H3 video generation struggles with cross physics in 'Passion of the Sponge'
BuffaloDesperate8357 · reddit · 2026-08-19
The author shares a workflow for generating 'Passion of the Sponge' (a Mel Gibson parody) with the H3 video model: standard workflow, 20 steps, 0.9MP, combining three 15-second prompt segments.
The hardest part was getting the physics of the cross to look right — even the shown result took great effort. The author has seen better H3 physics examples but can't crack it, and asks the community for prompt-writing tips on improving physics.
More from Multimodal
- PNS paper: trainable neural subdivision augments Loop subdivision with bounded corrections — ssh4net · 2026-08-19
- Creating Vampire Hunter D Style Character with T2V and R2VA — SIR_NVAX_A_LOT · 2026-08-19
- Seedance 2.5 video model test shows significant boost in consistency — SimplyAnnisa · 2026-08-19
- MLLM-Guided Semantic Correction for Text-to-Video Generation — kwangmoo_yi · 2026-08-19
- Help: Adding Video LoRAs to LTX 2.3 for fight scene animations — SayethTwat · 2026-08-19
- MiniMax H3 generates noise at end of 15-second clips — Tuckerdude615 · 2026-08-19