AfterGlow MIDI continuation¶
AfterGlow continues a MIDI piano phrase using a compatible Aria GGUF model. It generates MIDI events, not an audio waveform. Use a piano instrument or synthesizer to render the resulting MIDI as sound.
Continue a phrase¶
spum music --checkpoint aria.gguf --input phrase.mid \
--output continuation --tokens 192 --temperature 0.95 --seed 42
The output directory receives the continuation artifacts; full.mid contains the combined phrase and continuation. Keep an original copy of your input performance.
| Option | Meaning |
|---|---|
--checkpoint FILE |
Compatible Aria GGUF checkpoint |
--input FILE |
Input MIDI file |
--output DIRECTORY |
Output directory |
--tokens N |
Requested generation length, default 192 tokens |
--temperature T |
Sampling temperature, default 0.95 |
--seed N |
Sampling seed |
--cpu |
Select CPU execution |
--serve |
Read one JSON request per stdin line and emit one JSON reply per line |
Reuse a loaded model¶
Send a request line:
The reply includes success status, output path and generation measurements. This JSON-lines interface is separate from the HTTP chat server. It reads MIDI files from local paths; it is not a microphone or audio-transcription service.
Research interpretation¶
A musically coherent continuation does not demonstrate mood recognition or therapeutic efficacy. Evaluate those parts of a physical piano system separately. Model licensing and any permitted use of performances remain the integrating application's responsibility.