Downloads · 30 days
0
vital-jepa-v7
duckking032/vital-jepa-v7
vital-jepa-v7 is a machine learning model from duckking032. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as apache-2.0.
Stereo 22.05 kHz, six-second audio encoder initialized from V6 and trained on a mixture of native Vital interventions and synthetic one-shot articulations. V7 adds explicit articulation and pitch objectives while reta…
Access
Public
Updated Jul 12, 2026
Repo size
146 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.pt146 MB · 100%
At a glance
- License
- apache-2.0
- Access
- Public
- Created
- Jul 12, 2026
- Updated
- Jul 12, 2026
- SHA
- e5ebdfb0
Model card
From the Hugging Face model README
Vital JEPA V7
Stereo 22.05 kHz, six-second audio encoder initialized from V6 and trained on a mixture of native Vital interventions and synthetic one-shot articulations. V7 adds explicit articulation and pitch objectives while retaining acoustic-descriptor and exact-control intervention losses.
The selected checkpoint is step 16,000: it was chosen as the best balanced checkpoint across reconstruction, acoustic correlation, control identification/direction, retrieval, articulation, and pitch probes.
See evaluation/ for checkpoint comparison and the V6-to-V7 forward pass on guitar, bass-pluck, and snare test sounds.
- License
- apache-2.0
- Created
- Jul 12, 2026
- Updated
- Jul 12, 2026