Home / How it works
How Glimmer finds stream highlights without generative AI
Glimmer listens to and watches your whole stream, so you start from a first cut instead of a blank timeline. It finds most of the good moments. You fix the rest by drag and drop.
3 hstream in (example)
15 minvideo out
3things it checks: words, voice, screen
What does it look at?
| It checks | It catches | So you can spot |
|---|---|---|
| Your words | Hype words, fast talking, repeated phrases | The moments you got loud about |
| Your voice | Louder or higher than your normal, quiet to loud jumps | Screams and big reactions |
| Your screen | Deaths, boss bars, victory banners (you box them once) | Fails and wins |
How does it pick?
score = events x weight
+ audio excitement
+ speech signalsHighest scores that fit your video length win.
+ audio excitement
+ speech signalsHighest scores that fit your video length win.
What do you get?
- Cuts land in speech pauses, never mid-word
- Every clip shows why it was picked
- Say “not this one” and it ranks similar moments lower
- Your edits stay when you run again
- Learns from your edits, on your PC only
- A small program on your PC (Whisper) turns your speech into text. It only reads your words.