Tag
video generation
Video generation is shifting from making clips that merely move to models that can be controlled and reasoned about. Current work focuses on temporal consistency, playback speed, camera-object motion separation, and active vs. passive actions to better match user intent and real-world causality.
12 articles
Grok Imagine 1.5 turns prompts into 720p video
I break down Grok Imagine Video 1.5 and give you a copy-ready prompt workflow for fast 720p video generation.
OpenAI’s Sora hardware targets enterprise video
OpenAI’s Sora enterprise hardware brings local AI video generation to studios, agencies, and firms that need speed and privacy.
Crun AI turns Gemini Omni into chat video editing
Crun AI now exposes Gemini Omni so you can create and edit videos by chatting, with docs, credits, and usage tracking built in.
4 reasons Gemini users hit caps so fast
4 reasons Gemini users are hitting usage caps fast, plus what Google says about the new compute-based limit system.
Lumos-Nexus bridges reasoning and video quality
Lumos-Nexus separates training from inference to improve video quality without sacrificing reasoning-driven generation.
RefDecoder adds reference conditioning to video decoders
RefDecoder feeds reference image detail into video decoders, improving consistency and reconstruction without extra fine-tuning.
EntityBench Tackles Long-Range Video Consistency
EntityBench measures whether video models keep characters, objects, and locations consistent across long, multi-shot sequences.
ActCam adds joint camera and motion control
ActCam is a zero-shot way to steer both actor motion and camera path in video generation without training a new model.
How to Migrate from Sora 2 in 2026
Migrate Sora 2 video workflows to new models before OpenAI’s shutdown deadlines.
Teaching Video Models to Understand Time
A self-supervised way to detect speed changes, estimate playback speed, and generate or sharpen videos at controlled timing.
MoRight tackles motion control and causality
MoRight separates camera and object motion, and models active vs. passive actions so video generation can react more plausibly to user input.
Inside Sora's Shutdown and AI Vendor Risk
Sora peaked near 1 million users, then fell below 500,000 while burning about $1 million a day. The unit economics broke.