LLM based video clipping
Video content repurposing is a pretty saturated space, yet in talking to users I learned that most YouTubers view the solutions as having stagnated. In general most clips are unpublishable because of AI errors.
The first problem I wanted to tackle was inaccurate AI reframing for screenshare based content. When compared to the leading competitor OpusClip, SaralClip can more intelligently reframe these sorts of videos.
Demo videoSee how SaralClip's AI reframing compares to OpusClip for screenshare content.
Open Source
Contributing to Google's Gemini CLI was a very cool experience for me. After contributing early on, I was reached out to by the team and given the opportunity to directly collaborate with them in their group chat. I got to learn a lot about how a popular repository is run, as well as learn a tremendous amount about Software Engineering in general. It was also cool to see my gmail address in a sea of @google.coms.
Favorite PRs
Rendering a mutable object in React is problematic because updating it will not drive a new render. The infrastructure laid in this PR allowed for an idiomatic way to render the Settings configuration in the UI.
I like this PR because it shows why idiomatic code is important. It builds off #14915 to remove 1376 lines of code and fixes 5 bugs at the same time.
Those are just my favorites. For a full list see all my merged PRs.
I also have contributions to Vercel's AI SDK. I use this library extensively in the Frontend and Backend for SaralClip. In addition to code contributions, I also participate in discussions and posts where my priority is to help others maximize the AI SDK.
For a full list of PRs see all my merged PRs.