Need advice on how to optimize my youtubing workflow

[McJty]

Hi, as some of you know I regularly do youtube videos (both tutorials and my lets play) but I have some trouble getting my workflow to be optimal.The problem is that it is a lot of work to get the resulting video good quality and I have to do a lot of repetitive things. I can't imagine there are no better ways for this. I play, record and make the final result on Linux and I use about the following steps:

• Minecraft runs in window mode but fullscreen
• In OBS I set up a source called for 'Video Capture Device (V4L2)'. I have also tried 'Window Capture (Xcomposite)' but that doesn't seem to make any difference
• My Sennheiser microphone/headset is connected to the back of my PC as the front one gives a lot of static whenever the cpu/gpu is doing work
• OBS Stream is configured to 'YouTube / YouTube Gaming' but otherwise I don't stream
• OBS Microphone is set to 'Built-in Audio Analog Stereo'

One of the problem is that the microphone seems to catch a lot more static on Linux then it did on Windows. Luckily I can filter that out easily with Audacity. Now to continue:

I have hotkeys configured to start and stop recording. Whenever I play I constantly start and stop recording as my audience doesn't want to see everything that I'm doing. The end result of that is a lot of different videos (flv files)

When I'm done then for every flv file I do the following in Audacity:

• I open the flv file
• I select a suitable noise profile and eliminate noise. This works pretty well luckily
• I also do 'Compress Dynamics 1.2.6' as that seems to improve my voice considerably
• I export the audio as a wav file. Note that sometimes this wav file has a slightly different length then the original flv file (visible in kdenlive). Not sure what to do about that

Then in KdenLive I combine everything. That means I also have to match up the different shots again because of the slight length difference and I have to get rid of the audio of the flv there. Then finally everything is rendered to MP4 (H.264/AAC fast) profile (not sure if that's best but it works)

This is all a lot of work and time consuming. Can some of these steps be automated? Are there better ways to do things? What about my render output profile? Is that good for youtube?

Thanks for any advice

 

[Roefel]

i ain't a pro youtuber but i think everything you do kind of has to be done. it is a real time consuming thing.
But what i do is i record everything in 1 shot and then just cut out the pieces i dont need. that way i dont get many different files.
I don't think it will make that much of a difference in time (maybe it does in your case with the time difference thing).
but i just generally prefer to do it that way. idk if it's better or not tho.
also i use the same render setting and i think they're fine so, wouldn't really worry about that.

good luck with the channel

 

[Someone Else 37]

Is it necessary to do the noise reduction in Audacity on each individual segment, before stringing them together? It seems like it would be easier to piece them together first, then noise-reduce the audio afterward, when you only have one file to deal with.

As for why Audacity seems to change the length of the audio a bit... I don't know. Maybe Audacity and kdenlive disagree slightly about the sample rate? There might be a setting about that somewhere...

As a matter of fact, does OBS have an option to choose what format it outputs videos in? If so, you might have better luck if you configure it to produce MP4 files, as that's what your video winds up in anyway- it's possible that flv and MP4 just run at slightly different rates, or that they interpret the sample rate in slightly different ways. While you're at it, you could try having Audacity output the denoised audio as MP3 or OGG and see if those line up better.

As for automation, if you can find command-line utilities that do the things you're using Audacity and/or kdenlive for, you could write a shell script to handle it... but I don't know if such things exist (though they probably do) or what they might be called. Your Linux distro almost certainly has a built-in package manager (Debian and Ubuntu use apt-get, for example), and it likely has a function to search through its repositories... if that sentence made any sense at all to you, that's probably where I would start looking.

 

[McJty]

Well if I do the noise reduction to the entire thing then it means I have to render the end result to a file, extract audio, and then somehow combine the video and audio again. Is there an easy way to do that without quality loss (i.e. having to render the video twice)? It would make things a lot easier though.

 

[Roefel]

it is posibble to save obs recordings as mp4's just go to the output and wherever it exactly is you just need to save it as "video.mp4" for example, and it will save it as an mp4 if you do it like that

 

[McJty]

it is posibble to save obs recordings as mp4's just go to the output and wherever it exactly is you just need to save it as "video.mp4" for example, and it will save it as an mp4 if you do it like that

How is that better then flv? I'm a noob with regards to video

 

[Someone Else 37]

Well if I do the noise reduction to the entire thing then it means I have to render the end result to a file, extract audio, and then somehow combine the video and audio again. Is there an easy way to do that without quality loss (i.e. having to render the video twice)? It would make things a lot easier though.

*shrug* No idea.

If you can make kdenlive output the video from the first rendering in an uncompressed format (do uncompressed digital video formats even exist?), you'd get a huge file, but the quality would be perfect.

As a matter of fact, you could string the video segments together in kdenlive, save it as a kden project file (which I can only assume can be done), then render it as MP4 or what have you with the lowest possible video quality (but normal audio quality), import the audio from that into Audacity the same way you've been doing with the flv files, do noise reduction, export as MP3, and finally import that back into the kden project.

That's still several steps, but at least you don't have to perform them over and over again for each segment.

How is that better then flv? I'm a noob with regards to video

My thought was that MP3 and MP4 are probably related, so if you take the audio out of an MP4 video, save it as an MP3 file, then stuff it back into the MP4 file, it might not create the "slight length difference" you mentioned. flv (Adobe Flash video), wav (a Microsoft audio format), and MP4 (based on Apple QuickTime) are three entirely different formats developed by three different companies. While I am far from an expert on this sort of thing, I can only assume that formats following similar standards (e.g. MP4 and MP3) would be less likely to cause those sorts of issues.