Co-founder Gabe Cheifetz chats about the history of CrumplePop, why they made the move to audio plugins, and how they are using AI.
You co-founded CrumplePop twelve years ago. Can you provide a general overview of your background/career?
We were a small video production shop at the time. We basically ran around making short-form video content in a documentary style, in addition to doing more straightforward commercial work. I think the first video Jed (co-founder), and I worked on together was a short piece about a farm-to-table kitchen in Wisconsin. We worked on very low-budget music videos, an instructional video about making greenhouses, and stuff like that. At one point, our landlord, who had more connections than we did, got us a job shooting crescent dinner rolls and later on, sinks and showers.
Before that, I had experience working at a couple of small tech startups. I realized that there was a lot of opportunity to make good tools for video editors, specifically for Final Cut Pro 6, which was the main widely available editing platform at the time. We started with one simple product โ a lower-thirds template with a fun visual style โ and went from there.
CrumplePop began with a focus on video plugins and, in recent years, made the leap to focus solely on audio plugins. What was the decision behind it?
We made video plugins for many years, but our background was in editing and audio, not in visual effects. As time went on, we interacted with hundreds of video editors and learned about their needs. It became clear that audio was a huge unaddressed problem for video editors. Video people donโt want to think about audio. Often, they just pretend it doesnโt exist.
And this is often true, to one degree or another, from the most simple productions on up. We realized that video creators really need good audio tools, and they need to be designed for video people. This is what led us to audio. It helped that Jed and Pat had audio backgrounds โ and in Jedโs case, he has a lot of experience in analog audio synthesis. As we turned our development focus to audio, they could help with audio intuitions, and I could be the grumpy video editor. Itโs been a really good combination.
L-R, Jed and Patrick, WindRemover research: Field recording wind noise on various microphones
You released your first AI-powered audio plugin, WindRemover, in early 2020. How long has machine learning been on your radar? When did you first make the decision to focus on adding it to your tools?
We had been working on the underlying technology for a couple of years before that. Our first generation of audio plugins was based on classical audio DSP approaches, but we were very interested in the domain of problems that were impossible for this classical approach. Wind noise is a great example. There is no way to deal with that using classical denoising because wind noise is dynamic and has different components. So we knew that if we could deploy AI in the context of audio plugins, it could be really exciting. It took a while to figure out how.
What has the overall process been like?
They say, โBasketball is a game of mistakes.โ We had to make a lot of mistakes. It helped that we had already shipped many plugins over the years, so we kept R&D closely linked to product development. Everyone knew that whatever we were working on would have to work as a shipping product. This helped us keep focused. We also put a very big emphasis on process and streamlining R&D. If you can skip a three-month-long wild goose chase, you can ship three months earlier.
There has been a flood of new AI tools in the post-production space. What makes Crumplepopโs tools different from the competition?
Our approach is to combine AI and usability. I think our success with this is pretty unique. Editors want audio tools that solve their problems quickly. But a one-size-fits-all approach doesnโt work. So weโve designed AI audio tools with a very intentional focus on the needs of video and podcast creators.
How have your AI-powered plugins changed the way audio editors & podcasters work?
AI tools are giving audio and video editors new ways to work faster and deliver a more polished project. And itโs important to note that AI tools arenโt replacing audio experts. AI is giving everyone brand-new capabilities. For instance, we make a plugin, EchoRemover, that uses AI to remove reverb. This is something that wasnโt possible using classical EQ, compressors, limiters, etc. So this tool is useful for everyone from a one-person YouTube channel all the way up to large production teams.
Why are Boris FX and CrumplePop plugins a good fit for each other?
Boris FX has been the go-to for video editors for a long time. CrumplePop makes audio tools for video editors. Bringing audio tools โ designed for video editors โ to the Boris FX community just makes a ton of sense. Video editors are expected to do more and more all the time, and giving them easy ways to solve audio problems is a good idea. Weโre giving Boris FX users access to AI audio tools that are easy and fast. It couldnโt be a more natural complement.
How will your experience with building AI tech prove advantageous to our internal engineers and Boris FX products?
Building AI tech is different from typical software development. The process is different, and I would say the rhythm is different. If youโre used to the usual way of iterating through a development cycle, then developing AI can seem weird. We have a lot of experience with this by now, and this will be really helpful to the engineering team.
What new possibilities or creative opportunities does AI-powered software provide for visual effects artists and editors?
AI tools will help on the visual side in a couple of ways. There are opportunities for โfix-itโ type tools and utilities that are sort of the visual analog of CrumplePop tools. In other words, solving common problems and helping video editors work in a faster and more pleasant way. There are also some interesting paths that can open up on the side of generative AI in the future.
Learn more about CrumplePop and download the free Starter pack.
