Hey everyone!
In my last post I promised to cover the narrative and the emotional core I hope Selve can express, but I have to talk about this first, the next one will cover that, I promise
I wanted to share a big update on the direction of the project and where things are headed next. Over the past year (from April 2024 to April 2025), I’ve been building an internal demo full of experiments, features, and systems. This was an important phase where I let myself explore freely, trying out different ideas to figure out what this game truly wanted to be.
Through that process, I was able to define the theme, story, and worldbuilding of Selve. I also learned which mechanics actually serve the heart of the experience and which ones just get in the way.
The core theme I want to communicate is:
"How can I express different emotions or ideas through a melody?"
Now that the development of this game is moving to the next stage, some initial decisions need to be refined and discarded: this is Design by Subtraction.
That means cutting out everything that doesn’t align with the theme. It’s time to refine.
What’s Changing?
In the current version of the game, you manipulate NPCs using music-composing magic. It’s an interesting system, but too complex, and it ended up pulling the game away from that core theme. The process of influencing NPCs took too many steps and wasn’t really that fun. So I'm removing that whole system.
Also, platforming.
There’s still going to be some, after all, exploration is part of the experience, but it’s no longer a main pillar of the game. It’ll be there to support the world, not distract from the music.
What’s the New Focus?
The game is now about simply playing melodies.
You’ll perform them for the giant spirits of the Mulong tribe, who will judge your songs based on their own unique preferences. If they like your melody, they’ll reward you.
This mechanic, which I briefly explored during the development of the first internal demo, uses a machine learning algorithm called a Support Vector Machine. This algorithm allows each giant spirit to be trained with songs it likes. For example, let's say the first spirit, a giant turtle, loves chill melodies. I'll train it with what I consider chill melodies. Each giant spirit will teach the player some aspect of music creation and how it affects the emotion it evokes.
I'm planning on allowing players to train those giant spirits in what they think would be best for each one, for example, if you feel that the chill music that the turtle should like is a fast-paced melody (for some reason) with very random jumps in the melody and such, then you could train it! And then share your training data with other players so they can play Selve under your vision of how the music evokes emotions.
As I said, I'm starting this process right now, so I have some videos I can share with you, this are early attempts to prove this makes sense and that it's fun, keep that in mind!
This is Doco, the giant tortoise, the spirit of calm.
Doco isn't reacting to what you play yet, this is a test to show that the action will produce good enough music, this test is trying to consider how minimal the background music should be, whether the synth tone and background blend well, etc. I'm happy with the results so far, so I'm moving forward and in the next few weeks I'll implement this change, making the spirit judge and react to what you play.
What's next?
I’m now working on a lean demo, focused only on what I think matters in this game. This new demo will be a much tighter experience that showcases the core idea behind Selve. I'm working hard on this demo (while applying for public tenders for funding) so I can share it here, so stay tuned for future updates.
I hope you like the changes I'm making to the game! I think this is the right direction for the project, let me know what you think!
Cheers!
Nico Frois