Remastering old i2v videos with MiniMax H3: a depth-map + reference workflow
WaitAcademic1669 · reddit · 2026-09-07
A creator shares a full workflow for remastering 1-minute+ WAN i2v videos (drifty, low quality) using MiniMax H3's default ref2va template:
- Generate a depth map of the original video with Depth Anything V2, cut into 124-frame segments (5s each in MiniMax).
- Load separate reference images for scene (original first frame with characters removed by AI) and characters, allowing appearance adjustment or character replacement.
- The prompt is key: explicitly instruct the model to use the depth video for motion/camera guidance and reference images for appearance, with accurate action descriptions. Favorite samplers: seeds 2+ddimuniform, resmultistep+beta57.
Stitching tips: keep the same seed across segments, overlap a few frames for slight shifts, or create 5-22 frame transitions with flf2va.
More from coding & agent
- Claude Code + Monid MCP builds a $12K/year B2B prospecting engine on live data — aryanXmahajan · 2026-10-26
- Karpathy shares CLAUDE.md rulebook to stop AI coding agents from piling up tech debt — mdancho84 · 2026-10-26
- Rethinking dependencies in the agent era: drop libs that save 20 lines of code — gethackteam · 2026-10-26
- RustGIS hits 2,000 GitHub stars in two weeks with agent-first architecture — giswqs · 2026-10-26
- Vercel Ships 12 New Products at Its 2026 Conference as It Races Rivals for the Agent Entry Point — FinanceYF5 · 2026-10-26
- Build an AI agent from scratch in under 50 lines of pure Python: ReAct loop tutorial — adnan_hashmi · 2026-10-26