Threadscape: AI-Driven Visual Refinement
Exploring the intersection of AI-generated imagery, texture, and storytelling, Threadscape is a visual experiment in prompt engineering. The project takes three related AI-generated images—a wool-knit living room, a soft pink cat, and a knitted airplane—and refines them through iterative improvements, upscaling, and animation.
Step 1: Setting the Creative Vision
Defining the Mood, Materials, and Color Palette
🔹 Color Palette: Soft pink monochrome
🔹 Lighting: Warm, golden sunlight
🔹 Materials: Plush, cozy textures (knit, wool, velvet)
🔹 Mood: Dreamy, intimate, and inviting
Initial Inspiration
💡 The challenge was to create a consistent aesthetic across multiple AI-generated images while keeping each unique subject visually unified.
Step 2: Writing the Base Prompt
To maintain cohesion, I used a single structured prompt, changing only the subject:
📌 Base Prompt:
"A [subject] made of soft pink knit fabric, bathed in warm natural sunlight, surrounded by plush textures like wool, velvet, and cotton, with a dreamy and inviting atmosphere. Tight close-up shot."
🔄 Example Variations:
Cat: “A fluffy pink cat resting on a soft knit blanket…”
Living Room: “A cozy pink knit living room with oversized woolen furniture…”
Airplane: “A plush pink airplane made of knitted fabric, resting on a soft cushion…”
Step 3: Refining for Consistency
Challenges & Adjustments
❌ Framing Differences – Some images were close-up, while others were too wide.
❌ Texture Inconsistencies – Some results lost the knit/wool details.
❌ Lighting Variations – The warmth and softness weren’t always consistent.
🔧 Iterations & Fixes:
✅ Specified “close-up shot” for uniform framing.
✅ Strengthened “knitted wool fabric” to keep textures accurate.
✅ Reinforced “warm golden sunlight” for consistent mood.
Step 4: Final AI-Generated Images
After multiple refinements, I achieved a cohesive set of final AI-generated images that maintained:
✅ Cohesive color palette (soft pinks throughout)
✅ Matching material textures (knit, wool, and velvet across all images)
✅ Consistent warm sunlight (same golden glow in every image)
Step 5: AI Upscaling & Detail Enhancement
To enhance texture fidelity, I used Magnific AI to upscale the images, refining the fine-knit details and preserving realism while maintaining the original soft aesthetic.
Key Improvements:
✅ More defined wool & knit textures without looking artificial
✅ Sharpened lighting details for added depth
✅ Increased resolution for animation-ready assets
Step 6: Animating AI-Generated Content
💭 What’s Next?
I’m now taking these images into Runway AI to animate subtle movements, such as:
🌬 Soft wind gently moving the knitted curtains and blankets
✨ Subtle lighting shifts to enhance warmth
Final Thoughts & Takeaways
Key Learnings from This Process:
🔹 AI requires multiple iterations to achieve a polished look.
🔹 Strong prompt engineering is about guiding AI, not just accepting outputs.
🔹 Uprezzing and fine-tuning textures ensures images are ready for animation.