Optional metadata enrichment, built to your model's needs. Available at any licensing tier.
Every Orpheus engagement is different. These are common starting points.
Text descriptions paired with audio, such as "airy tone" or "ornamented phrase," for models that respond to prompts.
Note-level MIDI and timing aligned to the audio, for transcription and symbolic modeling.
Additional fields for dynamics and texture beyond the standard metadata.
Curated subsets and expected behaviors for testing how well a model follows intent.
Tell us what your model should learn or be tested on, and which datasets you're licensing.
HFA defines the fields, coverage, and format, and prices the enrichment as part of your license.
The enriched metadata ships alongside the dataset, with the same provenance and checksums.
Describe what your model needs to learn, and we'll scope the enrichment to fit.