Pull up the LLM-Stats video arena today and the number one slot belongs to an open-weights model. Not Runway.
OpenAI's Sora racked up a million downloads in its first week last November.
The first thing Runway Gen-4 got right wasn't the native audio or the improved physics.
Sora's shutdown got the column inches.
ByteDance pushed Seedance 2.0 into CapCut in late March.
OpenAI was spending fifteen million dollars a day running Sora. The product generated $2.
xAI reported 1.245 billion Grok Imagine videos generated in a single 30-day window.
Google shipped Veo 3.1 Lite this week at five cents per second of generated video, and the timing tells you everything.
Six months ago, every AI video generator shipped silent clips.
On April 26, OpenAI pulls the plug on its video generator. Six days from now, the most hyped AI video product ever built stops generating videos.
The marketing page buries the only number that matters: 13.6 GB.
Every AI video tool until now has shipped the same product: type a prompt, get a clip. Five seconds of a woman walking through a field.
OpenAI spent more money per day running Sora than most startups raise in a seed round.
Five cents. That's what Google charges for one second of 720p video from Veo 3.
A video model that ingests nine reference images, three video clips, and three audio tracks in a single generation pass just hit number one on Artificial...
Every video diffusion model released in the last year has followed the same playbook: train bigger, throw more VRAM at inference, charge accordingly.