sound problem, very persistent

Hi, I have a very persistent problem with playback sound. I normally use crunchbang, but was having extensive problems, so I installed kdenlive. I still have the problem, but it is now consistent rather than erratic. I am dealing with an interview where I have broken down the interviewee's comments into 3-4 second clips so I can reasseble them for diffrent purposes. I have tried to find the problem in its simplest form, and have even found it is consistent from a live disk, on 2 diffrent computers. Here is how I replicate it:

I put a clip in the project (nb: the clips play fine in a media player).

I put the clip in the timeline, and play the project.

The sound takes about a second to kick in after it should start BUT when it kicks in, it IS in sync (ie. its not a lag, its actually just not playing the sound at the start of the clip)

The problem can sometimes be overcome by fiddling with the bar which selects when you play from, for example, if you play from the exact begining of the sound starting in the clip, it sometimes plays it fine about 1/3 of the time, but if you start from any earlier or later, the problem is the same.

This problem even persists after rendering, so if I render several clips in a row, and play the video in a media player, the sound is still absent for the first second of each clip, as it was before.

I have tried seperating the sound from the video, as I used to do this to solve lag problems, but its still the same, even with the sound in and audio track rather than a video track.

BUT the music track I have underneath seems to play right from the start. This might be becuase there is no gap between the start of the clip and the sound in that case, most of my other clips have a little bit of either blank space or a pause at the start to give me some extra space for transistions etc.

I reallllly need to fix this so i can get back to work. I am self employed and depend on kdenlive for my job. Does anyone have any clues?



ps: sorry about my spelling, i think my spell check is broken on iceweasel and i am very dyslexic



I don't actually think I have ever come across this before, but for anyone to help you with a problem, you really have to let everyone know which version of Kdenlive and MLT you are using and whether it's Ubuntu, Debian, Mint or whatever. Oh, and their version as well.

I hope someone can help.

ok, so I am using avlinux5.0.3 with kdenlive_0.9.2 and mlt is 0.7.6avlinux-1 which is custom and includes


Which sound architecture do you use? Alsa or something else?

as far as i know its alsa - i havent set it up to do anything else :S I know that the system dosnt have any pulse audio on it though

AVlinux has audio and rendering issues with the current "stable" releases of Kdenlive and MLT. However, I've been using Kdenlive successfully on all my systems - including an AVlinux box - by compiling it with the build script and running it from the build directory, which probably doesn't really help you much unless you know how to retrieve and compile programs from their source code. :(

Btw, you are correct, there is no PulseAudio in AVlinux, it's strictly an ALSA and JACK system.



any idea how I could install and older versin of kdenlive on an os? I really don't mind which os I have to use in the end. perhaps that would aliviate my problem?

Hi, this issue was solved for a while, I had decided it was the clips that I was using. there were from a basic dv camera, but had been rendered using an incomplete version of kdenlive on crunchang (not all of mlt was installable)

I am now using ubuntu 12 with a kde desktop (kxstudio) and kdenlive is from the ppa, as update as it can be, no problems found by the setup menus. The same issue has happened again.

I render my video as ogv, my audio as ogg (in audacity) then sync them. Then I render this as one file. Then I cut up this file into 1/2 min clips and render those (as ogv) and then i can re-assemble them in my final file, set them to music etc.

the clips are fine after first render, fine in the syc file, but when i render the audio and video as one wile, the sound problem starts again.

I could try leaving the audio syncing till the end, and just using the camera sound as a guide, however, I still think I would have this problem with the final file anyway - and that way is much more work.

just not sure what to do!

ps: i think this is the camera i used

ogv & ogg are containers, what exactly are the audio & video codecs you are using within these? For example: under Ogg Codecs heading.

Are they the same each time you render at a point in your workflow. I don't use them personally so sorry to ask. Is the audacity audio codec the same as in kdenlive, same for video or are you inadvertently swapping between codecs at each stage.

When you say you render your file as one after first rendering an ogv & ogg from whatever your video / audio sources are are you actually rendering these or just multiplexing them into container format.

Obviously if you are not using a lossless codec every time you render you're loosing quality and introducing error and possible inconsistency. Your workflow does appear a bit conveluted.

What is your camera source and format that you're creating your oggs from, why ogg, why encode to them at all to import into kdenlive, does kdenlive not handle your camera source files straight off?

Hi, the codecs have always been whatever was set up already, so on kdenlive thats Theora, and on audacity thats vorbis (well, it says ogg vorbis files)- I really don't know much at all about these things. They are the same each time I render. I don't know if they are the same in audacity as kdenlive, I don't know how to tell - but this problem persisted before, which is why i started using audacity to render the audio to ogg before using it, as I thought it would help to make it the same as the video. What I mean by rendering is i press render, and select theora, and then render it as a file which I can play in xine or vlc.

My workflow has ended up like this because kdenlive kept crashing. I think my video files are very large, the camera for most of them was - the videos are hd and have large dimensions (bigger than my monitor anyway). I have 2 gb of ram on a modern pc, but it just used to get out of hand. This is also why I render the whole video before I even sync it: Kdenlive handles the video files fine, they just are MP4, but they are really to big to do very much with. I used to just break them up into clips and render the clips, but my computer still struggled. If I render them first, they seem to be easer for kdenlive to handle.

I watched the way my friends who make films and do animations handle their files in adobe/final cut - about half of them seem to use this method of editing each little scene in an individual file, and rendering it as a clip, then importing all the rendered clips into one file to build the final piece. Its been generally very helpful because it means that if I have a massive crash and stuff gets lost (this has happened a few times before) I only lose the jigsaw and not the pieces.

I use ogg because its patent free/etc/etc, I run an "ethical" company, most stuff we make is creative commons and we use only free software. The ogg thing seemed to fit. I didn't know about the lossless thing, I can see that theora is lossy, but there don't seem to be options for things like dirac. Is this not possible in kdenlive?

I really really appreciate this help, I am making stuff like this without hardly any knowledge of video, I am a quick study, as you can see, but time is a problem - I have a lot to do running my own business and all, I really just need to find a pattern that works that I can rely on. I have come a long way since I started a few months ago, but still having so many issues :S

hi, first to compare, generally I end up using an old PC for editing, with a slow processor (Sempron) and 2GB RAM.

The video files are generally of HD Canon DSLRs. h264 with a bitrate of approx 60 to 80 Mbps for the video stream alone.

Basically they're demanding files for kdenlive, so I rely on the proxy feature in kdenlive to make this manageable, the proxy feature means we use far less demanding video files to edit with but render from the full quality originals, this would cut out your whole ogg workflow and maintain better quality.

I use Ubuntu 12.04 with Gnome desktop and Kdenlive from the PPA, it very rarely crashes whilst editing. As you are using a bang up to date PPA build of kdenlive, proxies are an option, have you tried them using your original camera source files?

Your friends workflow is also possible in kdenlive but without the need for rendering an intermediate file, you can do small edits of a bigger project, save them as individual kdenlive project files and import them into a 'Master' kdenlive file as if they were camera video files. In theory anyway. :-) How well it works on complex projects I'm not sure about.

You're doing well, keep going. :-)

If you haven't tried the build scripts for kdenlive I think you should. If you can make them work you'll end up with a fresh compile of kdenlive that might work better.


ok, thanks for all this info, its very helpful. I think proxies is a great idea, however, it feels like a risk to me that you could do all of the work and it not come good in rendering - there seem to be some people with issues here. Also, I think the system I have now, or a simplified version of it, is useful for these specific projects (interviews) as I need to make a lot of similar vidoes using different combinations of the same clips. The stuff about lossy and losless is something I just didnt know, and this will be really helpfull I think - I am going to start running some tests on this today to see if it solved my problem. also, the "master file" and import method sounds good, I know some of my film making buddies use this too. I will take a look.

Can anyone recommend a book about general methods of video editing, methods of structuring your files etc?

Also, can anyone tell me if dirac is actually available and how. I cant seem to find it on the render menus, but lots of places for several years ago talk about it being used :S

ps: I ONLY JUST REALISED YOU CAN PUT KDENLIVE FILES INTO KDENLIVE FILES.... i kept looking for like an import button or something, it was something one of you said about the master file thing and just import it. I know it sounds daft but it never occourd to me to drag and drop like a normal clip. man!

Need to be careful over what the interpretation of 'proxy workflow' is, your editor friends may perhaps think this means encoding low resolution files and editing them, then doing some kind of folder swapping -> file relinking to the full resolution versions for rendering at the risk of loosing some links, where in fact kdenlive has an inbuilt proxy feature that many NLE's don't have, other NLE's rely on background rendering of low resolution files for playback or rely far more on hardware performance.

You could give proxies a go, it's not irreversible, simply tick the use proxys options in the project settings, choose a default proxy encode profile in there too, then when you add your full resolution files to your project kdenlive will automatically create the proxies, you get a progress bar on each thumbnail in the project tree and a 'P' for proxy when each has completed, you'll find playback a lot better using them.

At render time you don't have to do any fiddling and fudging swapping from proxies to full resolution, kdenlive just renders using the full resolution files you added to the project tree. :-)

I've used proxies in kdenlive and it has worked very well for me! It's a really great feature if you have a slower computer! It was a pain to move within MTS files because it took maybe 10 seconds for each mouse click I made.

But, in the newest build from script there seems to be a major speedup working with MTS files! :D So at the moment I don't need to use proxies because it goes really fast to work directly in the MTS files!