Strike a Chord! Modal Kinetic Typography
The paper’s core move is to animate letters through their own vibration modes, keeping glyphs legible while making motion carry meaning.
The method builds smooth looping motion from finite-element modes computed from each vector outline, plus closed-form rigid motion for letter parts. A frozen video diffusion model supervises only modal amplitudes and phases, limiting the jitter and tearing the authors associate with free-form point optimization. Because the motion basis comes from the glyph itself, the system avoids category-specific skeletons or keypoints, aside from a one-time language-model list of movable letter parts. The authors report smoother, more articulated results than Dynamic Typography and AniClipart, stronger human preference, and less tearing, including when the base shape is frozen. HF Daily Papers' note
The method builds smooth looping motion from finite-element modes computed from each vector outline, plus closed-form rigid motion for letter parts. A frozen video diffusion model supervises only modal amplitudes and phases, limiting the jitter and tearing the authors associate with free-form point optimization. Because the motion basis comes from the glyph itself, the system avoids category-specific skeletons or keypoints, aside from a one-time language-model list of movable letter parts. The authors report smoother, more articulated results than Dynamic Typography and AniClipart, stronger human preference, and less tearing, including when the base shape is frozen. HF Daily Papers' note
score 4