AI Radio: A way to inject radio-like sections to your MA playlists

[swiftbird07]

Hello guys,

just thought it would be fun to share the project I worked on over the weekend.

It is a python toolchain that enables me to define a specific MA playlist and the tool will generate the text and respective TTS mp3 files and insert them into a new playlist making it more like a radio like experience with weather reports, song intro and outros, news etc.

Using config files I can define what playlist to use, what the prompts are for each section and under what circumstances I want to add what section at which position.

I also recently added support for a "dynamic generation" mode where the tool doesn't create any new playlist but just add them to the current player queue one after another (mainly to save costs on the OpenAI billing 😅).

As said this currently works as a standalone script chain and is not integrated directly into MA as a plugin or something - mainly because I didn't research yet if and how this is possible. If there is interest I may consider doing more research in that regard.

I hope you find this project interesting. At least for me it's very funny to have my own personal radio station now commenting on my songs, giving me fun facts about artists etc. (and most importantly there are no annoying AD breaks!).

Thanks to the Music Assistant developers for making such a nice platform (and the SDK/API that goes with it) otherwise it would have been way harder to develop a tool like this!

Link: https://github.com/swiftbird07/AI-Radio

If you have any questions or feedback just ask :)

[MarvinSchenkel]

Nice work! It would be really nice if we could turn it into a Music Assistant plugin? That would allow people to just add it to MA without any extra steps. For example, someone is working on party mode that you could use as inspiration on how to inject queue items from within a plugin. Of course we are here to assist in case you have questions or need help 🙏

[swiftbird07]

@MarvinSchenkel
Hello, thanks for the encouragement! I am working on a plugin right now.

But I indeed have two questions: First, for this plugin I would need to modify server and frontend components at the same time (server for the AI radio logic and frontend to implement it into the interface). How would you like me to start the contribution process: Should I open a PR on the server component first and wait for the approval of the backend logic and then open a PR with the frontend stuff or should I open PRs for both simultaneously?

I have to implement frontend changes as plugins in the current form can't add UI elements or context menus, needed for AI radio, by themselves.

This brings me to my second question: Should I overhaul the frontend logic so that plugins in general can modify it by providing e.g. html / css assets or should I just 'hardcode' the plugin into the frontend and check on load if the plugin is installed + enabled (and only then display its UI)? Or is there another way I am not seeing?

Thanks in advance
Martin

[MarvinSchenkel]

Sorry about the delay, I am not actively monitoring the Discussions sections.

I would recommend to just develop them simultaneously. You can have both repos checked out and simply install the new frontend in the server repo by running uv pip install ../frontend. This will allow you to basically develop both side by side and test it as a whole

As for the frontend changes, we do like to keep them isolated as much as possible. For frontend, it's a bit harder than for the backend. This will be the second 'plugin' that also involves frontend changes, so we are still figuring this out ourselves. Your best example is to look at how party mode does this:

• Frontend: Feature: Party mode frontend#1344
• Backend: Party mode  server#2911

For example, party mode uses the extra_attributes to pass information.

Could you jump on Discord and start a topic in the dev-support section? We'll help you along the way and come up with something clever for the frontend!