Integrate AudioMuse-AI for Sonic Similarity advanced functionality

[NeptuneHub]

Hi,
I'm the AudioMuse-AI developer and I just saw that you're implementing Sonic Analysis by reading those page:

• https://www.music-assistant.io/audio-analysis/sonic-analysis/
• https://www.music-assistant.io/blog/2026/06/10/music-assistant-2-9/#good-vibrations

I'm writing because AudioMuse-AI is free and opensource and it work specifically on sonic analysis in the last year. I would like to ask if you can be interested in some sort of integration of AudioMuse-AI in Music Assistant to don't have to reinvent the wheel. Or eventually to be added as an alternatives for those user that already work on that music server.

It support multiple music server like Jellyfin, Navidrome (and Open Subonic one), Emby and Lyrion.

It also support multiple Sonic Analysis functionality like:

• Clustering: Automatically groups sonically similar songs, creating genre-defying playlists based on the music's actual sound.
• Instant Playlists: Simply tell the AI what you want to hear—like "high-tempo, low-energy music" and it will instantly generate a playlist for you.
• Music Map: Discover your music collection visually with a vibrant, genre-based 2D map.
• Playlist from Similar Songs: Pick a track you love, and AudioMuse-AI will find all the songs in your library that share its sonic signature, creating a new discovery playlist.
• Song Paths: Create a seamless listening journey between two songs. AudioMuse-AI finds the perfect tracks to bridge the sonic gap.
• Sonic Fingerprint: Generates playlists based on your listening habits, finding tracks similar to what you've been playing most often.
• Song Alchemy: Mix your ideal vibe, mark tracks as "ADD" or "SUBTRACT" to get a curated playlist and a 2D preview. Export the final selection directly to your media server.
• Text Search: search your song with simple text that can contains mood, instruments and genre like calm piano songs.
• Lyrics Search: search your library by theme, story or meaning, like love songs, not just the sound.

It work both in a container version or can be natively be installed on Windows, MacOS and Linux.

It have multiple API that can be easy to be integrated, and if needed other integration can be developed off course.

Let me know what do you think and if there is the possibility to make some integration in the future.
Thanks!

[chrisuthe]

Hi @NeptuneHub!

Music Assistant is open for a reason, you should absolutely integrate AudioMuse-AI with it! AudioMuse-AI would make an awesome plugin, which can then provide things like similar tracks, similar artists, similar albums and a search provider. The other pieces (Music Maps, song paths, Lyrics Search) could be added either as part of your plugins interface (IE: not native MA functions) or MA could be expanded to make use of them, if people agree.

Note that Music Assistant already has explicit places it uses things like "similar track" where providers (like an AudioMuse-AI plugin) are handed a track and asked for similar tracks. Radio mode, don't stop the music, the similar track listing, smart playlist functionality, etc.

I suggest you start by taking a look at Smart Playlists, and how similar track is used throughout the system, but a plugin for users using AudioMuse-AI seems like an easy fit, as a community maintainer: more options more better.

The Sonic Analysis provider I added to music assistant is specifically meant to be on device and with no additional work by users. It's not as complete as AudioMuse-AI though most of what your system does is on the drawing board for the native sonic analysis too, for full transparency.

A note on the Sonic Similarity/Analysis in MA: It's broken up into two sections: The Audio Analysis portion (MA has support for N number of Audio Analysis providers: things like smart fades that handles beats and keys for fading between tracks, Acoustid lookup for getting tracks tagged properly, and Sonic Analysis which handles what smart_fades doesn't sonic signature wise. The second part is the plugin (Sonic Similarity) which takes the Audio Analysis data (in this case, from both the Sonic Analysis and Smart Fades providers) and make use of them, exposing similar_track, exposing a search function, a discover row, etc. The plugin builds two indexes from data and uses them to provide the output without actually shipping any custom front end pieces (except the settings page).

AudioMuse-AI could VERY easily copy/paste the Sonic Similarity plugin, rename it AudioMuse-AI and just replumb it to hit API's. No need for you to provide an Audio Analysis provider since you're keeping all the data in AudioMuse-AI.

Let me know if you want help doing so!

[dtpdx]

Please please do integrate audiomuse-AI into music assistant. I have been using audio assistant AI for my library and it is amazing. I also am looking how to integrate my navidrome server onto my home devices and with integration with Music assistant, just seems like pieces of puzzle fitting together. very complementary, force multiplier kind of thing... Please do this!

[chrisuthe]

@NeptuneHub - here is a branch where I've duplicated the plugin, and set it up to use AudioMuse-AI API's:

https://github.com/chrisuthe/server/tree/audiomuse-ai

I haven't ran it (no audioMuse in my stack at the moment) but if you or someone else wants to take it, test, build, add etc and then open a PR it's a solid base!

[NeptuneHub]

This sounds very good, if you arrive at a point of having a built image to test I’m off course ready to support!

Meanwhile I saw that MA already use getSimilarSongs from Navidrome, and the AudioMuse-AI Navidrome plugin override this API. I need to do a test for confirm but this should mean that the similar song functionality will came for free using Navidrome, AudioMuse-AI, AudioMuse-AI Navidrome plugin. It’s not the full set of feature but will be a nice start.

On Jellyfin plugin similar song functionality it doesn’t come for free because the AudioMuse-AI plugin override the /Item/id/InstantMix endpoint (that is the same used from Jellyfin web frontend when you click InstantMix) instead MA use /Item/id/Similar

Will be even more great if we are able to integrate more functionality. Let's think for a minute in a scenario where you have:

Alexa -> Music Assistant > Navidrome/Jellyfin -> AudioMuse-AI

Where you're able to tell alexa "Play a song path between song A and song B" or "Play song that talk about "Summer Holiday", it will be great. It will be the power of open source against closed source.

[NeptuneHub]

So I did a test in this way:

• Navidrome 62.0
• AudioMuse-AI Navidrome Plugin v8
• AudioMuse-AI last v3.2.3
• Music Assistant container latest

On Music Assistant when you click on a track you have the Similar Track section, it callgetSimilarTrack of navidrome, Navidrome then call the AudioMuse-AI API, in fact you have this in Navidrome logs:

time="2026-06-25T11:08:58Z" level=info msg="[AudioMuse] Calling similar_tracks API: http://192.168.3.204:8000/api/similar_tracks?eliminate_duplicates=true&item_id=fSGwI8ZqhrPWvSXAMuSzey&n=25&radius_similarity=true" plugin=audiomuseai

and this on AudioMuse-AI flask container log:

10.42.0.1 - - [25/Jun/2026:13:08:58 +0200] "GET /api/similar_tracks?eliminate_duplicates=true&item_id=fSGwI8ZqhrPWvSXAMuSzey&n=25&radius_similarity=true HTTP/1.1" 200 9709 "-" "Go-http-client/1.1"    

For AudioMuse-AI, instead of populate the similar track section, it would be possible to add a functionality (a button and/or a menu voice) that directly enqueue song and start playing, like when on Navidrome you click on a song and then select Instant Mix?

Will be also useful a similar apporch also for Jellyfin by calling /Item/id/InstantMix

Then off course if it possible to integrate even more functionality by an ad hoc Music Assistant plugin that talk directly with AudioMuse-AI, even better. But the two things above could be an easy win.

What do you think?

[NeptuneHub]

Just to add that my test went futher:

• I installed also home assistant
• I added Music Assistant on home assistant
• I added on home assistant the blueprint: https://github.com/music-assistant/voice-support/blob/main/local-assist-blueprint/mass_assist_blueprint_en.yaml
• I installed Music Assistant Desktop App

What is missing here is that there is no way to say to the blueprint to call the getSimilarTrack of navidrome.

@chrisuthe On my side I don't know where to put mu hand on the plugin part. But if you, or someone, is able to do this and provide a container build with the change I can defintly help with test!

[dtpdx]

This sounds incredible!!

[Chaphasilor]

Just wanted to chime in to note that I'm also a big fan of AudioMuse and helped with fine-tuning some of the sonic analysis algorithms and alpha testing. I'm also a Music Assistant user, although my main library lives in Jellyfin.
One personal benefit would be that I already have AudioMuse set up (on my home server / NAS) and connected to Jellyfin, including a full sonic (& co) analysis of my entire library. If I wanted to use the existing sonic similarity support within Music Assistant, I'd have to re-analyze my entire library, but this time via the (comparatively weak) Raspberry Pi 5 running Home Assistant, and over the network instead of via directly-attached storage.

So while I love that Music Assistant now has a super simple way to add sonic analysis for a personal library, AudioMuse support would provide a really nice migration path for existing Jellyfin/Navidrome/Emby/Lyrion users, that either want to integrate their existing media server with Music Assistant, or fully replace their media server with Music Assistant.

Also, it's incredible that you already whipped up a proof of concept plugin for AudioMuse, that really helps kickstart things. I'll give that a try, if I can get it installed <3

[morgancurrie]

Same situation here. I spent a month analyzing 26k tracks on low-powered hardware, and I would prefer to keep this functionality separate from Music Assistant. I also don't want to add more memory to my previously performant Home Assistant setup. Audiomuse already supports all of our phones' various music clients, and our kitchen dashboard which uses Symfonium. I would love to help test this or help in any way if you need more eyes/hands on any of this :)

[NeptuneHub]

@chrisuthe I just forked your code from here:

• https://github.com/chrisuthe/server/tree/audiomuse-ai

And did some change to have it working (tested with Jellyfin), here the PR:

• Audiomuse ai plugin chrisuthe/server#2

Can you give a check and let me know if something is missing to have it merged?

Thanks!

[NeptuneHub]

Hi all,
is there anythik helse needed to proceed to have the AudioMuse-AI provider merged?