Wan 2.2 Image-to-Video on 8GB VRAM: 4-Step Clips With Auto Last-Frame Saving
Equivalent-Ad1959 · reddit · 2026-11-22
The author released a beginner-friendly Wan 2.2 I2V-A14B workflow tuned for 8GB VRAM cards (tested on RTX 4060 + 32GB RAM), using Q4KM GGUF quants and LightX2V 4-step LoRAs.
- Start + optional end frame: animate a single image or lock the final pose/face with an anchor end frame, toggled off with Ctrl+B.
- Automatic last-frame saver: saves the final frame of every generation to the output folder; a load node pulls it back in one click to chain 3–4 clips into a 20+ second sequence.
- Speed: 4 steps (2 high-noise + 2 low-noise) render a 7s clip (480×832, 113 frames at 16fps).
- Built-in notes cover setup and chaining, including refreshing faces between long clips.
- Workflow and model links are on Civitai.
More from Multimodal
- AI short film 'The Wow Lunch': Midjourney images animated by Seedance 2.5 via Claude MCP — azed_ai · 2026-11-22
- Zero Custom Nodes: A ComfyUI Pipeline That Puts a Model Into Any Reference Video's Motion — lmcdesign · 2026-11-22
- Open-source Short-Drama Skill helped a creator ship 20+ AI dramas with 50M+ views — vista8 · 2026-11-22
- A burger menu designed as an arcade game screen, prices denominated in HP — gizakdag · 2026-11-22
- AI-generated short film: an AI actress starts questioning her world — THEBLUEHORIZON_AI · 2026-11-22
- pi-H3 distillation test: 4 steps passable, 2 steps break motion and audio — BittiAI · 2026-11-22