AI Copilot Zumi
Discover your creative partner woven into the canvas: how Zumi proactively evaluates generation quality, diagnoses canvas structure, and deepens co-creation.
Most AI assistants passively sit in a chat corner, waiting for creators to type prompts. In complex visual storytelling, creators need something much more active: a collaborative partner with cinematic sensibilities and technical alertness.
Zumi is AZ8's resident AI Agent. Seamlessly integrated with the infinite canvas, Zumi monitors image consistency and motion smoothness in the background, offers recommendations at critical milestones, and even helps reorganize cluttered board layouts.
1. Three Proactive Diagnostic Triggers
Zumi follows a "non-intrusive, context-aware" principle, delivering proactive insights only when real creative blocks or inspection moments arise:
Trigger 1: Video Node Quality Assessment & Scoring
Whenever a video finishes rendering, Zumi evaluates the output across visual coherence, limb motion physics, and fidelity, delivering strengths, shortcomings, and targeted tuning tips.
Trigger 2: Canvas Scale & Storyboard Structure Suggestions
When the canvas hits critical complexity thresholds (e.g. 5 shots drafted, or 20 nodes forming a sprawling board), Zumi surfaces spatial layout tips to clarify narrative branches and prevent clutter.
Canvas Structural Advice
Global Workflow Progression Recommendations
Trigger 3: Writer's Block & Real-Time Prompt Expansion
If your cursor stays idle in a prompt input for more than 10 seconds, or the system senses repetitive edits, Zumi analyzes the upstream shot context and recommends cinematic keywords and lens parameters.
2. On-Demand Activation & Multimodal Co-Creation
Beyond proactive hints, you can summon Zumi at any moment for deep creative dialogue:
Entry 1: Dedicated Sidebar Chat Drawer
Click the Zumi avatar in the left toolbar to open the full conversation drawer for worldbuilding, brainstorming, and shot list breakdowns.
Sidebar Zumi Entry
Multi-Turn Discussion Window
Entry 2: In-Node & In-Input Action Triggers
Stay focused on your shot without switching contexts. Hover over the top bar of any node or inside the prompt input to access Zumi micro-actions: Refine Prompt, Diagnose Flaws, or Derive Next Shot.
Node Toolbar Action
One-Click Prompt Optimization
Entry 3: Box-Select Multimodal Assets for Complex Queries
Select a block of script text, two concept frames, and a generated video clip simultaneously, then send the selection to Zumi with a question like: "How can I heighten the visual tension in this shot while maintaining character fidelity?" Zumi will synthesize all inputs and offer actionable next steps.
