UGG unifies video generation and 4D geometric prediction
One model treats RGB, depth, and camera pose as conditional completion and can search scene hypotheses at inference.
Stories per day · 30 days
9 RSS
Relatedworld-models11diffusion7minimax7huggingface5lora5streaming5autoregressive4benchmark4distillation4gemini4multimodal4benchmarks3camera-control3google3multi-agent3open-weights3text-to-video3world-model3
161 stories, newest first · one card per story
Stories tagged #video-generation
APEX few-step streaming video drops the extra fake-score network
Condition shifting estimates the generated velocity inside the generator and cuts peak GPU memory 4.4% and iteration time 9.2% versus DMD.
PhysOPSD uses executable code as privilege for video generation
Authors say the method beats supervised fine-tuning and reinforcement learning with 3,000 samples while keeping general video quality.
SemanticKV compresses KV cache for agentic video generation
Authors report a 3.35x KV-memory cut and 1.22x speedup on Causal Forcing at compression ratio 0.7.
WorldFoundry catalogs 270 models for world-model evaluation
The open-source stack also lists 74 benchmarks and 62 metrics, separating execution evidence from scores.
Google rolls out Gemini Omni 1.2 Flash with 15-second clips
The video model is generally available on the Gemini API at $0.07 per second of 720p, 30% below Omni 1.1.
1
Video-Coder trains code-driven agents for video generation
Video-Coder achieves 7-point gains on PhyGenBench and 24-point gains on VBench-2.0 via agentic reinforcement learning
Video Rebirth publishes HyperFlow H3 Flash, a compressed MiniMax-H3
The 13.7B–13.9B model makes 5.2–15 second video with stereo audio in 4–8 steps, under a restricted MiniMax license.
EmoReward self-evolving reward calibration advances emotional video RL
EmoReward distills reliability from training traces to adaptively allocate weights for facial motion and chromatic cues in video generation.
Kling multimodal understanding insights for video generation
Kling-MLLM unifies image video audio inputs; captioning and prompt enhancement act as semantic contracts that maximize description density in the generation chain.
History Forcing improves long-horizon video world models
History Forcing is a distillation framework that strengthens historical supervision for few-step autoregressive video generators, achieving state-of-the-art results.
arXiv paper presents Wave Forcing for speed-of-light video generation
Wave Forcing co-designs generation training and execution for continuous video output achieving 1F1C with 147.6 DiT FPS on eight H200 GPUs.
Pika releases Pika H3 Flex open weights for video and audio
Local inference outputs native 480p or 768p with stereo audio; the hosted API adds up to 4K.
3
Agentic Video Generation and Recursive Self-Improvement Studied
Agent harness coordinates video models for agentic generation, world modeling, and recursive self-improvement in multi-shot video production.
Wizard enables real-time agentic control of game worlds
Wizard uses skeleton and geometry proxies with adaptive masking to allow zero-shot control of characters in novel scenes via an agentic framework.
WeVisCut trains a 9B agent for narration-to-video editing
Trained on more than 130,000 examples, the 9B policy matches Claude Opus 5 while using nearly 33 times fewer tokens.
VibeVideoBench scores agents that render video by writing code
Across 2,151 rollouts, the strongest of nine systems meets every core requirement on 76.3% of 239 tasks.
Hy-Nomas generates multi-turn interactive video worlds from text and camera
The 8B diffusion model uses action-aware causal attention and a prompt enhancer; authors release weights, code, and benchmarks.
Open-H3-Regenerate-2K upscales MiniMax-H3 drafts to 2560×1440
Authors report a five-second clip falling from 87.26 seconds to 22.45 seconds on one B200 with the fast decoder.
3
Visual memory raises recall but lowers precision in long video
On VMem-Bench, retrieval lifted entity recall from 0.559 to 0.628-0.693 while precision and forbidden-entity avoidance fell.
OmniShowcase generates subject-consistent video from multi-view images
Training covers 12 camera motions and 88 trajectories, including full 360-degree orbits, with surface-normal conditioning from reference views.
Licon VBVR LoRA adapts MiniMax-H3 for image-to-video
The adapter, H3-VBVR-I2V.safetensors, is trained on rebuilt VBVR data and loads in a MiniMax-H3 ComfyUI workflow.
WorldCast generates character video under large-angle camera motion
CLACT adds multi-view character data, and GAMA injects geometrically aligned features to limit identity drift.
World-in-Context unifies view, action, and camera video controls
A 5B model matches 22B EchoWM navigation at 81.5 versus 81.6 and switches views 93.8% of the time.