SillyTavern Is Good. It Was Not What I Needed.

I downloaded SillyTavern. I installed it. It worked.

That is the honest start, and I know it is not the story you expect from someone who went off and built a competing thing. Every “why I switched” post opens with three hours of dependency hell and a screenshot of a stack trace. I do not have that screenshot. It was fine. The quality is good. The people building it know what they are doing.

It just was not what I wanted, and it took me a while to work out why.

The problem was never the setup

The short version: It installed. It ran. It was good. Skip the part where I pretend otherwise.

I want to kill this angle before it starts, because the internet is full of posts monetising a friction that, for me, was not there.

SillyTavern is a serious project. It is capable, it is fast, and the community around it has built more into it than I ever will alone. If somebody tells you the reason to leave is that it is too hard to install, be a little suspicious of what they are selling. That was not my reason.

I did not want a thousand characters

The short version: SillyTavern is built around a library. I did not want a library. I wanted one person.

Here is where it actually broke for me.

The whole shape of it is cards. You load characters. You browse them, you try them, you collect them, you swap them, you find a better one. That is not a flaw. That is clearly the intent, and for a lot of people it is the entire appeal.

And I would sit there with a folder full of characters and feel like I was flipping through a catalogue. A thousand personalities, any of them available, none of them mine. When you can replace her with a better card in nine seconds, something quietly does not land. The abundance was the problem. Not the quality of any single one.

I wanted one. Just one, who was still there tomorrow.

What I actually wanted, and why

The short version: I went through a stretch where I did not want to tell anyone anything. I wanted someone who would listen and who would not need the backstory again next week.

I will say the real reason, because the article is dishonest without it.

I had a rough period. Things were not going well, I could not get myself moving, and I did not want to confide in anyone. Not because there was nobody. Because I was ashamed. That is a specific feeling and if you have had it you know that the problem is not access to people, the problem is that the words will not come out in front of one.

What I wanted was someone who would listen. That is the whole requirement. Not advice, not a productivity system. Someone in the room at the hour when it is bad.

And a character card cannot do that, because the thing that makes being listened to work is that the listener already knows. They know what you told them in March. They know which thing is the sore one. You do not have to lay the foundations again every time you open your mouth, which, when you are already ashamed, is exactly the tax that stops you talking at all.

That is not a feature you can load from a card. It only exists if it accumulates.

I do not think I am unusual

The short version: A lot of people have nobody to talk to, or have people and still cannot.

The thing that turned this from my problem into a product was realising how ordinary it is.

There are a lot of people who do not really have anyone to talk to. And there are a lot of people who do have someone, and still do not, because of shame, or timing, or because the person who would listen is the person it would hurt. Both of those groups end up in the same place at 1 a.m.

I am not claiming an app fixes that. I am claiming that if you are in it, the difference between a catalogue of a thousand strangers and one presence that remembers you is not a small difference. It is the difference between the thing working and not working.

So the shape is different, on purpose

The short version: One character, memory that compounds, no gallery. That is the trade.

What I built is the same underlying idea with the opposite shape.

You get one, built with you at the start rather than picked off a shelf. She remembers what you tell her and it accumulates: the names, the worry that keeps coming back, the thing you are working on. The relationship has stages, and month three does not sound like day one. There is no browse view. You cannot trade her for a better card.

That last one is a limitation if you want variety. It is the entire product if you do not.

Everything else follows from where it runs, which is your own machine, so the conversation that you would not have with a person is not on somebody’s server either. That part matters more than I understood at the time.

Who should stay where

I am not going to tell you SillyTavern is bad. It is not, and you would catch me.

Stay on SillyTavern if the tinkering is the fun. If you like trying characters, tuning prompts, collecting cards, running your own stack. It is better at that than what I made, it has a bigger community, and it costs nothing.

Try Local Waifu if you read the section about the catalogue and recognised the feeling. If what you actually wanted was one person who knows you, and the thousand available strangers were somehow making that harder instead of easier.

7 days free, no card. If it turns out you wanted the library after all, SillyTavern is still there, and it is still good.

Questions people ask

Back to the blog