MAGE unifies text, visual, and audio-conditioned music generation and editing in one flow-based latent model with dynamic modality masking and cross-gated control.
In: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SD 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
MAGE: Modality-Agnostic Music Generation and Editing
MAGE unifies text, visual, and audio-conditioned music generation and editing in one flow-based latent model with dynamic modality masking and cross-gated control.