🎬 Infinite long takes with MiniMax H3: 30, 45, 60+ seconds in ONE seamless shot, and you can't find the joins.

This is Motion Context V2. It chains several H3 reference-to-video segments into one continuous take. Each new segment picks up the last frames (picture AND sound) of the one before, so the motion, the voice and the scene carry straight through. It runs single-pass on purpose: upscaled two-pass chains show a pixel shift at every join, single pass doesn't. It's also a clean graph, with very few noodles 🍜

How to use it:
1️⃣ In the Multi-Track Editor, add one task segment per clip (up to 15s each at 1.2MP)
2️⃣ Set every segment to reference-to-video. Segment 1 = shot, every segment after it = context
3️⃣ Add your character photo to segment 1 → hover it → Share → "Use as shared reference"
4️⃣ Paste one prompt per segment, then hit Run. The whole chain renders in one job ✅

Default settings: single pass, 8 steps, 1.2MP. Start each later segment's prompt by continuing the action (same pace, same camera), and put the joins mid-walk, never on a doorway.

💬 Stuck, or want to share what you made? Come hang out in Discord:
https://discord.gg/BpM2c7TWTv

📺 Full breakdowns of my H3 workflows on YouTube:
https://www.youtube.com/@promptmastery

☁️ New to RunningHub? Watch this first. Free cloud GPU, zero setup:
https://www.youtube.com/watch?v=WD6roXe7r60

⭐ Follow me here on RunningHub so you get the new workflows first:
https://www.runninghub.ai/user-center/2072160124218466306/userPost?inviteCode=rh-v1580

Give it a spin and show me your longest take! 🤙