Andrew Price breaks down a practical workflow for using Blender to direct AI instead of fighting it, combining depth passes, ComfyUI, model generation, and world building to create concept art faster while keeping the important creative decisions in human hands.

This talk was recorded at the recent Blender Conference (BCON Austin), hosted by Autotroph: https://autotroph.com. You can find Andrew Price’s YT channel here @blenderguru

00:00 Why 3D artists still have an edge in the AI era
03:17 The moon colony idea that text prompting couldn’t solve
04:21 Using ComfyUI and depth passes to control image generation
06:33 Turning generated images into usable 3D placeholders
08:24 Building a stronger world with AI-assisted worldbibles
10:43 Batch-generating 3D assets for fast scene building
11:40 Using Polydex to assemble scenes like Lego
13:25 Why curated asset libraries still matter
14:27 Labor vs. judgment: the skill shift artists need to understand
15:50 What physically accurate shaders taught us about automation
17:29 “Automate only the skills you’re willing to lose”

AI image generation can make pretty pictures from almost nothing, but getting specific, believable results is a different problem. That’s where Blender becomes a control tool, not just a modeling app.

In this talk, Andrew shows how 3D artists can use blocking, depth maps, asset generation, and iteration to automate the boring parts without giving up authorship.

The bigger point is not just speed. It’s about focusing on judgment, taste, and worldbuilding, the parts of the process that actually matter most.

#Blender #ConceptArt #ComfyUI #3DArt