Text to Video News and AI Updates
Follow Text to Video developments across AI companies, labs, and open-source projects.
Latest Text to Video news, research, benchmarks, product releases, and industry adoption updates.
Latest Articles
- ReVi: Recursive Recap Video Generation — ReVi is a multi-agent system that turns a full novel into a short recap video within a fixed time budget, planning story, narration and shots before generating video, and beating baselines on recap quality and duration compliance.
- What Remains in Sight? Autoregressive Video Decoding as Representation-Guided Context Rewriting — ReCR is a training-free method that treats which past frames an autoregressive video model sees as an explicit decision, improving long-horizon generation across several backbones at matched token budgets.
- OmniOPD: Stable On-Policy Distillation for Joint Audio-Video Generation — OmniOPD stabilizes on-policy distillation for joint audio-video generation by giving the audio teacher cleaner visual context and more audio supervision, improving video quality while keeping audio quality.
- GenDSR: Transferring Representations from Video Generation Models for Dynamic Spatial Reasoning — GenDSR feeds intermediate features from video generation models into vision-language models for dynamic spatial reasoning, reaching a state-of-the-art 67.2% on DSR-Bench.
- LPA-CWM: A Learned Physical Adjudicator for Motion Reasoning with Counterfactual World Models — LPA-CWM adds a lightweight 3M-parameter adjudicator to counterfactual world models for video motion tracking, predicting response reliability rather than using uniform aggregation—improving tracking accuracy by 60% on DAVIS and 29% on Kinetics without modifying the underlying video model.
- Dynamic Learning Solutions: A System for Personalized Educational Video Generation — Researchers built a pipeline that converts PDF textbooks into animated video explanations by chaining RAG retrieval, Stable Diffusion image generation, DynamiCrafter animation, and Google TTS — demonstrated on NCERT educational materials.
- BEACON: Behavior and Appearance Control for Subject-Specific Video Generation — BEACON generates person-specific videos that preserve both visual identity and characteristic facial dynamics by disentangling the two into separate conditioning signals — updating only ~1% of the Wan video diffusion model trained on 2,000 pairs, with improved expressivity over state-of-the-art methods.
- New York Seizes a Dozen Celebrity Deepfake Websites — The Manhattan DA's Office seized 12 nonconsensual deepfake pornography websites targeting ~1,200 victims — the largest such enforcement action to date — using the US Take It Down Act, which now gives law enforcement authority to seize domains hosting synthetic sexual imagery.
- MindTopo: Can Foundation Models Reason in Topological Space? — MindTopo — 11,030 instances across continuity, separation, order, enclosure, and knot tasks — finds all 14 MLLMs tested fall far below human performance on topological planning, with fine-tuning improving reasoning more than planning and video generation failing to preserve topology across transitions.
- From Evaluation to Enhancement: Benchmarking and Improving Think-with-Video Reasoning for Video Generative Models — VWG-Bench exposes a stark gap in video generation: models score well on visual quality but consistently fail logic-heavy and rule-constrained tasks across 9 reasoning dimensions. Vid-PRE, a RL-trained prompt rewriter, closes much of this gap without touching model weights.