At 11:41 PM (KST) on August 17, 2026, AI artist Timeless (@Timeless_aiart) posted "Fall of the Pantheon," a 15-second short video. It's a brief sequence of a temple collapsing, and the artist said it was made using Midjourney together with ByteDance's latest video generation model, Seedance 2.5.
In less than a day it drew more than 40,000 views, with reactions ranging from "the best thing I've seen today" to a comment saying "I thought it was a 2D studio production."
Design With Images, Finish With Video
In a follow-up reply, the artist posted reference images with the words "Visuals used." They're black-and-white concept art in the composition of a Gothic cathedral with a crumbling spire, and appear to be the reference that became the backbone of the final video. Designing still images in Midjourney first, then finishing them as moving video in Seedance 2.5 is a two-step workflow that has become common among AI video creators lately.
Seedance 2.5 is a video model ByteDance officially released on July 31. It generates up to 30 seconds in one go and produces sound in the same pass. It was applied first to Jimeng AI and Doubao Pro in China, and it's also available on third-party platforms such as Higgsfield.
"The Collapse Physics Doesn't Look Off"
The standout comments were about physics. One user noted, "The scale and impact of the destruction scene are clean. The debris doesn't hang in midair." One common flaw in AI-generated video is objects ignoring gravity and floating halfway, and in this work that awkwardness isn't prominent. There was also a reaction to the anime-style fight scene: "the first example of AI-made work that doesn't feel off."
The Prompt Hasn't Been Shared
Several people asked which prompt produced the scene, but as of August 18 (KST), no public reply from the artist had been confirmed. For now, rather than trying to reproduce the same result, it's more realistic to borrow the two-step flow itself, designing with image generation and video generation as separate stages.