Hacker News Reader

Comment by dceddia | Hacker News Reader

I'm making a video editor that removes silence from videos. After creating a bunch of code screencasts, I've found most of my editing time is spent manually cutting out chunks of silence, and it's always felt like a job the computer should be doing.

So I'm making a native Mac app to do it for me. It's in private beta right now, and feedback has been good so far!

I'm hoping to hoping to get it launched in the next few weeks. Aiming for a minimal useful feature set initially – recording the screen, removing silence, and exporting (either an edited video, or the timeline of cuts, to enable editing in Resolve/Premiere/ScreenFlow), and I'll build up from there.

https://getrecut.com

View on HN

I would suggest you to support exporting to OpenTimelineIO - opentimeline.io "Open Source API and interchange format for editorial time line information." Most of the softwares you mention already support it, if I am not mistaken.

+1 I'm interested in automatic video editing too. Both automatic jumpcutting, and also using audio cues like "cut!" to control the auto-editor during recording.

Take a look here https://davidbieber.com/snippets/2020-02-21-jump-cut-program... and at some of the follow-up snippets for what I'm thinking about. Feel free to get in touch.

Nice! I’ve had some beta users request the ability to add markers or audio cues. It could be a nice opportunity to get into some ML stuff.

I'm on mobile so I cant link it, but carykh on YouTube made a bunch of tools like this that remove silence or edit clips on the fly based on you putting a thumbs up or down along with other things. Definitely worth checking out if you want to see something that already exists.

If you do see their animated videos you'll find out that actually most of the animations are automatically created via a script that works with phonetics and emotions they hint it to show. Pretty cool stuff.

I think I’ve seen that! It was impressive. The jumpcutter python script was one of the first things I stumbled upon, and it looked super useful. Then I wanted the ability to tweak the params and visualize in real time, which led me to start in on this project.

Not trying to be discouraging, but my fear would be that your product is just a feature that other video editing software products will add if it is popular. Have you thought about your endgame? It is a good idea though -- I'm surprised they don't already do that.

Yeah, that's something I've considered. I'm surprised they don't yet too. Even though it might be a short term opportunity, it feels worth pursuing for now because it's a tool that can save me and others a good chunk of time.

My other line of thinking is that this could expand into an editor that's purpose-built for screencasters, with whatever other niche features that might entail.

Great idea. I've been using the command line application from https://github.com/carykh/jumpcutter for my videos to automatically remove silence but it's not as user friendly as yours looks.

Others here mentioned carykh's jumpcutter. I tried to get it to install but couldn't, then discovered it hasn't been updated in a long time.

There's even an issue with links to alternatives and forks: https://github.com/carykh/jumpcutter/issues/180

I tried out Auto-editor [1] and it works great for what I needed! Although for "removing" silence, I prefer to increase play speed instead of jump cutting.

[1]: https://github.com/WyattBlue/auto-editor

Very interesting. I'd be interested in an editor that could do a whole host of common mundane tasks in this way while maintaining the non-destructive editing promise. I often have to cut out coughs, umms and urrs as well. That takes up more of my time than anything.

Neat. Does this really need an app? I think it may be able to do with the command line. Something like ffmpeg that performs video operations and something to detect a noise level treshold.

There are some command line tools to do this (see other comments), and I think they use ffmpeg under the hood.

They work, and I built a version of one myself as a proof of concept before I started this, but I quickly realized I wanted visual feedback and ability to preview the edits before exporting. Without that it’s a lot of guess-and-check.

With Recut I can tweak the padding value, hit play, see if it sounds right, and only export once.

Yes, because not everyone is comfortable with the command line

Can you elaborate what it means to remove the chunks of silence? Aren't there valid cases where there's no sound but you're actually showing/doing something on the video.

Yep there are! This is a harder problem to solve but I have plans to handle it down the road.

For now, I’m targeting it at folks who make videos in the “egghead style” [0] - short, tightly-edited code screencasts where there’s very little dead air.

0: https://egghead.io and https://howtoegghead.com

i'd think it wouldn't be too hard to use something like opencv to detect frames that don't have any/much change between them and then correlate that with the audio detection to figure out what can be safely culled like in that fashion.

You can use webRTC VAD or the VAD from RNNoise.

https://github.com/xiph/rnnoise

Interesting, thanks for the link. This would be much fancier than the simple thresholding I'm doing currently.

It might be overkill for the use cases I'm targeting right now (clean recordings, probably done with audio gear) but I'll make a note to check it out.

Reply on news.ycombinator.com