Provenance

All this because of a little "Made with AI" tag.

It all started with a simple "Made with AI" tag on X (Twitter).

On a video I spent a couple of weeks editing.

Great.

But to be fair, it was partly my mistake. I had used an AI-generated voiceover from ElevenLabs.

Which is fine. Nothing wrong with that.

But if I'd known about the tag before posting the video, I probably would've handled things differently.

So I did what anyone would do.

Googled it.

Apparently, there are already tools that let you check if your video contains information about AI-generated content.

Nice.

But there's a catch.

You have to upload your file to their servers.

Now, I don't know about you, but uploading my unreleased work to some random website just to check a little tag?

Yeah. No thanks.

So I started digging.

That's when I came across C2PA (Coalition for Content Provenance and Authenticity).

Basically, it's a standard that lets files carry information about where they came from, what tools were used, and how they might have been edited.

These records are called Content Credentials.

Think of it like metadata, except you can't just right-click a file, open Properties, and expect to see everything.

Okay, cool.

But how do I actually read this information from a video that's several gigabytes in size?

Without uploading it to some random dude's server?

Turns out, there's already an official C2PA command-line tool.

It runs in Terminal on macOS or Command Prompt on Windows.

You give it a file. It gives you the information.

Simple, right?

Well, technically.

Except it also gives you a shitload of JSON.

And I have absolutely no idea what half of it means.

Which is fine.

I'm not saying terminals are bad. They're great.

But they're not for everyone.

A marketer, content writer, or social media manager shouldn't have to learn terminal commands just to check a video before posting it.

They have other things to worry about.

Like why the video needs to be in six different aspect ratios.

At this point, I was already thinking about building a simple website for this.

But I had one requirement.

No file uploads. Anywhere.

Your file. Your computer. That's it.

Now, running the same command-line tool through a typical web server would mean sending the file somewhere for processing.

Which is exactly what I didn't want.

And before someone tells me there are probably better ways to do this, you're probably right.

I'm a motion designer. Not a full-time developer.

Anyway.

While digging through C2PA's GitHub repositories, I found that they also provide WebAssembly (WASM) support.

Wait.

So I can run the credential-checking code directly inside someone's browser?

Without sending the file to a server?

That's exactly what I was looking for.

So I started building a little UI around it.

And that's how Provenance was born.

A tool that lets you check Content Credentials in images, videos, audio files, and PDFs.

Right inside your browser.

For now, Provenance only supports files under 1 GB.

Not ideal, especially when you're working with videos.

But I thought that's fine for a first version.

Who knows? Maybe one day I'll wake up at 2 PM like a cockroach and decide to fix the file size limitation.

Until then, 1 GB it is.

Drop a file. Check its credentials. That's pretty much it.

Oh, and one important thing.

No Content Credentials doesn't mean a file is fake. And having them doesn't automatically mean it's real.

It simply shows whatever provenance information is available and whether the credentials can be verified.

It's not an AI detector or a magical truth machine.

I wish it were that easy.

Anyway, I don't want to explain every little thing about the tool here.

Go try Provenance

It runs entirely in your browser.

I don't get to see your files.

Trust me. Or don't. That's kind of why I built it this way.