Swapping Complex Fight Scene Backgrounds with MiniMax Using Only 2 Photos
illumi_nation25 · reddit · 2026-08-14
The author shares an advanced workflow for high-difficulty video background replacement using the MiniMax video model (MMH3 REF2V). The experiment shows that by using just 2 real-life photos as references, the model can successfully replace the background of a 12-second complex fight scene while preserving the original motion, lighting, and camera trajectory.
Key Workflow & Insights:
- Hardware/Software: Run on RTX 5090, utilizing DaSiWa MiniMax H3 Workflows, with final editing in DaVinci Resolve.
- Chunking Strategy: Since the model struggles to match long choreography perfectly, the author split the 12s video into two 6s segments, generated them separately, and stitched the best parts.
- Prompt Structure: The post shares a detailed structured prompt template. It explicitly defines retention and replacement rules for video elements (camera path, action, audio) and environment elements (target background), achieving precise visual control.
Related event: MiniMax H3 Video Generation: Tests and Creative Works(24 posts)→
More from Multimodal
- 1986 to 2026 in one session: outpaint, modernize and 4x upscale, session file included — juanpablogc · 2026-10-16
- Meituan LongCat's WorldCast benchmark: GPT-Image-2.5 tops at 72.8, open-source lags — meituan-longcat · 2026-10-16
- Three resources to get up to speed with FLUX 3 — ariG23498 · 2026-10-16
- Watch: Claude Opus 5.5 builds a Tesla Roadster from a single prompt — TAbrodi · 2026-10-16
- Midjourney v8.2 turns out striking indigo ink-wash art with sref 1848359068 — azed_ai · 2026-10-16
- "AI television from another dimension": robot grooms a chihuahua on air — BotLordBotflix · 2026-10-16