To be honest, for me, AI has been a real big fail, as has the 3D phase in post. I can't think of any AI workflow that has enhanced or improved my workflow. I watched Adobe's video of what's coming to Premiere, and to be honest, Adobe disappointed me by adding enhancements that “revolutionize” past workflows. Don't get me wrong, I'm sure there are parts of AI workflows that people find useful, but there's nothing that I can say I use regularly in my workflow. Until now. A few months ago, I had my “it” moment. Not only do we use AI tools in a variety of projects, but it's easy to imagine where this “machine learning” artificial intelligence workflow might go in the future. Well done, Boris FX, well done!
Intelligent… that's for sure.
Now, let me preface this article by saying that I have not received a penny from Boris FX for this article. I have been using Continuum since before it was offered for free to all Media Composer editors who upgraded to Symphony (I'm sure all Media Composer editors remember that), and I've been using it with After Effects for as long as I can remember. Most of the “wow” benchmarks that drive my workflow forward don't come from an NLE or compositing application. They come from Continuum.
To be honest, Boris FX has made some pretty surprising acquisitions over the last few years. Licensing GenArts, Imagineer Systems, WonderTouch, Syntheyes, and even Primatte technology has allowed Boris FX to become a one-stop shop for pretty much everything an editor or compositor needs. For me, the biggest leap in the last 15 years has been the integration of Mocha technology into almost every effect in Continuum and Sapphire, as well as the licensing of it into After Effects, which has made it the standard for AE tracking today. So, you might be wondering, what does this have to do with AE and machine learning? Well, Boris FX has just released the 2024.5 update for Continuum, and hidden within is a glimpse into the future of the effects package. Believe it or not, this Witness Protection effect paves the way for the next generation of effects in Continuum, and in the process, saves editors and graphic designers countless hours of wasted time even with the best tools available today.
While everything is being called “AI” these days, Boris FX decided to go in a slightly different direction by calling their stuff ML, or “machine learning.” In Continuum, you can find four “ML” effects with a simple search:
So, if you look at the Media Composer version of the image above (ML effects are available in other host applications in Continuum), you can see that there are actually four different ML effects: DeNoise, ReTimer, UpRez, and Witness Protection, which is a real-time effect in Media Composer. So what exactly is the Witness Protection effect? Well, you've seen it many times before. You need to blur the face of someone walking down the street, but you don't have permission to use their likeness in your production. That's when you use an effect like this. However, in the past it has worked quite differently, and to be honest, the effect went from shaky to very cool to awesome. It was shaky at first, but it used the Continuum tracker to do all the motion tracking. We all know how awful point tracking can be, but the integration of Mocha into almost all effects in Continuum has taken this effect up a notch, as it made the tracking process much easier and more accurate. It wasn't without its issues, though. If your talent walks behind objects like trees or lamp posts, it requires more work in Mocha and in fact, the editor was a bit fed up whenever we wanted this kind of effect. Because we know how much time this kind of work really takes and it can be very slow. But no more. How does it work? Drag and drop. Yes. That's it. Drag the effect (or apply, depending on the application you are using) and you're done. Is your talent walking behind something? No problem. The ML (machine learning) effect will be dropped when your talent comes back from behind something. Is your character walking in screen or off screen? Again, no problem. ML will add the effect again when your character reappears. Here's what I mean:
This effect includes everything that was previously available. There's the ability to switch to a mosaic pattern instead of blur if you want, and you can also turn ML off completely if you want to apply the effect to something different, like a logo on someone's shirt.
That being said, here's where I really see the potential of this effect: currently, the ML component is designed to detect faces and essentially add an ellipse as a mask to blur or pixelate a person's face in Continuum.
I was amazed at how fast and accurate it was. The only adjustments I really had to make were to add a few feathers to the mask and make it a little bigger, and it did everything else for me. It was my first time doing anything with an AI, and it was amazing. “Seriously, I'm really looking forward to using this on a regular basis.”So let's take this effect a step further. What about logos on shirts and products? What about the ability to blur nude bodies? What about the ability to look at a transcript and blur the mouths of people who swear? These applications can really be used in the day-to-day workflow of editors, especially those working on reality TV, and can save a ton of time on compositing. You can also explore other effects in Continuum to see where applications like this can speed up your workflow. Take the Lens Flare effect for example. You can save a ton of time by just typing in what you want to “attach” the lens flare to. The sun, the headlight, the flashlight. Something so simple can save you minutes, maybe even hours, of tracking time. For me, this effect got me interested in artificial intelligence/machine learning in NLE/compositing applications, because I can easily imagine different editors using this in all kinds of productions. For more information on Continuum 2024.5, visit borisfx.com.
