MiniMax H3 Test: Single Prompt Generates Split-Screen and Pseudo-Mocap
Tokyo_Jab · reddit · 2026-08-25
A Reddit user showcased an impressive video generated with MiniMax H3. Using a single prompt, the model created a split-screen effect, synchronized dialogue between two characters, and pseudo-motion-capture level movement consistency. It even accurately rendered a stencil with cut-out logo details. The test highlights the model's capability for complex instruction following in video generation.
More from Multimodal
- The math: 5 rerolls at 480p plus a 4K upscale costs $5 vs $9 at 720p — HeyAmit_ · 2026-10-15
- Dreamina workflow: generate at 480p, upscale to 4K, cut video costs 44% — HeyAmit_ · 2026-10-15
- One-prompt GPT-6 Astra writes two Palestrina-style motets in LilyPond — deanwball · 2026-10-15
- DiffSynth-Studio Adds OPD Workflow: Few-Step Diffusion Training in ~2 GPU-Hours — bdsqlsz · 2026-10-15
- AI film and TV hit the market: Prime Video picks up AI spin-off, pilots screen at Mipcom — lmoroney · 2026-10-15
- Ramayana film taps DNEG's massive VFX and AI-assisted production for global release — CurieuxExplorer · 2026-10-15