Ryze Adventure
AI Storytelling & Video Generation | 03 .2026 | Weavy
📚 Overview
Curious about the creative limits of generative AI, I built this passion project to benchmark the rapid evolution of video models and test how a traditional compositor adapts to Weavy’s node-based workflow.
This is a personal, non-commercial passion project inspired by "Ry," the brand character for RYZE Mushroom Coffee. While working as a compositor on an official RYZE commercial at Brand New School, I became deeply curious about the creative boundaries of generative AI. I wanted to see how far I could push narrative storytelling independently using these emerging, professional tools.
Using Weavy, this personal exploration serves as a benchmark to evaluate the rapid evolution of video AI.
The project had two primary objectives:
-
Workflow & Interface Evaluation: To test how Weavy’s node-based interface benefits the AI generation process, and to evaluate how easily a traditional VFX compositor can adapt to and pick up its node-graph logic.
-
Technology Benchmarking: To produce both single-shot and multi-shot content, analyzing how much AI has progressed over just a few short months—specifically focusing on improvements in text rendering, physics simulations, character consistency, and overall visual fidelity.
💡Key Learnings
• Upfront planning is critical: Similar to storyboarding—defining approach early (first/last frame, reference-based, text-to-video, etc.) improves cost efficiency
• Reference handling matters: Labeling inputs as “Image 1, 2, 3” works better than repeatedly describing subjects
• Reverse engineering helps: Generate video → extract frame → regenerate for better results
• Model strategy matters: Test prompts on cheaper models first, then move to higher-cost models like Nano
• Editing > regenerating (sometimes): Using tools like Kling Edit on imperfect outputs can be more efficient (check with Justin)
• Reframing trade-offs: Tools like Qwen Multi-angle tend to lower quality—better to prompt directly in higher-quality models
⚠️ Challenges / Limitations
• Body part artifacts : High-quality models (e.g., Veo3) still introduce unwanted human-like features (fingers, toes). Certain trigger actions (e.g., “sipping,” “holding”) can cause this—prompt adjustments are required
• Cross-model inconsistency : Different engines produce varying styles and color profiles, which can create challenges in maintaining consistency during editing
• Camera direction & spatial logic: Still struggles with accurate perspective and object relationships across shots (e.g., backpack position remaining static despite camera angle changes)
⚙️ Workflow / Tools Used
🔍 Infrastructure & Workflow Insights
Evaluating Weavy's node-based architecture highlighted major advantages for technical artists:
-
Non-Destructive Tracking: The node-graph layout makes it incredibly easy to trace progress, reverse-engineer generations, and track structural changes visually.
-
Low Friction for Compositors: For an artist accustomed to working in Foundry Nuke, adapting to Weavy’s node-based framework felt highly intuitive, significantly flattening the learning curve.
-
Areas for Growth: Despite its strengths, the architecture still needs maturity. The current Composite node, for example, is not yet a true compositor. It suffers from clear technical bugs—most notably struggling to accurately merge formats with mismatched resolutions.
🚀 Potential Use Cases
Given the current strengths and limitations of this workflow, the ecosystem is exceptionally well-suited for:
-
Short-Form Social Content: Rapid-turnaround, seasonal campaigns, or real-time marketing tied to major live events (e.g., NFL, NBA, holidays).
-
Children’s Entertainment & Content: Projects utilizing simpler, stylized character designs that align perfectly with current AI consistency models.
Ultimately, while the AI ecosystem isn't ready to completely replace traditional high-fidelity finish work, tools like Weavy prove that the bridge between generative AI and professional compositing logic is narrowing rapidly.
📍 Overall Takeaway
Generative AI has advanced meaningfully in text rendering, physical simulations, character consistency, and visual fidelity over a very short period. This project demonstrates that complex storytelling with a completely consistent character is now fully achievable. AI tools show exceptional potential for rapid, flexible content creation, while traditional VFX pipelines remain essential for projects requiring absolute precision, pixel-perfect timing, and strict spatial logic.