Mindset Be honest about the limits of your tool
Removes any sound you can describe from your videos
Camil told the room exactly where SoundScrub breaks down (no spatial selection, a practical clip-length limit) and it made his strong claim, that it beats DaVinci Resolve at isolating a named sound, more believable, not less. Naming your limits signals that you understand your own product and are not overselling, which is precisely what makes your genuine advantages credible. Buyers trust a founder who volunteers the edges. Hiding the limits only means the user discovers them later, at the worst possible moment, and stops trusting everything else you said.
Related advice
Camil Stand on an open model and build the experience around it
SoundScrub's audio separation is powered by Meta's open-source SAM Audio, not a model Camil trained. The leverage for a solo builder in 2026 is enormous: the hardest research is published and free, so your job is the product around it (the interface, the workflow, the trust). Do not try to out-research a frontier lab. Find the capable open model that does the hard part, and compete on the experience, the focus, and the polish that the raw model will never have on its own.
Camil A personal fix becomes a product when others share the pain
SoundScrub began as a script Camil wrote to clean the background noise out of his own holiday footage. The leap to a product came from one question: do enough other people have this exact pain? Scratching your own itch is a great way to start, but it only becomes a business when you confirm the itch is widely shared and worth paying to scratch. Build the thing that solves your problem first, because you understand it deeply, then go check that travelers and videographers are loudly describing the same problem before you commit to productizing it.
Camil Wrap your script in an app for the people who fear the terminal
Camil's noise-removal script already worked; the product was wrapping it in a desktop app for everyone who would never touch a terminal. The market for a command-line tool is other developers; the market for the same capability behind a clean interface is everyone. If you have a script that genuinely solves a problem, the productizing work is mostly removing the technical barriers around it, not adding features. A good interface over a working script reaches an audience an order of magnitude larger than the script ever could.
Camil Let your architecture hand you a feature worth selling
SoundScrub uploads only the extracted audio, never the video, and deletes it right after, so users get a real privacy guarantee that fell out of the technical design rather than a marketing decision. When a property of your architecture happens to answer a fear your users have (privacy, speed, offline use), name it and put it on the page. The most credible selling points are the ones that are simply true because of how the thing is built. Look at your own stack for promises you can already make and are not yet making.
Extracted from
SoundScrub: From a Terminal Script to a Product