H3 text-to-video demo: lively scene generated with no reference image

SIR_NVAX_A_LOT · reddit · 2026-08-20

A Reddit user shares a text-to-video demo using H3: generated purely from text with no reference image (bf16/50 steps). The clip shows a 'transformative' scene where 'the world is alive,' with the author impressed by the model's realism.

Original post →

More from Multimodal

Multimodal channel →