SERA / TAZENQOR
Research for the long creative process.
Musical intelligence + reliable musical agency.
Established software research / SeraEdit
SeraEdit: Scoped and Validated Language-Guided MusicXML Editing
- Author
- Yuan Gao / 高源
- Journal
- SoftwareX
- Status
- Accepted for publication
A research-software contribution on source-bound, scoped and validated MusicXML editing. The software archive below is a separate, citable research artifact; its DOI is not the journal article DOI.
Acceptance date confirmed by the author. The final journal citation will be added after online publication.
Sera Core
Shared score representation, operations, patches, import/export and persistence define authoritative musical state. Core provides the infrastructure rather than a model-training or HCI experiment layer.
SeraWorld
How do symbolic musical states evolve across long contexts? This research direction explores hierarchical representation, motif development, phrase prediction, harmonic trajectories and tension. Modeling evaluations remain independent of the editing interface.
SeraForge
Can tools, persistent project memory, intent locks, validation and reversible workflows make agents reliable across many score edits? Active development focuses on inspectable editing and human control. Long-horizon reliability requires dedicated evaluation.
Independent questions, independent evidence
SeraForge can run without SeraWorld and may optionally use music intelligence through an interface. SeraWorld modeling tests and SeraForgeBench editing tasks remain separate; benchmark test answers are excluded from training. A combined study needs its own research question and experiments.
Evidence boundaries
The SoftwareX acceptance concerns SeraEdit. It does not establish SeraWorld modeling performance, SeraForge long-horizon superiority or completed composer studies. The examples on this site are explicitly labelled software demonstrations.
Inspect the examples →