> For the complete documentation index, see [llms.txt](https://musenai.gitbook.io/musenai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://musenai.gitbook.io/musenai/start-here/what-is-musen.md).

# What is musen?

Version: v4.0.0\
Last updated: September 2026

### 1. musen is AI Radio

musen is a continuous personal radio shaped by what you ask for and how you listen.

The current product is free on iPhone and Android. Press play and your Live Radio begins. You can change its direction with a typed Request, your voice, or a picture, then leave the radio to continue without building a playlist or managing a queue. An optional AI Host can add short spoken moments between songs. ([musen](https://musen.live/?utm_source=chatgpt.com))

The music itself comes from existing recordings made available under the appropriate rights arrangements. musen uses AI to understand intent, select and sequence music, preserve continuity and decide how the radio should evolve. It does not currently generate the songs you hear. ([musen](https://musen.live/about?utm_source=chatgpt.com))

The simplest description is:

**Press play. Tell the radio what the moment needs. Let it take care of what comes next.**

### 2. Radio is an experience unfolding through time

Most digital music products begin with individual items.

A song. An album. A playlist. A search result. A recommendation.

musen begins with the radio itself.

A good listening experience depends not only on whether an individual song is relevant, but also on what came before it, what the listener wants now, how quickly the energy should change, whether speech would help or interrupt, and where the experience should go next.

That leads to the central question behind musen:

**What should be happening in this radio right now?**

The musen White Paper defines AI Radio around this continuous objective: the system maintains an evolving listening state, reasons over permitted context, expressed intent, prior observations and governing constraints, then continues adapting as the experience unfolds.

Radio is therefore not treated as a sequence of isolated recommendations.

**Radio is a computation unfolding over time.**

### 3. The listener gives direction, not instructions for every song

musen is designed around intent.

You might ask for:

> Funk with enough energy to keep me focused for the next hour.

Or:

> Make this drive feel like the beginning of a movie.

Or simply show the radio a picture and let the scene become part of the direction.

A Request does not produce a page of results. It influences what the radio does next. The radio can change course while preserving the experience already in progress. ([musen](https://musen.live/how-it-works?utm_source=chatgpt.com))

This matters because musen is not trying to turn natural language into another search interface.

The listener expresses what the moment needs.

The radio remains responsible for the radio.

### 4. Taste and moment are different

A person's general taste is useful, but it is not enough to describe what they want at every moment.

Someone can love techno and want jazz on Sunday morning. They can listen to ambient music for a week without becoming an ambient music listener forever.

musen therefore uses different kinds of signals for different purposes.

#### Train Radio

Explicit choices help establish broad taste preferences.

#### Requests

Text, voice and picture Requests express what the listener wants now.

#### Listening

What someone stays with, when they listen and how the radio unfolds can help the system understand what works over time. ([musen](https://musen.live/blog/how-musen-trains-itself?utm_source=chatgpt.com))

The objective is gradual adaptation without requiring constant interaction.

### 5. Memory creates continuity

A personal radio becomes more useful when it does not need to meet the listener for the first time at the beginning of every session.

musen can preserve relevant listening context and preferences across sessions so that future radio can begin with more understanding than the session before it.

But memory is not intended to become a permanent definition of a person.

Preferences change. Context changes. Requests can override previous direction. Old information can lose relevance.

The governing principle is:

**Memory should create continuity, not captivity.**

The White Paper treats memory as governed state rather than unrestricted accumulation. Scope, retention, correction, forgetting and isolation all matter because stale or uncorrectable memory can make an adaptive radio progressively less adaptive.

### 6. AI orchestrates the radio

AI in musen is primarily used to shape the listening experience.

It can help interpret intent, understand relevant context, reason about pacing and continuity, select among eligible music and determine how the experience should evolve.

But intelligence does not automatically become authority.

A model may propose what should happen. Authoritative musen systems remain responsible for the state, rights, policy and execution constraints that determine what can actually happen.

This distinction is part of the broader AI Radio architecture:

**Fujin remembers. Intelligence proposes. Runtime commits. Admiral executes.**

The detailed internal mechanisms are intentionally outside the scope of the public White Paper, but the governing idea is simple: a probabilistic system can help make editorial decisions without becoming the source of legal, rights or system truth.

### 7. The AI Host is part of the radio, not a chatbot

The AI Host is optional synthetic speech inside Live Radio.

It can introduce tracks, add short spoken moments and bring another layer of personality to the stream. The listener can configure it or switch it off completely for music only. ([musen](https://musen.live/how-it-works?utm_source=chatgpt.com))

The Host is not the system controlling musen and it is not a conversational assistant that requires the listener to keep talking.

It lives inside the radio.

The distinction is important: AI Host speech is an expressive layer. It does not gain rights, system or professional authority by speaking.

### 8. Artists remain the authors

musen uses AI to orchestrate culture, not replace the people who create it.

The current product plays existing music recordings rather than generated songs. AI helps decide how music should be selected, sequenced and experienced, while the recordings themselves remain the work of artists. ([musen](https://musen.live/about?utm_source=chatgpt.com))

That leads to another governing principle:

**Rights before reach.**

Having access to music is not the same as having permission to use it. Storage does not imply streaming. Streaming does not imply analysis. Analysis does not imply model training.

AI Radio only works as a credible medium if culture, permission and attribution remain visible rather than becoming implementation details.

### 9. What exists today

As of September 2026, the public musen product is deliberately focused.

**Available now**

• Free personal Live Radio\
• iPhone and Android apps\
• Account based continuity\
• Text Requests\
• Voice Requests\
• Picture Requests\
• Train Radio taste controls\
• Listening based personalisation\
• Optional AI Host\
• Existing music recordings\
• Continuous radio playback ([musen](https://musen.live/?utm_source=chatgpt.com))

Other ideas around creator participation, broader catalogue infrastructure, web access, developer interfaces, machine interfaces, attributable listening economics and Large Radio Models sit at different stages of planning, testing or research. They are not current product capabilities merely because they exist in musen's architecture or research programme.

### 10. The idea stays simple

musen is building a radio that can understand direction without requiring continuous management.

You should be able to press play, tell the radio what the moment needs, change direction when you want to, and then return your attention to whatever you were doing.

The interface can change. The models can change. The catalogue can grow. The architecture can evolve.

The invariant is the radio:

**continuous, coherent listening over time.**

That is musen.
