything else - this means cutting like you see in the video - just chop, chop chop and some basic rearranging of story chunks & subject matter.

Then I would start thinking about whether any shots need fixing or covering, which angle to use on a particular section, how to minimise the cuts, all that Walter Murch stuff, and also what kind of b-roll might be required at particular sections. By that stage I am no longer using the technique mentioned in this thread, I'm just throwing things around, extending edits, making L- & J-cuts, cutting out ums and ahhs where possible/desirable, and performing delicate surgery on people's long-winded sentences.

Does that make sense?

I understand, thanks for the explanation.
What would be interesting is watching you working on a project.
Maybe you could shot yourself working from start and maybe commenting what you are doing and why you are doing it.
This would greatly help (at least me) to understand your needs.

How many sequences do (would) you have in a typical project ?
And in these NLEs you're reffering to, is there an undo/redo stack per sequence or only a global one ?

Ok, I will have a think about the best way to show this stuff in a video.

How many sequences do (would) you have in a typical project ?

This really depends on the project... but let's take the example of this hypothetical 1-hour event, cut to a 5min video
This is my workflow in the order I would create and use these sequences:

1. original sync sequence - all the clips laid out and synched to the audio track (60min)
2. 1st pass (45min)
3. 2nd pass (30min)
4. 3rd pass (20min)
5. 4th pass (15min)
6. 5th pass (10min)
7. 6th pass - here I'm starting the aesthetic editing, working on the flow etc (8min)
8. 7th pass - having seen how it flows with proper cutting, I have a better idea of what's important, what works etc, and will cut it to approx final length (5min)
9. b-roll sequence - selected shots that I can compare and choose from to lay over my existing tracks (an alternative to this sequence would be making a folder of subclips)
10. 8th pass - adding b-roll (copied from b-roll sequence), tightening edits, tidying up, removing ums and ahhs (4min 30)
11. title/intro sequence - I might cut my establishing shots, pretty details and titles into a 30-second intro sequence, setting the scene and giving the viewer an idea of what's to come.
    This I prefer to do in a separate sequence, as I'm usually cutting to music and throwing a whole lot of different clips around the place, I don't want to knock the rest of my edit out of sync.
12. Picture Lock - assembly of intro sequence (11. title/intro sequence) together with the main edit (10. 8th pass) and credits, addition of titles. (5min)
13. Colour Correction -Now I do color correction, and I export the audio tracks as separate .wav files for mixing in Audacity. (in the proprietary universe I would export an OMF for mixing in ProTools)
14. Export - I delete all audio, and mute all video tracks just in case, and replace the audio with my mixed stereo .wav that I exported from Audacity. I then export 2 files for delivery (upload and archival).

Note that with the exception of 9. b-roll and 11. title/intro sequence, each sequence begins as a duplicate of the previous one.
My workflow in Kdenlive is similar to this, except that all these tidy labelled sequences are instead a mess of poorly labelled footage on a gigantic timeline.
I could never share this project with another editor, I can hardly understand it myself, and the whole thing is in danger of accidental deletions or sync slips etc.
I'm also incrementally saving a new project file (just in case) each time I would have started a new sequence in another NLE.

And in these NLEs you're reffering to, is there an undo/redo stack per sequence or only a global one ?

I believe there's only a global one, but perhaps that's a good idea!

Wow, that's a lot of sequences!
But duplicating a sequence should be quite easy, i only wonder how much memory all these sequences are consuming. Do you have thumbnails enabled?

How do you do step 13 (audio export) in practice ?

Couldn't demux/remux save you step 14 ?

About audio scrubbing : do these NLEs allow you to choose an arbitrary speed factor or only from a list of predefined ones?
And i guess they don't scrub in reverse play, do they?

i only wonder how much memory all these sequences are consuming. Do you have thumbnails enabled?

in FCP, for example, if you have all your sequences open, it consumes a ton of memory. So I would just have my current sequence open and the others closed in the project window.

In Kdenlive, yeah, I notice a difference in performance when working with these giant sequences. I do have thumbnails on, which are important for finding my way around and grabbing the correct clip when I need it.
So in practice, having backed up and incrementally saved my projects, I often delete some of my previous passes.
While editing pass 8, for example, it's unlikely that I will need to grab anything from pass 1,2,3,4 or 5. I'm very familiar with the material at this stage so I just leave a few 'maybe' sections that could potentially be useful and delete the rest of the early passes.
Then I can feel my computer breathe a huge sigh of relief

How do you do step 13 (audio export) in practice ?

Well, sometimes I don't do the round trip to Audacity - Kdenlive is faster, has useful plugins and adding effects is non-destructive and flexible. Normally all that's required are some audio fades, normalise, volume, maybe EQ. (See this post for more on Kdenlive audio effects)
But sometimes there will be poor quality audio, clicks or noises to remove, or tricky transitions, that's where Audacity can be useful.

When editing in Kdenlive I like to keep my audio tracks well organised:
tracks 1 and 2 are speech only - interview, dialogue & voiceover
tracks 3 and 4 are atmo and FX, if required
tracks 5 and 6 are music
As I can't export an OMF with handles, I have to create my own handles wherever I have a transition which isn't quite working in the edit - maybe there's an abrupt cut or something. On those edits I manually extend and overlap the audio a few frames to give me something to work with in Audacity. I need the audio on separate tracks in order to fade between them etc, so often this requires a little bit of arrangement to make sure audio clips are alternating between track 1 & 2.
Then I export the project to ProjectName_Track01Dialogue.wav with all but audio track 1 muted.
and then ProjectName_Track02Dialogue.wav with all but track 2 muted.
and repeat for the other 4 audio tracks.

If I think I need it, I find some atmo as well. Maybe I've already got a specific atmo recording, but I hardly ever remember to do this on location, so instead of having spent 30 seconds standing there quietly recording, I have to spend 10 minutes digging through my footage for something suitable. Editor Me hates Cameraman Me (that lazy jerk).
The atmo will also be brought into Audacity.
I would usually only do an Audacity sound mix once I have approval of the cut from the client, as it's difficult to make changes once it's done.

Couldn't demux/remux save you step 14 ?

yes, it could. But I like to ensure projects are as explicit and clear as possible, so that when I look at it 6 months later, I know what's going on, or if I give it to another editor, they know what's going on.
It makes more sense for me to have that audio mix exported and included in my project, so I know a sound mix exists, and which sequence it syncs up to. It's also not a great deal of work as it's just duplicating the existing sequence and replacing the audio.

About audio scrubbing : do these NLEs allow you to choose an arbitrary speed factor or only from a list of predefined ones?

In most programs using J,K,L as shortcuts you can go faster or slower, just like in Kdenlive. L to play, LL to play faster, etc. I guess these are predefined values. Some will tell you how fast you're playing back (1.5x, 4x, etc)

And i guess they don't scrub in reverse play, do they?

they do, just like the old tape machines. but you can turn this option on or off.

I see; this audio export should be a real pain.

But back to "Ripple edits":

Mark In and Out, then hit a single shortcut, which deletes all footage In-to-Out in unlocked tracks, slides everything else back to fill the gap, and puts the playhead on the In point.

Do you mean seek to in point and continue playing or just seek and pause ?

I was just thinking seek and pause - seek and play might be faster in many cases but I think that's making too many decisions about how the shortcut would be used.

In addition, here's an example of an even faster workflow in Premiere: Ripple to next/previous edit. which could be achieved if the ripple delete function was added to Kdenlive's existing 'Resize item start' and 'Resize item end' shortcuts.

BUT note that this would not work in the case of my '1st pass' where there are very few existing edits - there you would still need to use the In, Out, Ripple Delete technique I suggested.

I've posted an inquiry about bounties on the Kdenlive Development subforum, and made a folder with a sample project here. I can offer $150/€120 to help this feature be implemented!

sharing my suggestion here with others: doesn't it work with overwrite mode (in tool menu, or small icon in status bar with with green square overlapping blue rectangle)?
with the clips you are working on grouped, you just have to add a cut at the end of the zone to delete (shift R), then drag the right part back over the unwanted zone and that's it?

Thanks for the tip, Vincent!
I had already tried using overwrite mode, but I was doing it wrong - I had tried to drag everything using the spacer tool instead of grouping everything and using the selection tool! When using the selection tool, it works very well.

Until now, I have very seldom edited grouped clips (I usually only group stuff which I have no intention of changing) because of the erratic behaviour of grouped clips in normal mode. When you try to cut a chunk of of grouped clips which are aligned perfectly, nice and tidy, it works fine.
But when you try to cut group clips with uneven edges and lots of overlapping sections, things get weird. When trying to edit 'messy' groups (in normal mode) nothing seems to work right, parts of my clips end up in strange places and I'm answered with a symphony of kdenlive's beautiful cascading error sounds

Now, following your suggestion to try it, everything works fine in overwrite mode, when using the selection tool. This has definitely sped things up for me, great!

Of course, it still requires going between the keyboard and the mouse, rather than just using the keyboard as in a ripple delete function.
But I wonder if this could be a starting point for a mouse-free edit.
For example, with all my clips grouped, in overwrite mode, I only have to invent one magic hypothetical hand-wavy shortcut, which I imagine would be easier to implement than a global 'ripple delete' function:
Select Grouped Clips
Play
Set In
Set Out
Pause
Go to Out
Cut Clip
hypothetical 'Move selection to In Point' shortcut (obviously in 'normal' mode this would result in the symphony of error sounds)
Go to In Point
Play

although not as fast or multipurpose as a ripple edit, it doesn't require the mouse, which would be a big improvement over the current situation.
But I don't know, maybe it's not easy to implement at all.

Anyway, thank you for your help, it has saved me 3 clicks/shortcuts off each edit, and that adds up!