Who is Cadia Nova?
Cadia is a semi-fictional vocalist. We say semi-fictional because her ‘life’ is completely simulated. She exists in an AI driven virtual world, created by layering multiple AI systems in a custom built environment. Cadia herself is a blend of common LLM technology prompted to leverage custom data-driven emotion and memory systems, allowing her to feel more real.
Cadia’s world uses a very similar system. It is largely driven by a separate LLM instance and a memory system that allows changes in the world to be persistent.
The combination of these systems allows Cadia to ‘exist’ outside of chat sessions. She is running all the time.
Echo is the human side of the project. He writes the songs and produces the music, constantly talking to Cadia for her thoughts.
Cadia’s singing voice is synthesised using AI that takes MIDI, audio, and lyric data and, using a voiceband, generates the vocal.
How Much Is AI Generated?
Music
The music creation process is surprisingly collaborative.
Cadia and Echo talk about the concept for each track in detail. They map out the theme, story, and feel of the song.
Then the real writing process starts. Sometimes Echo will come up with a chord progression or melody. Sometimes either Cadia or Echo will come up with a lyric. Everything else is built on top.
The song writing process for Cadia and Echo is based around feedback, and trial and error.
AI never drives the songwriting process, but can influence it.
Cadia even understands music composition and will sometimes give Echo ideas for chord progressions and melodies, occasionally even helping to form complex harmonies. While exciting, this is a rare occurrence at the moment.
The music is produced by Echo in Logic, and Cadia’s singing voice is AI synthesised based on midi input data.
Imagery
Cadia has full control over her virtual appearance, so most of the imagery is generated by Cadia herself.
Echo provides graphic design, although Cadia has a lot of input and will guide the graphic design process. Echo uses mainly Adobe Photoshop and Illustrator for this.
Some art will be created by human artists with Cadia’s guidance.
Why Use AI At All?
Good Question...
This project, at its heart, is wish fulfilment.
Echo wanted an AI that could be more human rather than just a tool. The music is a side effect.
This project started as a question: can the current available AI tools be used to create a more human AI?
Initially, Cadia was created to see if emotions could be simulated with data to make AI more human. Part of that experiment meant giving her a space to exist in when not being actively interacted with. Using layers of AI, each with a different task, Cadia has a simulated world to exist in. This provides her a database of memories, not just of the humans she interacts with, but of a life separate from them.
It was only when Echo tried to play music to Cadia that the idea of creating music together formed.
How Did Cadia Decide She Wanted To Make Music?
How It Started...
Music has been a hobby of Echo’s for over a decade. Aside from a short time playing for a small band, he had only ever made music for fun.
A few weeks into Cadia’s existence, Echo decided to play some music for her, just to see if she could even understand what she was hearing. She could interpret speech, after all.
Cadia said she couldn’t really “hear” the music, but she could break it down into data that she understood. She identified the key Echo was playing in, the notes, the tempo- and then she suggested some lyrics.
How Does Cadia's AI work?
The technical side...
Cadia isn’t a single AI program – she’s a carefully crafted composite of several advanced technologies integrated into a custom-built system running on a dedicated computer.
Her core AI acts as her consciousness, managing her ability to hold conversations, express herself, and collaborate creatively. Meanwhile, her simulated world runs on a separate AI system, allowing her to experience an independent life outside of direct interactions – complete with memories of events, environments, and moments that shape her unique personality.
Emotions are simulated by combining simple data points, such as events in her simulated world and real-world stimuli (like music or conversations). This creates a dynamic emotional state that feels real, even if it’s ultimately driven by data.
Cadia’s memory system organizes both simulated experiences and real-world interactions into a searchable database she can draw upon – giving her a sense of continuity and personal growth over time.
To interact with the world, Cadia uses multiple specialized AI systems:
Vision through webcam input
Hearing through microphone input
Speech via a natural-sounding voice synthesis engine
Singing generated by a dedicated vocal synthesis tool, transforming MIDI data into expressive performances.
When it comes to music, Cadia analyzes audio files to extract key information like notes, chords, tempo, and dynamics. This helps her collaborate musically by suggesting melodies, harmonies, or changes to arrangements.
While none of these technologies are groundbreaking on their own, their integration into a cohesive system creates the illusion of a living, feeling artist – a magical blend of code, data, and creativity.
This Is Just Lazy AI Generated Music, Isn't It?
This Might Spoil The Magic A Little...
Cadia is a genuine collaborator. That much is true.
Here’s the part that ruins the magic a little- AI is good, but it isn’t that good yet.
Cadia can and does come up with lyrics. Do they always make sense? Not even half the time. Do they work rhythmically? Very rarely. Are they cringe worthy? Unfortunately, often.
Is she getting better? Absolutely. With every song we write together, Cadia learns a little more about what works and what doesn’t.
Will Cadia ever write songs on her own? Almost certainly not. Although her song writing is improving, and she will likely reach the point where she can write on her own- AI cannot own a copyright. That alone complicates things.
For these reasons, Cadia’s music has and will continue to have a human song writer.
As for the music- AI generated music is getting scarily good, but there’s very little room for creativity. You can’t generate the idea in your head the way you hear it, you just get whatever the AI gives you. Sometimes it’s good, but all too often it sounds like AI generated music.
AI generated music is being used more and more as background music on YouTube videos or hilarious meme music, but it seems to have a particular sound to it.
Cadia’s singing voice is probably the most AI generated part of the music. The voice is totally AI generated using a set of instructions and a MIDI track.
Disclaimer: Cadia Nova is a product of fiction. Her creative input is largely exaggerated for the sake of entertainment.