An Eklipse alternative for clips the kill feed misses
Eklipse is built around detecting in-game events. It reads your screen for the things games put there on purpose: kill feed text, victory banners, scoreboard changes. Within a supported title that is a precise and very effective way to find highlights, and if that is all you need, it is hard to beat.
OptimusClips takes a different route. A self-hosted vision model watches the footage frame by frame and reasons about what is happening, which means it is not limited to games it has been taught, and it can find the moments that never produce an on-screen event at all.
OptimusClips vs Eklipse
| OptimusClips | Eklipse | |
|---|---|---|
| How moments are found | A vision model watches the frames and reasons about what is unfolding, alongside audio and transcript. | Reads on-screen game events such as kill feeds, victory screens and scoreboard changes. |
| Game coverage | No game list. It works on footage, so an unsupported or brand-new title is not a concept that applies. | Detection tuned per title, with over 1,000 games stated as supported. |
| Precision on competitive shooters | Finds the play and the reaction, but will not key off a kill feed the way a purpose-built detector does. | Direct and reliable when the game prints the event on screen. |
| Non-gaming footage | Podcasts, interviews, talks and demos work the same way, since nothing depends on game events. | Built around gaming. |
| Sources | YouTube, Twitch and Kick links, or a direct upload. | Connects to streaming accounts to pull VODs. |
If you play a supported competitive title and the clips you want are kills, wins and objective plays, Eklipse's event detection is purpose-built for exactly that and will be more direct about it than we are. Eklipse states support for over 1,000 titles. Our advantage only shows up when the moment worth clipping is not one the game announces.
What an event detector cannot see
Event detection works by looking for the things a game deliberately displays. That makes it accurate about kills and wins, and blind to everything the game does not announce. A lot of what makes a stream clip travel is in that blind spot.
The near miss that produces a genuinely funny reaction. The moment someone realises what is about to happen a beat before it does. The teammate saying something ridiculous while nothing at all is occurring on screen. The slow build where the interesting part is your face, not the scoreboard. None of that generates an event, so none of it gets flagged.
Because we run a vision model over the frames rather than looking for specific overlays, those moments are the ones our approach is actually aimed at. It is also why the same pipeline works unchanged on a podcast or a talk: nothing in it assumes a game is being played.
Where this genuinely does not beat Eklipse
If you are clipping a competitive shooter and you want every multi-kill, an event detector built for that title is a more direct instrument than a general model, and it is faster. It is looking for one specific pattern and it finds it.
We are also slower in absolute terms. Watching every second takes roughly the runtime of the source video plus overhead, where reading a kill feed does not. If your workflow depends on clips landing within a few minutes of a stream ending, weigh that seriously.
The practical test
Take one VOD you already know well, ideally one where you can remember a moment you would definitely clip by hand, and run it through both. The comparison that matters is not the clip count or the processing time, it is whether the moment you were thinking of came back.
The free tier covers 60 minutes of source video a month, which is enough to run that test without paying for it.
Questions
- Does OptimusClips need to support my game?
- No. There is no game list, because nothing in the pipeline is keyed to a specific title. The model watches the footage the way a person would, so a brand-new game, a modded one, or something with no competitive scene at all works exactly the same as anything else.
- Is it better than Eklipse for Valorant or Apex highlights?
- Not necessarily, and it depends what you mean by highlights. For clipping kills and round wins in a supported competitive title, Eklipse's event detection is aimed precisely at that and is more direct about it. Where we do better is the reaction, the commentary and the moments the game never puts on screen.
- Can I use it on footage that is not gaming?
- Yes, and it needs no different setup. Podcasts, interviews, talks and product demos run through the same pipeline, because the model is watching and listening rather than looking for game events.
- How fast is it?
- Roughly the runtime of the source video plus overhead, because it genuinely watches every second rather than scanning for overlays. That is the real trade for the kind of moments it finds, and it is the honest reason to pick something else if speed is your binding constraint.
- Do I have to connect my Twitch account?
- No. You paste a link to a YouTube, Twitch or Kick video, or upload a file directly. There is no account connection required to get clips back.
See what it finds in your footage
The free tier gives you 60 minutes of source video a month. Paste a YouTube, Twitch or Kick link, or upload your own file.
No card required. See pricing for what the paid tiers add.