fal to Retrain MiniMax H3: Graphic Live Video Production, 5-second Video Output in 3 Seconds

Dynamic Beat AI News Flash: AI reasoning platform fal has released H3 Max. It is a video model trained by fal based on the open-source MiniMax H3, with a focus on enhancing keyword understanding and visual effects, and has also made specific optimizations for its inference system.

In Artificial Analysis's Audio-Visual Leaderboard, the H3 Max text-to-video ranks first with 1205 Elo, surpassing ByteDance's Seedance 2.0 and the original H3; the video-to-video ranks third with 1235 Elo, just 4 points behind the leader Wan 3.0.

The speed is also remarkable. fal states that generating a 5-second 768p video takes less than 3 seconds, with a throughput approximately 35 times that of the original H3 interface. Currently, it is launched with a 50% discount, priced at only $0.04 per second for 768p, and generating a 1-minute video costs only $2.4; this offer will end on September 1, and the price will revert to $0.08 per second.

However, the capabilities of H3 Max are more limited. It currently only supports up to 768p resolution, while the original H3 can generate 2K videos; the launch only includes video-to-video and text-to-video, with Reference-to-Video not yet available.