Generative Video News
The latest Generative Video coverage curated for game developers.
arXiv cs.GR details Fluid-Gen-Zero, a training-free way to make pretrained video generators obey fluid physics without retraining. It splits motion from appearance, using a simulator for dynamics and a VLM-guided …
TrajLoc tackles a nasty multi-object video-control problem by replacing dense shared conditioning with per-object attention masks. In practice, that means better identity preservation when trajectories cross or objects occlude each …
A new research tool lets artists poke at video diffusion transformers in real time, not just via prompts. AttentionBender exposes self-attention, cross-attention, feed-forward blocks, layers, tokens, and even hidden neurons …
A new compound AI pipeline claims to push brand-compliant video generation from 42% to 89% by adding retrieval-based brand extraction and a multi-agent QC loop. For teams using generative media …
A training-free video-editing method is using audio and text conditioning tweaks, not weight updates, to push generative models toward specific motion changes. For teams building AI-assisted tools, the interesting bit …
For teams using generative video in design animation, this paper tackles the missing piece: objective evaluation. It proposes a fully automated framework that checks layout fidelity, motion correctness, temporal quality, …
A paper from arXiv shows a 2-stage way to turn video diffusion models into reusable neural materials. The interesting part for game teams is that it uses 17 generated frames …