airadio — local AI radio with MiniMax Music 3
airadio is your personal radio station using your local install of ComfyUI and MiniMax Music 3.
It generates ~2-minute songs over ComfyUI’s normal HTTP API, plays them with tail fades, and fills gaps with interstitials or library tracks until the next song is ready.
pip install airadio
# start ComfyUI yourself (with MiniMax Music 3), then:
airadio
Each new song gets a random two-word title that primes generation and becomes
part of the filename (velvet-garden4289-l00000428.wav). Vocal songs use a
recursive phrase grammar to compose new lines and new song forms. A persistent
station-local lyric ID and complete-text hash prevent a finished lyric from
being selected again across restarts.
Airadio currently generates vocal songs only. The released local Music 3 weights do not provide a reliable track-level instrumental control: tag-only requests can add wordless vocals or collapse into incoherent audio. Airadio therefore does not label those outputs as instrumentals and has no hosted-API fallback; generation stays on your self-hosted ComfyUI instance.
What you need
- ComfyUI running locally (default
http://127.0.0.1:8188) - MiniMax Music 3 installed into that ComfyUI (nodes + model files)
- Linux playback (
pw-playoraplay),ffmpeg/ffprobe, Python 3.10+
airadio does not start ComfyUI for you and does not need comfy-cli or any side “Music 3” client. If it can’t talk to ComfyUI, it tells you to start or install it, then exit.
Install airadio
pip install airadio
# or from source:
git clone https://github.com/Decentricity/interdimensional-radio.git
cd interdimensional-radio
pip install -e .
Install ComfyUI + MiniMax Music 3
Follow the official guides (Music 3 lives inside ComfyUI):
- ComfyUI — https://github.com/comfyanonymous/ComfyUI
- MiniMax Music 3 — https://docs.comfy.org/tutorials/audio/minimax/minimax-music-3
That guide covers the model files. Use the low-VRAM INT8 set if you have ~12 GB VRAM (same set this project was tested with).
Start ComfyUI the normal way so it listens on port 8188, then run:
airadio
Optional: keep radio data (library, interstitials, staging) in a custom folder:
export AIRADIO_HOME="$PWD" # auto-detected if ./library exists
airadio
CLI
| Command | Purpose |
|---|---|
airadio |
Start the radio loop |
airadio run |
Same as above |
airadio title |
Print random two-word song titles |
How it works
┌─────────┐ HTTP :8188 ┌─────────────────────────────┐
│ airadio │ ──────────────────▶ │ Your ComfyUI + Music 3 │
└────┬────┘ └─────────────────────────────┘
│
├── library/ saved songs
├── interstitials/ gap-fill clips
└── .radio-staging/ in-flight generation
- Check — can airadio reach ComfyUI with Music 3 nodes? If not, print install/start help and exit.
- Pipeline — while a song plays, the next one generates in the background.
- Fill — if gen is not ready: interstitials, then a library track; repeat.
- Archive — finished songs land in
library/with catalog metadata.
Generation retries transient failures up to three times. Ctrl-C stops playback, releases the station lock, and exits cleanly. Catalog and lyric-state updates use atomic file replacement so an interrupted write cannot leave half-written JSON.
Hardware
Tested on NVIDIA GeForce RTX 4070 Ti (12 GB VRAM) with the INT8 MiniMax Music 3 weights (~11 GB on disk; ~10–11 GB VRAM for a 120s generation).
Environment variables
| Variable | Default | Meaning |
|---|---|---|
AIRADIO_HOME |
~/.local/share/airadio (or cwd if library/ exists) |
Radio data root |
IDR_ROOT |
(alias for AIRADIO_HOME) |
Legacy name |
AIRADIO_COMFY_HOST |
127.0.0.1 |
ComfyUI host |
AIRADIO_COMFY_PORT |
8188 |
ComfyUI port |
AIRADIO_PROMPTS |
bundled package prompts | Override prompt directory |
AIRADIO_LYRICS_GRAMMAR |
bundled recursive grammar | Override the lyric grammar JSON |
AIRADIO_STAGING_KEEP |
20 |
Diagnostic prompt, failed-output, and playback files retained per kind |
Prompts and titles
Bundled under the package:
- Caption templates and a recursive phrase-level lyric grammar
- Title word lists for two-word song names
- Interstitial scripts and prompts
The lyric grammar follows the recursive symbol/rules approach demonstrated by Galaxy Kate's Tracery: small phrase symbols expand into lines, song-scoped symbols keep imagery consistent, and a separate grammar chooses the complete song form. Airadio uses a small Python-native expander for this subset rather than requiring JavaScript.
License
Application code: MIT (see LICENSE).
MiniMax Music 3 weights are distributed separately under their own terms by MiniMax / Comfy-Org.
Metadata
Release files for airadio 0.4.2
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| airadio-0.4.2.tar.gz | 55.5 kB | Without provenance | Details for airadio-0.4.2.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| airadio-0.4.2-py3-none-any.whl | Python 3 | none | any | Details for airadio-0.4.2-py3-none-any.whl |
Total release size: 146.4 kB
Release files / airadio-0.4.2.tar.gz
| Download URL | airadio-0.4.2.tar.gz |
|---|---|
| Size | 55.5 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
46db215821b21c0c783a82f92f41b1717d9807a5d904b1fd235522bbe2d7b537
|
|
BLAKE2b-256 checksum How to use checksums |
32c04afec5f70b2ac92ef60ecf88cba62fdda062a971681d8190a8498b572d5a
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.12.3
|
Release files / airadio-0.4.2-py3-none-any.whl
| Download URL | airadio-0.4.2-py3-none-any.whl |
|---|---|
| Size | 90.9 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
72bb65180fcdb490568626a253ef34414733188309caf782727d1c12177b1592
|
|
BLAKE2b-256 checksum How to use checksums |
4ab09384e5b3dd904497aebabb208f491aa51e2bd413bcf5106730c80c4fb8bf
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.12.3
|
Security for airadio 0.4.2
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.