Sunny Ray hosts Saurabh Ray, founder of OpenBrew AI, who moved into artificial intelligence from a background in video games and streaming media. Saurabh explains how OpenBrew makes legacy creative software AI native, letting non-technical people create at scale, and traces the idea back to frustrations fixing repetitive bugs and cross-functional miscommunication while working on Paramount's streaming team. He argues that local, on-device AI models are often sufficient for function calling and tool use, reducing reliance on expensive cloud APIs, and predicts consumer hardware will soon run creative AI apps without cloud costs. He describes building Open Brew's popular Tauri-based desktop scaffolding, winning a government SBIR phase one grant and a DoD contract for an agent workspace tied to SharePoint, and launching Motion Buff, a cloud tool that turns documents or decks into narrated videos in minutes for proposals and pitches. Saurabh connects this to the metaverse's unmet demand for 3D content, arguing more diverse non-creative builders are needed to fill virtual worlds, and closes by pointing listeners to openbrew.ai and motionbuff.openbrew.ai.
A game designer turned founder on why local AI beats cloud APIs and how OpenBrew builds AI native tools for non-creatives.
What is Open Brew, and why should people care?
We're building harnesses starting with legacy creative software and making it AI native, so even non-creatives and non-technical people can use our tools. We hope to bring more people into creative endeavors and eventually expand into other tools, ecosystems, and industries beyond creative software.
What does a game designer see in AI that an engineer might miss?
Game designers are builders who use whatever tools get the job done, and a lot of game design is smoke and mirrors, tricks that make people feel and experience things. AI right now is similar, it's really just statistics dressed up as intelligence. When you wrap software around it, like armor around Tony Stark, a regular person can feel like a superhero and get far more done.
Why build on local AI models instead of just calling someone's API?
A lot of what these harnesses need is just function calling, understanding language well enough to turn it into structured data for a deterministic program, and local models handle that fine. APIs are expensive and slow compared to that. I'd rather apply real intelligence where it's needed and let local models handle the grunt work, which is where I think enterprise is already heading.
Your most starred public code is the Tauri desktop scaffolding, not the AI engine. What does that tell you about what developers want?
Most people building local AI want to work on the AI backend, but they don't know how to package it into an app people can actually install. Python is great for the backend but not for a nice-looking frontend, so I combined Tauri with everything needed to talk to a local AI backend, packaged it up, and let people plug in whatever backend they wanted.
You had a working prototype and wrote a funding proposal to the government. What did you actually have to put in it?
I wrote several SBIR, small business innovation research, proposals. One was for a workspace where agents draw on your data, built to connect to SharePoint since government uses that heavily, letting agents grab documents and help teammates collaborate on text or SharePoint sites. I got funded for phase one, and separately the DoD picked it up through a program accelerating AI onboarding from private industry.
What made you go all in on this rather than keep it as a side project?
I was working at another startup in the same space when I got notice that my phase one was accepted, and I'd always wanted to run my own company. Games and entertainment are hard businesses unless you're Disney, whereas software felt more like something people actually expect to buy. Getting that validation was the push I needed to finally commit.
You just stealth launched Motion Buff. What does it do that a video team cannot?
Motion Buff is a cloud tool that builds and hosts your video for you. Normally a video team goes back and forth with other teams on writing, edits, and voiceover, which eats time small businesses like mine don't have. You plug in your documents or slides and it builds a narrated, captioned video in about five minutes, so you can spend your time on the actual proposal.
What does the first yes look like in the next 90 days?
Someone desperate to get a video out near the end of the month. Government funding submissions happen monthly, and if you miss the deadline you wait another 30 days before hearing back, so effectively 60 days lost. Someone who needs it done right now, that's my first yes.
Building something daring? Sunny talks to founders like this every day. Fifteen minutes to see if your story belongs on the stage.
Claim your pre-interview