Generating 10 Chained AI Video Clips on RTX 3090 Takes ~4 Hours

DeliciousGorilla · reddit · 2026-08-08

A developer shared their practical experience generating continuous videos using Stable Diffusion. To solve the continuity issues of a long take, they adopted community advice to use cuts instead, splitting the same scene into 10 separate video generations.

Hardware and parameters included a single RTX 3090, a resolution of 1344x768, 20 steps, and native audio generation enabled. The entire process took 3 hours and 46 minutes to complete.

Original post →

More from Multimodal

Multimodal channel →