Geometry Duo for ComfyUI brings flicker-free depth and normal guidance to video diffusion
Narrow-Particular202 · reddit · 2026-10-06
A developer released two ComfyUI nodes to fix geometry instability (morphing characters, wobbling backgrounds) in video models like MiniMax H3, Wan 2.1 and CogVideoX:
- ComfyUI-Depth-Anything-3: video-batching depth node with pure PyTorch GPU colormaps (zero CPU stalls), minmax temporal normalization for flicker-free results, and metric checkpoints outputting real-world distances in meters.
- ComfyUI-Lotus-2: dual-engine generative normals/disparity (SD2.1 fast engine + FLUX.1-dev detail engine) with 2-wire plug-and-play wiring; a built-in 2MB precomputed empty-prompt vector removes the 10GB T5 CLIP dependency, and a Stage 2 Detail Sharpener (steps 2–4) recovers hair/fur detail.
The post includes a workflow selection table: DA3 for temporal consistency and 3D camera tracking, Lotus-2 for micro-detail normals. Both are on GitHub and available via ComfyUI Manager.
More from coding & agent
- Gemini CLI PR adds custom OTLP headers to telemetry, enabling auth'd exports to Grafana and Datadog — jesussamuel-byte · 2026-10-06
- A 4-step Opus 5.5 + Hermes workflow that turns competitor pages and reviews into testable positioning angles — VibeMarketer_ · 2026-10-06
- Jeff Weinstein courts SaaS platforms to embed agent-driven purchasing flows with approval gates — jeff_weinstein · 2026-10-06
- Is Product the new DevRel? DevRel teams are merging into product and PM roles — gethackteam · 2026-10-06
- How 4 humans coordinate Claude Code, Codex, Muse and Hermes via Microsoft Planner kanban — DJAI9LAB · 2026-10-06
- HyperFrames Studio launches: a video editor built for agents, co-editing on one timeline — toolstelegraph · 2026-10-06