You still have hours of footage to go through. You have to find the moments worth keeping, remember what happened earlier, work out whether any of it connects into a story, and somehow turn the whole thing into something another person would actually want to watch. It is one of those jobs that sounds simple until you have done it fifty times.
So when the idea for ShadyCut clicked, I got completely obsessed with it. Apart from a short family vacation, I have worked on it pretty much every day since. But being excited about an idea and proving that the idea actually works are two very different things.
At first, I only wanted to prove the pipeline was possible
The first versions were intentionally simple. I was not trying to build the perfect system; I just wanted to connect the entire process from beginning to end and answer one question:
Can I actually make this work?
Could I give it a long recording, understand enough of what happened, find useful material, build some kind of structure around it, and eventually get something editable out the other side?
The first real tests gave me an answer, just not the one I hoped for. The core idea was possible, but the original approach was nowhere near deep enough.
A stream cannot really be understood from a handful of obvious moments. The system needed more context, better understanding, better story reasoning, and much better decisions about what should survive the cut and why. That was probably the first time I looked at ShadyCut and thought:
Oh. This is a much bigger problem than I thought.
Then came the part where nothing looked exciting
Instead of piling more features onto a weak foundation, I went back to the beginning. From the moment ShadyCut first reads a video, I started moving through the pipeline step by step, improving each part before moving on.
For a couple of weeks, that meant doing a lot of work without getting the satisfying part at the end. No polished demo. No magical new video every evening. Just lots of smaller pieces slowly becoming less stupid.
When you are building something alone, that is a strange phase. You are spending every available hour on it, but you still do not really know whether all those improvements will add up to something good. There were definitely days when I wondered whether I had bitten off way more than I could chew.
The Genshin test was the scary one
Eventually I had another real stream ready for testing: a Genshin Impact session from another creator. There is a case study from that test on the site now, but what the case study does not show is that I was genuinely nervous about running it.
I desperately wanted to see the result, and at exactly the same time I really did not want to see it. Once you have spent weeks telling yourself that the individual pieces are getting better, eventually you have to put real footage through the entire system and find out whether you were right.
There is always an easy excuse while something is unfinished:
Of course the result is not good yet. I am still building it.
Eventually that excuse disappears.
So I ran it.
And for the first time, ShadyCut surprised me. Not because the result was perfect — it was not — and I immediately found things I wanted to improve. But the important part was already there.
It understood enough of the session to find a story, turned that story into an actual first cut, and gave me something I could import into Premiere Pro. The scenes were there, the structure was there, the notes were there, and the editorial guidance was there.
I remember looking at it and thinking:
Holy shit. This is basically what I imagined.
That was the point where the project changed for me. Until then, I was mostly trying to prove that I could build ShadyCut. After that test, I started wondering how far I could actually take it.
Then I sent it back to 1ceStream — who would tell me if it sucked
Testing your own product only gets you so far. When you have built every part of something, you know what it is supposed to do, and your brain starts filling in gaps automatically.
So the next step was obvious: I sent the result back to 1ceStream, the creator whose Genshin Impact session ShadyCut had just turned into a first cut.
That made his reaction particularly important. He already knew the original stream, what actually happened during it, and what the footage felt like before ShadyCut touched it. And I knew he was not going to tell me something was good just to be polite.
His first reaction was:
“This is actually fu*king dope.”
Then came the part I cared about even more:
“Somehow I like the flow. It’s not fast, it’s not slow, just right.”
That hit differently, because pacing and story flow are exactly the kind of things that are difficult to reduce to a technical test. The person who actually created the source footage was watching ShadyCut’s version of his own session and the edit felt right to him.
He also questioned why the opening section felt so long. In this case, that was intentional: ShadyCut had deliberately left more room in the intro so the editor could add context, record a short voiceover, or reshape the opening before the tighter story begins.
But his reaction was still useful. If the intention behind an editorial decision is not immediately clear to the person receiving the cut, then the guidance around that decision probably needs to be clearer too. That is exactly the kind of feedback I need right now — not just whether the edit works, but whether ShadyCut communicates why it made certain choices.
I am still nowhere near answering every question
One successful test does not magically prove everything. There are still plenty of things I do not know.
Will the same approach work equally well for completely different streamer personalities? What happens with streams that jump between games? What if one session contains several completely different stories? Can we eventually build one strong video from multiple separate streams?
Those are real problems I still need to solve.
But I know something today that I did not know two months ago:
The core idea works.
ShadyCut can take a real long-form gaming session, understand enough of what happened to find a story, and turn that into an editable first cut that a human can actually continue working with.
That is no longer the pitch. It is something I can open in Premiere.
Building it alone is still kind of ridiculous
From the outside, ShadyCut is a website and some nice screenshots. Behind that is a system that has grown into something much larger than I expected when I started.
There are days when the scope feels like the kind of thing you would normally split between a whole team. Right now, it is still me: building it, breaking it, testing it, rebuilding it, talking to creators, watching outputs, and trying to make every run slightly better than the last one.
That can be exhausting. But getting the first real signs that another person can use the result and say “yeah, this actually works” gives you a ridiculous amount of energy back.
The goal has not changed
I still want the same thing I wanted when I started.
I want ShadyCut to remove the part of video editing that makes you look at a six-hour stream and think:
I really do not want to go through all of that again.
I want it to do the exhausting groundwork: understand the footage, find the story, build the first cut, and then hand it back to the creator or editor so they can do the part that actually makes the video theirs.
As AI gets better, I will keep looking for ways to bring those improvements into ShadyCut. As new games become important, I want to keep expanding what the system understands and supports.
There is still a stupid amount of work ahead. But two months in, something important has changed.
I am not really asking myself anymore whether ShadyCut can exist.
Now I am asking how good we can make it.