Hacktoberfest 2026 Week 2 β "Side Hustle: build something with open-source AI that someone would actually pay for"
What I Built
Wondful AI Renderer is a free, open-source (MIT) Blender plugin that turns any 3D white model into a finished, art-directed scene image β without leaving Blender.
The core promise: your geometry stays locked. On our benchmark (SheenChair Nordic living room), the rendered result holds 99.2% IoU against the original model's silhouette. The AI changes materials, lighting, and environment β never your product's shape.
v1.2.0 (released yesterday) adds: Layered output β transparent PNG, shadow/projection layer, clean background, and layered PSD, so the result drops straight into a Photoshop workflow
Pixel-exact detail paste-back β license plates, logos, and nameplates are composited back unmodified ("the AI doesn't touch your product")
Custom terrain β adjustable four-corner slopes with presets
Material-based color calibration (default 0.7) Who Would Pay
3D artists and small studios doing product visualization who need finished marketing shots from CAD/white models β without a photo shoot, without manual compositing.
The free MIT version is fully functional. The paid version ($29.90 on Gumroad, includes 1 year of updates) gets every new feature 3 months early. People pay for early access to the pipeline, not for locked features.
How I Built It
Blender Python addon (operators + N-panel UI), MIT licensed Appearance pipeline: the plugin sanitizes the prompt, renders structure guides (mask/depth/normal) from the viewport, and drives an image model for the final re-render
v1.2.0's layered output was built to answer a real artist complaint: "AI images can't go into my PS workflow" β now they can
208 offline tests passing; install is download-zip β Preferences β Add-ons β Install from Disk
Demo
White model β finished shot (all inside Blender): SheenChair white model β Nordic living room (IoU 99.2%)
YU7 car white model β desert / grassland / snow scenes
Shoe, watch, camera, lamp β 4 more product cases in docs/CASES.md
Before/after comparisons and measured IoU numbers are in the repo's case documentation.
Code
GitHub (MIT): [https://github.com/tiandaoyouchang-png/wondful-ai-renderer](https://github.com/tiandaoyouchang-png/wondful-ai-renderer)
Paid version (early access): [https://tiandao1.gumroad.com/l/wondful](https://tiandao1.gumroad.com/l/wondful)
Built solo. If you've ever needed a product shot but only had a white model, this is for you.