What happens when AI becomes part of the creation process?
When generative AI began producing usable 3D geometry, our goal was to make AI a genuine creative partner inside the active sculpting canvas rather than a detached prompt box.
Lead Product Designer
Interaction & systems owner
July 2023 - Nov 2025
Iterative production releases
Womp Core Team
AI Eng, graphics team & founders
In-Canvas Companion
Multimodal chat to 3D clay
From asking questions to generating editable 3D foundations
Conversational iteration, background queues, and zero-click scene injection
Project Context
Lowering the creation threshold
Womp has a unique spatial workflow that takes time to master. Creators struggled with blank-canvas paralysis; generative foundations provided an immediate starting point.
Editable assets, not static novelties
Artists rejected non-editable meshes. They needed a solid foundation: a character base or prop they could immediately sculpt, recolor, and physically print.
Balancing experimentation and compute cost
3D generation is computationally heavy. We designed a daily replenishing credit system that encouraged exploration while reserving priority capacity for Pro subscribers.
Preserving active scene context
A full modal broke spatial continuity. Spark was docked directly into the canvas so creators could inspect and orbit active geometry while generating variations.
Design Process
Validating with a documentation bot
We launched a lightweight support chatbot trained on Womp docs and shortcuts. It resolved 85% of common questions, proving creators welcomed conversational guidance in the app.
First-Generation Spark Support Chatbot
[UI Placeholder: Support drawer showing instant tutorial video cards and tool shortcuts triggered by natural queries]
Figure 1: Initial support bot validating that creators valued in-app conversational guidance.
Docking as a persistent canvas drawer
Instead of a full-screen takeover that interrupted workflow, we anchored Spark as a 320px right-hand panel, keeping the 3D viewport live and interactive while generating.
Split Viewport: Canvas + Persistent Side Panel
[UI Placeholder: Active 3D canvas viewport alongside docked Spark panel showing simultaneous camera orbit and chat]
Figure 2: Docking Spark alongside the 3D viewport to keep conversations tied to active geometry.
Prioritizing mesh quality over speed
Testing showed fast 3-second models produced unusable geometry. Users gladly traded 15 seconds for a watertight mesh, supported by clear GPU progress indicators.
Purposeful Waiting Experience
[UI Placeholder: Status cards explaining compute stages: prompt analysis, multiview generation, and mesh baking]
Figure 3: Replaced generic spinners with contextual progress states reflecting real GPU phases.
Conversational refinement and version branching
Rather than resetting on every prompt, Spark maintained thread history so users could iterate on details in natural language without losing prior drafts.
Continuous Refinement Timeline
[UI Placeholder: Chat thread showcasing prompt thumbnails, conversational modifiers, and quick adjustment chips]
Figure 4: Non-destructive conversation branching allowing creators to explore variations safely.
Single-click injection into native clay
Clicking Add to Canvas instantly converted meshes into native Signed Distance Field clay, ready for direct manipulation with Womp's sculpting brushes.
Mesh Drop onto Womp 3D Grid
[UI Placeholder: Active canvas showing injected model with transformation gizmos, ready for sculpting]
Figure 5: Immediate transition from conversation to active clay sculpting on the grid.
Predictable credit economy
Free users received 300 daily replenishing credits, building sustainable engagement without sudden paywalls, while Pro members unlocked frontier model tiers.
Transparent Credit Spend & Pro Access
[UI Placeholder: Credit consumption badge next to generate button with clear upgrade sheet for frontier models]
Figure 6: Transparent credit meters ensuring creators never face surprise charges.
Key Trade-offs
Side Drawer over Modal Takeover
A full-screen modal provided more layout space, but broke spatial continuity. A docked drawer kept the active object always in view.
Output Quality over Raw Speed
Fast 3-second generations produced melted geometry. Testing proved users preferred waiting 15 seconds for watertight, usable shapes.
Branching History over State Overwrites
Preserving every generation in-thread let creators branch in multiple directions without losing previous variations.
Daily Replenishing Credits over One-Time Grants
One-time grants burned out in one session. Daily 300-credit refreshes established an organic habit of returning to create.
Outcomes & Impact
Direct Problem Resolution
Beginners found instant guidance in chat, lowering support volume and onboarding drop-offs.
Organic Community Adoption
Skepticism faded once creators saw outputs were native, fully editable clay rather than static imports.
Complete In-Canvas Creation Loop
Shipped an end-to-end experience connecting natural language prompts directly to active 3D canvas sculpting.
Current Status
Shipped to production worldwide. The experience established conversational AI as an active companion for 3D modeling, giving artists an intelligent foundation without compromising their creative control.
What I Learned
Introducing AI into a creative tool is not about bolting on a general-purpose chat. It is about identifying where AI genuinely removes creative friction, and giving artists full control over the final object.
Evolving through smaller, validated steps shaped the product into an active creative companion rather than a standalone gimmick.
"Make AI a partner in the process, not a replacement for creative agency."