I have to take all this.

Throw it in the timeline.

There is nine hours of me coding.

So, ideally, I would like to say Take this nine hours of footage, cut it down to be 30 seconds long and sink it to a certain kind of music.

You would still have a human directing the storyline, but you would basically get the algorithm to do the grunt work off, doing actual cuts for you.

So I think this is possible, but it's kind of impossible.

So we're gonna see how far we can get in this video.

So there's already a lot of code that's written that can that can process video.

So I don't have to, like, reinvent the wheel.

Pretty sure I'm gonna use FFM.

Hey, honestly, I feel like looking at the source code will be very good.

I think that's a good place to start, man.

See, programming has been a while.

What am I looking at right now?

Is what I get when I google.

Why CBC are discreet.

Co sign transform doing the math for this right now.

See if it makes sense I've copied this code from the example, but what I'm doing is I'm changing some of the numbers to see if I understand what's going on.

If I do zero, that's gonna be like a grayscale image.

It's not a grayscale image, so I was actually generating video files through the code here, which was really, really awesome.

The only problem is, I tried to generate certain colors, and I was not able to generate the colors that I was expecting.

I'm Googling all these questions and 99% of the time I don't find what I'm looking for.

So that just means I need to do like 100 Google searches.

Why do you guys even watch this channel?

I am actually the biggest idiots.

The problem was, I was using the wrong command line arguments.

I was trying to convert this string MPEG one into an integer and process it through the image.

Um, let's try to create Deep Green.

The red needs to be all the way down, and the blue needs to be all the way down.

16 16 secrete green row, Row A o Day three.

Got a ton of papers on the desk it feels good.

E basically been looking at their functions on, like, open Kodak contacts, how they actually find the stream and how they find the decoder from the stream and all that stuff.

And I've basically been writing my own little version of this yesterday way made so much progress with the clip.

Now we're actually able to get the packets from a clip given a start in an endpoint, that means I could take those packets.

I could decode them if I want.

If I could start doing processing on them way wrote 1000 lines of code in the past few days.

All right, we're gonna see if this blows up or goes horribly wrong.

Look at this right sequence to output file.

No streams to mucks were specified.

The first frame is kind of there, but it looks super messed up way export more than a few frames, and it actually, um, actually works.

It's glitchy.

It's so glitchy.

Well, that way I've spent the last week basically building up all the data structures that would allow me to make cuts and splice them together into a sequence.

Right.

So, essentially, now I have the basic building block that every other algorithm will build on top of.

Finally, we're going to start writing the algorithm now.

This is gonna be the first demo where we take a bunch of clips.

I think what we're gonna do is the user says, Here's a directory with a bunch of videos in it.

I want you to randomly edit these videos together so way we've lost a lot of memory way.

Scan a directory for all the videos, and we put them in the sequence one after the other.

That's good.

That's good.

So I could show you my sheet of paper here.

Can you see what's going on here?

So, look, we got our sequence here right now.

We gotta like we got all the clips laid out one after each other.

Now, we just gotta, like, randomly make a cut.

Imagine editing a video for eight hours, and then just seeing a guy beside you do.

And he's got the same edit In Majesty building it.

Look, it's building the edit Man prince sequence.

Oh, it's brilliant at it for us.

Cool.

We should get a 12th video out of this is like random.

Uh, it's basically just me, Like walking around with the camera, you know?

And I'm like, uh, look, this clip sequence read frame Kern owed equals no.

Well, I think I basically need to do this.

They could always point to the head.

Oh, it's outputting the video audio.

Okay, so it's doing audio packets.

Video packet.

Let's hit play.

Let's see what our algorithm randomly edited together.

Oh, all this footage I shot January 2nd January 3rd.

Now we're gonna get our algorithm to edit this for because this is a bit more like this is me getting out, like doing things.

Let's say 20.

We're gonna do a quick at it.

Oh, if this is really aspirin, so Okay, this So I bought 14 video and it Yeah.

No, it's not.

It's not really good for talking footage.

What if we just take out the audio?

We just mute the audio and we overlay it with music.

Will people be able to tell that it was edited by an algorithm and not by me?

I know when I had it, my videos.

When I make those cuts to the music, I'll do them.

I'll be like every 10 frames or something I'll do a cut.

So that's what I told the algorithm here.

I told the algorithm, Yo, each cut is gonna be, on average, 13 frames, but it can very up to 10.

So it could be three frames.

Or it could be, like 20.

What if I said, like, 40 seconds?

Just okay, let's do it.

Let's let's see what's possible here.

Um, it doesn't look as bad as I thought.

Well, once you remove the audio, remove the yacht, all we're done.

Okay, let's watch it with sound.

I guess it's a lot of me talking in.

It is gonna be really awkward.

There's putting my wall.

Thanks, E.

Everything is trying.

It's fine.

Wait.

Okay, let's put it to music.

Honestly, that was It's not as bad as I felt like.

It's actually kind of watchable.

What is annoying is that every frame is roughly like the same length.

So we could we could probably play around with the variety of that.

You know, maybe I only use, like, six seconds of that.

But if I use six seconds of that in a bigger video, that would actually work, you know?

I mean, and that's like that would probably that would literally save me, like an hour of going through that footage and cutting that down just by randomly do, do, do, do, do do.

Obviously, this is not like a solution.

This is this is just experimental work where I'm just like seeing what's possible.

This is actually a perfect way for me to learn video processing.

I'm just trying to learn how this stuff works.

You guys could be the source code.

However, there's really no documentation.

I haven't had time to, like, fully tested, so I don't recommend trying to run it.

And I don't recommend trying to, like, downloaded.

I'm gonna fix it up and keep working on it.

I have so many ideas for other algorithms I could actually code up and just see what the outcome is.

If you guys wanna support me, sub to this channel, subscribe and also check out my patri on that would be super cool and super helpful.