Project One
7 Aug. 25
🍃 Moss Merry Way — AI Concept Art Generator
Status: Prototype Complete · Stack: Google Colab, Stable Diffusion, 🤗 Hugging Face, xFormers, NVIDIA T4
Overview:
This project is the first creative tech sprint from Glowlock Labs, focused on building a text-to-image concept art generator for visualizing environments in the world of Glowlock. For this sprint, we brought to life the realm of Moss Merry Way—a fog-drenched, magical forest filled with sleepy light and emotional stillness.
Goal:
To generate high-quality concept art with the mood of a handcrafted fairytale world using minimal prompts and lightweight customizations.
Use Case:
These images are intended as a base for looping animated backgrounds, storybook illustration, and AI-assisted visual development for fantasy IPs.
🔧 Tech Stack
-
Platform: Google Colab
-
Model: Stable Diffusion XL (via Hugging Face)
-
Libraries: diffusers, accelerate, xformers, transformers
-
Hardware: NVIDIA T4 GPU (Colab Pro)
-
Environment Fixes:
-
Forced PyTorch upgrade to 2.1+
-
Downgraded NumPy to <2.0 to resolve compatibility
-
Hugging Face Token authorization
-
🧪 Sprint Outcome
-
✅ Successful installation and setup of Kohya LoRA Trainer (XL version)
-
✅ Resolved NumPy compatibility errors with patched Colab environment
-
✅ Generated final render capturing the atmosphere of Moss Merry Way
-
✅ Deployed to GitHub and documented workflow for reuse and iteration
-
✅ Published visual + copy across Instagram and Facebook
Project Two
20 Aug. 25
🎥 Dylan Dewlock — AI Video Sprint
Status: Prototype Complete · Stack: OpenArt, Photoshop, Generative Fill, Stable Diffusion, After Effects
Overview
This sprint from Glowlock Labs focused on producing a short animated sequence of Dylan Dewlock, swinging through a giant charm bracelet into his magical world. The video combined AI image generation, iterative editing, and frame-by-frame enhancement to achieve a playful, storybook atmosphere.
Goal
To prototype a narrative animation clip that demonstrates Dylan Dewlock’s entrance into the world—balancing AI efficiency with handcrafted polish for a whimsical, cartoon-fantasy aesthetic.
Use Case
The clip serves as:
-
A pilot animation test for the Dylan Dewlock series
-
A workflow experiment in mixing AI video + human edits
-
A visual reference for charm bracelet world transitions
🔧 Tech Stack
Platform: OpenArt Video + Adobe Suite
Models/Tools:
-
OpenArt video generator (for base animation)
-
Stable Diffusion (for background frames + enhancements)
-
Photoshop (Generative Fill + frame corrections)
-
After Effects (sequencing + lightning strike effect)
Process Fixes:
-
Expanded charm bracelet image to show multiple charms for realistic swing motion
-
Converted video into frames for detailed Photoshop editing
-
Used Generative Fill to patch inconsistencies in Dylan’s motion
-
Reassembled frames into smooth looping animation
🧪 Sprint Outcome
✅ Generated base swing + landing animation in OpenArt
✅ Expanded charm bracelet visuals for multi-charm sequence
✅ Added lightning strike effect while keeping Dylan smiling/heroic
✅ Patched inconsistencies in Dylan’s movement with Photoshop frame edits
✅ Final render captured playful + magical energy of Dylan Dewlock’s entrance
✅ Exported for portfolio/demo use and documented workflow for future episodes