Obskura started as a solution to a simple problem: recording livestreams reliably, automatically, and locally. That part works. But once you've got hours of recorded video sitting on your drive, a new problem appears: how do you actually find the moment you're looking for?

That's where the next phase of work is headed. It's currently in research and development — we're figuring out what's feasible before making promises. Here's what we're exploring.

Clip Search: Finding the Moment, Not the File

If you remember a funny reaction, a clutch play, or a specific phrase from a stream you recorded last week, the current workflow is painful. You scrub through hours of video, hoping you're in the right ballpark. Sometimes you give up and just never watch it.

Clip search is about changing that. The idea is to run ML models across your recordings so you can query them naturally — by what happened, what was said, or what was on screen — and jump straight to the moment.

There are real tradeoffs here: running models locally means respecting your hardware, which is why we're taking time to benchmark different approaches. The goal is something that actually works on a regular machine, not a workstation.

Automatic Clipping: Highlights Without the Hunt

The further-out idea is automatic clipping — surfacing the moments worth watching without you having to look for them. Think of it as the recording equivalent of a good editor: the boring parts get skipped, the highlights float to the top.

This one is harder. "Highlight" means different things to different viewers, and a model that's too aggressive will annoy you by cutting things you actually cared about. It's on the roadmap as something to explore once clip search is solid.

Why Take the Time?

It would be easy to ship something fast and call it done. We'd rather spend time making sure what ships is actually useful. Obskura's whole point is reliability — a recorder you can trust to be there when you need it. Anything we add to that should meet the same bar.

So: no firm timeline yet. But if you've ever wished you could jump to a specific moment in a recording without the scrub-hunt, this is the direction we're heading.

Want to Follow Along?

The best place to hear about R&D progress is our Discord community (available for active subscribers). Feedback from real users has shaped the roadmap more than anything else.

And if you're curious about the bigger picture — why this project exists and where the subscription money goes — the About page has more.