Step-by-step workflow
1
Download Wan 2.1 I2V model
Separate model from the text-to-video variant.
# Best quality (720p):
huggingface-cli download Wan-AI/Wan2.1-I2V-14B-720P --local-dir ./wan21_i2v_14b
# 8GB VRAM option:
huggingface-cli download Wan-AI/Wan2.1-I2V-1.3B-480P --local-dir ./wan21_i2v_1b
# Place in: ComfyUI/models/diffusion_models/
2
Load I2V workflow in ComfyUI
Download a community Wan 2.1 I2V workflow JSON. Drag into ComfyUI.
3
Write your motion prompt
Describe MOTION not the scene — the image is the scene.
"Gentle waves lap against the shore, seagulls fly across frame, realistic ocean motion"
"Candle flame flickers in slight breeze, warm light dances on wall"
"The person takes a deep breath, chest rises and falls, hair moves slightly"
# Settings: Frames: 81, Steps: 30, CFG: 5.0, Image Strength: 0.75
4
Adjust image_strength for motion control
Lower (0.5-0.6) = more creative motion but drifts from original. Higher (0.8+) = faithful to image.
Pro tips
→
Best for: nature, food, products, architecture, portraits
→
For talking heads: use SadTalker or MuseTalk instead
→
Combine with Kokoro TTS for complete video production pipeline
Why this matters for India
// india context
Animate product catalogue photos, food menu images, real estate listings — professional quality for free