Hey Product Hunt, I’m Ritik. Today I’m introducing S-Roll, an agentic harness for video understanding that turns long videos into clips through natural-language prompts. Long videos are full of valuable moments, but finding them still means scrubbing through timelines or uploading entire recordings to the cloud. I built S-Roll to replace that workflow with a conversation. Open a video and describe what you are looking for, such as “Find the KO clip” or “Find where they discuss AI safety.” S-Roll understands the recording, finds the relevant moments, and creates clips for review. From there, it can track the subject, reframe the shot, generate captions, and prepare the final export. The entire system runs locally on Apple silicon. Videos, transcripts, and chats stay on the user’s Mac. S-Roll is the first step toward Saliency, an operating system for video understanding that will own both the models and the agentic harness around them. S-Roll is available free on the Mac App Store with no watermarks. I would love feedback on what works, what does not, and which video workflows should come next.

S-Roll

An agentic harness that turns long videos into clips

An agentic harness that turns long videos into clips