Welcome back to the channel. My name is Matt and I generally make FPV related videos and content, and today’s video is going to be a little bit different because we’re going to be talking about FPV and AI tools, and how I believe there’s a way to merge the two, where we’re actually streamlining our workflow and making little tools, either for ourselves or for other people, where we can actually get things done in a much simpler way, much faster.
Why Build Tools?
And if you’re helping others with it, maybe you can sell it to somebody or whatever you want to do. I am going to put together a little program here. Hopefully, we’ll see how it goes. This is going to be a ride-along session where I actually code it with you guys, and we’re going to be using a couple different AI tools to do it.
The Problem with AVI Files
I’ve got some test footage here, and what I really am trying to get out of this is a video converter that’s going to convert Avi to MP4 or something similar moves fine as well. Yes, you can do all of this without coding an AI program or whatever it is that you want to call this. You don’t need to do any of this to get that done. In fact, you can actually make yourself a little quick action and you can have it just reformat it to an MP4 in like three seconds.
Demonstrating Possibilities
But that’s not the point. The point is, is that we can actually do a lot with something like this. And I just want to kind of like demonstrate that, and maybe I’ll help you guys spark some ideas of what you can do in relation to the FPV hobby, because I think there’s some room for growth, especially in the workflow space.
Storage Challenges in FPV
People are dealing with a lot of files. I’ve got probably over ten terabytes worth of drone footage. I’m sure there’s other people out there that have tons of footage as well. I don’t like to delete stuff, and I’m probably on the lower end of the storage scale there. I, I’m sure that there’s people out there that have a lot of stuff, and they’re just trying to figure out how to manage it. For me personally, I hate these Avi files because they just don’t play down very well and I’ve had compatibility issues.
Real-Life Issues with File Formats
Or if I want to just like send this as a iMessage to somebody, it just gets all distorted and terrible. So I figured I’d make this a little app and thought it’d be kind of fun to just bring you guys along. So if that sounds good, smash the like button. Let’s get started.
Introducing FFmpeg
All right. So first up there is a piece of software that I think we’re going to use. It’s called FFmpeg which if you haven’t heard of it it’s been around for a long time. And it’s just a free open-source piece of software that can do tons of stuff. It can do so much freaking stuff, it’s unbelievable.
FFmpeg and Its Capabilities
But you do need to know code in order to get it going. Not a ton of code. I’m not a coder. I’m more of like a builder, if you want to call it that. I can kind of like come up with a concept of a piece of software. And now with these AI tools, it allows me to actually execute it, which I think is pretty cool.
FFmpeg runs on a lot of software and platforms that you probably don’t even hear about, but for instance, YouTube. I’m pretty sure they are using FFmpeg to encode their videos, and that means that they’re taking whatever trash or good file that they’re getting. They have to be able to take any kind of file, and then they’re converting it into like probably ten different formats, at least both for low res, high res, and then the different screen sizes and whatnot and different, uh, frame rates and things like that. So they’ve got a very complicated operation going on in the background.
Using AI Tools for Research
All right. So I’m using a tool called Perplexity. This is actually the free tier. So you guys can sign up for this yourself. Super easy to use if you haven’t messed around with AI yet. It’s I don’t know what you’re waiting for. It’s groundbreaking. Um, but this is a really good tool for research. And what it’ll do is they’ll scan the internet and come up with, like, ten different sources of information, and then we can use that to feed the other AIs and kind of like compile data. So that’s basically what we’re doing is compiling data.
Exploring FFmpeg’s Functions
So let’s say, um, so research FFmpeg and its functions. And we just want to get a lay of the land of what this tool can do, and then we just want to give it to the other AI. It’s a little redundant because I’m sure it already knows all about this software, in fact. So you could just see a couple. We’ll just pause on that. We can see a couple of the main functionalities of this piece of software.
You can transcode. You could do basic editing, video scaling, super powerful post-production effects. You could do streaming with it. You can use hardware acceleration for it, all kinds of filtering, metadata extraction, etc. And now with the AI tools that are coming out, they’re able to scan images and tell you what’s in the image. And they’re using something called OCR, I believe, and that is OCR scanning.
The Power of OCR
Right. So that’s the optical character recognition where they’re actually converting images into like a machine-readable text. It’s kind of cool stuff that’s been around for a while, but now that it’s being implemented with the AI stuff, I think it’s really getting quite powerful.
Introducing Bolt DIY
So I’m going to copy this information because it’s just good. And I’m using a tool called Bolt DIY. Now, this is a tool that is an offshoot of Bolt Dot New. We’re going to talk about all kinds of AI tools. It is essentially the same piece of software. The only difference is this you’re paying $20 a month for, and it’s using I believe it’s using Claude on the back end. Claude is yet another AI tool, and it’s actually quite good as well.
I don’t think I’ve come across a bad AI tool, to be honest, but Bolt DIY is a locally run service. Let’s see here. I’ve got my terminal here, and what I’m doing is I’m actually running this locally. And how you can do this yourself is you just pull the GitHub repository for it and you literally run the script, and I’m running it locally. You can even put it into Docker if you’re more advanced. I’m keeping it simple right now, and I’m just running the script in the background that allows me to run this.
Local vs. Paid Options
The advantage of running this over the paid version is that I can use whatever AI on the backend that I want, because what Bolt or anything does is it basically is a tool that you can use for coding, and it gives you a really good interface to do all that with on the backend. Like I said, it’s already using Claude in the paid version. When you run it yourself, you can actually choose out of all these different choices what it is that you want to run on the backend.
Choosing Backend AI Tools
So you can run Grok, which is X’s version. I’m on the X Grok beta, and we can use OpenAI, which is ChatGPT. And you can use Anthropic, which is Claude. With that said, you do have to have an API key for these things, and you do have to pay for the credits to use the API. Until the end of the year, Grok/X is running a deal where you can use their system for free. You get $25 worth of credit per month. So you really have like 11 days to take advantage of that.
But you can mess around with this stuff for free if you wanted to. You’d have to just go to X, find the API dashboard, sign up for an API key, and that’s pretty much all you need to do. I don’t even think you need a credit card to get done.
Building the Application
All right. So let’s actually go with Bolt New because I know it’s going to do a really good job. So let’s say I want to create a simple drag-and-drop application. Let’s say a web app where I can upload a video file or set of video files and have a dropdown where I can re-encode them to different extensions.
For example, I have .AVI files that I want to be turned into .MP4, etc. Please make me a modern pretty interface using FFmpeg on the back end to do the work. Or actually, let’s ask it what the best way is, let’s not tell it.
Debugging Early Errors
Unknown error. Oh boy. All right, let’s try this again. We’re going to reload. Oh, kicked me out, that’s why. Ha ha ha! Oh, classic. All right, let’s try this again. Doo doo doo. Start new chat paste. Go.
And what’s great about this tool is it will create all the different files that are needed for the app to run. So now it’s asking. I need you to inform you that video transcoding directly in the browser has significant limitations.
Integrating Backend Services
Okay. Back end service. So we’ll have to talk to a backend service. Created a simplified version that only handles browser-supported conversions. Right. Because we don’t want to crash the browser. I am okay with running FFmpeg on the back end from my computer. How would I link the two so that they can speak to one another?
Experimenting with Colors
Let’s make this prettier. Can we? Now I think there’s something we can use to get better colors. So I want like a front-end color picker. There was something that I heard about, and I can’t remember what it’s called. Color Hunt. Yes. All right, let’s pick something FPV worthy.
Handling More Errors
Error module not found. Okay, so we got a problem. Can I find that? All right. So let’s give this back and do do do. I want to give it the command as well. And actually let’s ask Chat Jebediah. What does this mean?
Your file JS likely contains something like import express from express contains express. When Node.js tries to remember common reasons express is not installed. Okay, so we need Express installed. Now I do not have that installed. I can tell you that much because I’ve never heard of this software.
Progressing Through Challenges
So let’s see what happens when I run this. I hope you guys are liking this because this is just, like, totally experimental. All right, so we’ve run things. There’s a new version available to update run that. Sure. Let’s update. Might as well. If I’m going to do this, I might as well do it right.
Refining and Fixing Issues
After much experimentation and debugging:
- Multiple file support was added.
- Individual progress tracking for each file was implemented.
- Error handling was improved to identify issues with transcoding.
Final Product and Testing
Eventually, the application successfully processed AVI files into MP4, displaying functionality for single and queued file processing. The proof-of-concept showed potential for practical use and further enhancements in the FPV community.