Recreating videos in MiniMax H3: a 4-step frame-extraction workflow
Negative-Whereas3307 · reddit · 2026-09-18
A Reddit user demonstrates recreating a Mario talking-head video in MiniMax H3, arguing that "make something similar" leaves too much unspecified. The 4-step workflow (from Hypit, also doable manually):
- Extract reference frames at points where expression, pose, or on-screen content changes.
- Break down the scene: describe framing, setting, delivery, and overlays separately (presenter, ranking numbers, airplane card are independent parts).
- Decide what to keep: specify what H3 should recreate vs. what can change, keeping performance instructions separate from graphics instructions.
- Compare both versions side by side and fix specific gaps in the next pass.
The demo keeps the talking-head setup and numbered list but misses the airplane card — exactly the kind of gap the breakdown step is meant to catch.
More from Multimodal
- Free course covers the 4 hard problems of production generative media agents — Al_Grigor · 2026-11-04
- PixelBench: testing multimodal models pixel-by-pixel instead of with QA — bdsqlsz · 2026-11-04
- Kitten-TTS 2 offers free voice cloning on laptops, no GPU needed — solyarisoftware · 2026-11-04
- "We'll Take Your RAM": a fully local-hardware cyberpunk meme from 2030 San Francisco — Wonderful-Yak-3348 · 2026-11-04
- Openfilm: open-source AI video agent turns coding agents into video makers — solyarisoftware · 2026-11-04
- Microsoft reopens VibeVoice: 10-second voice cloning, 90-minute local TTS — solyarisoftware · 2026-11-04