Lamdu’s second 2 minute video: this time focusing ...
# two-minute-week
y
Lamdu’s second 2 minute video: this time focusing on localization, which I haven’t yet made any video or post about
👏🏽 8
❤️ 3
i
This is beautiful and elegant. So inspiring to see. I wish I could say more than that, but what you've shown seems very well-considered and I don't have anything meaningful to add. I also really appreciate the effort you're putting into these videos — the planning and editing are a real benefit. If there was a way to adjust how you're processing the audio so it doesn't cut in and out so abruptly, that'd help me focus more on the substance of what you're saying. I'd probably find it less distracting if you didn't filter/cut out the "silences" (room tone) between your sentences at all.
❤️ 2
y
@Ivan Reese true. I’m actually quite embarrased about the sound. for our youtube videos I actually went and recorded the sound with a proper setup but here it’s been a very quick thing..
i
Quick is good! All I mean to say is — the background noise is not bad. Embrace it, for this format. I'm also going to be embracing a "quick and dirty" aesthetic for my videos. Your audio has good levels, and your voice is loud and clear and pleasing to listen to. No need to try to carve out silences.
y
It’s not really a conscious effort to carve out, but more of an edit of different takes I’m recording within final cut as I make the edit, where final cut doesn’t make it easy if you want to cross fade etc as by default it’s just layed out as layers attached to the magnetic timeline, but yeah I can try to group together the clips. wrt carving silences actually in our youtube videos my audio engineer friend did put a noise gate on them but the recordings were much better to begin with and didn’t have much noticeable noise..
s
@yairchu i recorded my video with a pretty bad setup (webcam) and did some very basic noise removal in Audacity - its very easy to do:

https://www.youtube.com/watch?v=xkpzHJGE4Dk

That and fade in/out should take care of most of it I think :)
y
@s-ol thanks! I also recorded with a webcam and I do have access and some knowledge of audio processing tools, I develop them for a living.. It’s just that I didn’t bother “going pro” for a quick 2 minutes video. For Lamdu’s videos on youtube I’ve explicitly went to record at the pro recording studio at my office (which at the moment I don’t have access to while quarantining at home) and recorded in a quality room with quality hardware and had a pro audio engineer process it a bit (I’m just a programmer after all..). For here I just wanted something ready without much effort and recorded with my webcam and only applied the untouched compressor preset I had working on my last youtube video.. But I get it, and will make an effort to improve the sound editing for next week’s video 🙂
s
@yairchu oh, I see. In any case, I'm really liking your to-the-point videos! You definitely have my attention ;)
👍 1
i
The best example of the problem I have is, say, from 0:10 to 0:20 in the video. The room noise in the background is less distracting than the fact that it keeps cutting in and out. So when editing in Final Cut, you can the "empty" room tone between your sentences, rather than cutting your audio clips so that they only exist when you're speaking. Again, this isn't super important, it's just a small thing that'd hopefully not cost you any additional effort but make the videos easier to follow. As for Sol's suggestion to use noise removal, I find that that always introduces artifacts, and the artifacts are just as distracting as the noise. Better to just live with the noise floor, I say — but that's just my personal taste.
👍 1
y
@Ivan Reese I know that some very serious folks swear by some noise removal tools like Izotope’s, and also Accusonus’s not bad I think
i
Yeah, Izotope makes some amazing tools. But a lot of your success with noise removal depends on the kind of noise being removed. Is it hiss? Is it a pure tone, like 60hz line hum? Those are easier to remove without side effects. But lots of noises are less easily characterized, and even the best tools I've seen will have trouble with them. That said.. I've yet to see anyone apply ML to this problem, and it seems like the sort of thing that'd be very well suited. Could be there are better solutions now since the last time I checked, even.