Step-by-step workflow
1
Create character reference image with Flux
Generate a high-quality portrait as your anchor image.
"Portrait of a 28-year-old Indian woman, shoulder-length black hair, dark brown eyes, warm olive skin, confident smile, light blue kurta, professional, studio lighting, sharp focus, photorealistic"
# Save the seed number!
2
Install IP-Adapter FaceID in ComfyUI
Locks face identity across all scenes.
# In ComfyUI Manager, install:
# 1. ComfyUI_IPAdapter_plus
# 2. ComfyUI-WanVideoWrapper
# Download: ip-adapter-faceid-plusv2_sdxl.bin → ComfyUI/models/ipadapter/
pip install insightface onnxruntime
3
Generate scenes with consistent face
Use this template for each location or action.
"Priya at a modern office desk, working on laptop, warm indoor lighting"
"Priya walking through a mall, casual style, natural daylight"
"Priya speaking to camera, confident expression, YouTube talking head shot"
# IP-Adapter weight: 0.8
4
Animate each scene with Wan 2.1 I2V
Take each generated image and add motion.
"Priya looks up from laptop and smiles, slow zoom in, warm lighting, realistic motion"
5
Edit scenes together
Import clips into CapCut or DaVinci. Add Kokoro TTS voiceover. Add music. Export.
Pro tips
→
Face weight 0.8 = sweet spot
→
LTX Studio (ltxstudio.com) has built-in consistent character feature — no GPU needed
→
Keep video prompts simple — AI will handle the motion
Why this matters for India
// india context
Create YouTube series and brand storytelling with a consistent AI character — without hiring actors or paying AI avatar tools