During my CT&D Summer Fellowship, I explored how Bunny Quest could expand into AI-assisted workflows for different types of social content.
I developed two initial directions: one for structured campaign content and another for flexible student-life content.
TIMELINE
Summer 2026
ROLE
AI Workflow Designer
Art Director
Mentor
TEAM
Haein Noh
Ying Chen
Vy Nguyen
TOOLS
Figma Weave
Flora
OVERVIEW
Designing a system, not one deliverable.
Bunny Quest began as a character system representing FIT’s minor programs. During the fellowship, I explored how AI workflows could help students create new campaign content while preserving the characters’ identities and visual language.
THE GOAL IS TO CREATE A WORKFLOW THAT WAS…
Reusable
Flexible enough for multiple campaign ideas.
Consistent
Able to preserve the Bunny Quest character system.
Teachable
Clear enough for another student to understand and modify.
THE CHALLENGE
The character system was flexible, but the creation process was not.
Bunny Quest could support many content directions, but each new post still required rebuilding the process from scratch.
I wanted to understand how reusable workflows could help designers create new content while preserving each character’s visual identity and personality.
RESEARCH
I learned Figma Weave through exploration, then applied it with intention.
Because Figma Weave was new to me, I began with general YouTube tutorials and small practice workflows, bringing what I know from other node based AI tools such as Flora.
As I became more familiar with its tools, my research became more specific. I started identifying needs within Bunny Quest, searching for relevant tools, testing them, and applying them to my workflow.
INITIAL WORKFLOW
I built two workflows for different content needs.
I created two reusable Weave workflows to explore how the Bunny Quest character system could support different types of social content.
CLICK TOGGLE TO SWITCH WORKFLOWS
PROTOTYPING AND TESTING
Testing whether the workflows were truly reusable.
I asked my mentee, Ying, to review and use both workflows with minimal guidance. I observed how she used the workflow, how she felt, and got tons of valuable feedback.
KEY USER INSIGHTS
Campaign Content Workflow
Fixed vs. flexible was unclear
The computer-based composition appeared customizable even though it was embedded in the master prompt.
Iterations lacked labels
Multiple image and video nodes did not explain how they differed.
A key transition was missing
The dark-screen computer image needed for animation was created outside the visible workflow.
Navigation needed clearer structure
Some assets and paths were difficult to understand, even though the video directions worked well.
YING'S FINAL OUTPUT
Student Life Content Workflow
Some defaults were hidden
The master prompt contained preset values that were not clearly exposed as editable inputs.
The starting frame was unclear
The workflow reused an image from another concept, making the intended video setup confusing.
Character motion needed more control
The idle pose, eye movement, and action direction needed clearer, more specific instructions.
The video path felt disconnected
Start frame, end frame, character reference, and final video nodes were spread apart and difficult to follow.
YING'S FINAL OUTPUT
The workflow produced successful outputs, but testing revealed opportunities to improve clarity, flexibility, and navigation for first-time users.
HOW I CHANGED THE WORKFLOWS
Campaign Content Workflow
I revised the campaign workflow based on Ying’s feedback, turning hidden prompt decisions into clearer, editable controls.
1
Added visual direction controls
I added dedicated nodes for environment, accent color, supporting colors and text placement so users could shape the scene without rewriting the master prompt.
2
Broadened the central object
I broadened the fixed computer setup into a device input so the workflow could support different campaign concepts and objects.
3
Made the aspect ratio editable
The original master prompt hardcoded a 9:16 format. I removed that fixed instruction and added a dedicated aspect-ratio input so users could choose the format.
4
Generated the start and end frames
I added a dedicated frame-generation step so the animation had clearer visual anchors before moving into video.
1
Clarified the image-to-video path
I added a dedicated frame-generation stage so users could understand which images guided the animation.
2
Separated device actions
I divided the device animation into an intro action and ongoing motion for more precise control.
3
Added personality-based movement
I expanded the character-action node so movement reflected the character’s personality, not only the required action.
4
Improved workflow navigation
I expanded the character-action node so movement reflected the character’s personality, not only the required action.
Student Life Content Workflow
I revised the campaign workflow based on Ying’s feedback, turning hidden prompt decisions into clearer, editable controls.
1
Made the format editable
I separated aspect ratio from the prompt so users could adapt the same workflow to different social formats.
2
Made seasonal direction flexible
I turned the season from a hardcoded prompt detail into an editable input so the concept could be reused year-round.
1
Clarified the image-to-video path
I added a separate stage between image creation and video generation to create the visual states needed for motion.
2
Strengthened preservation rules
I made character and scene constraints more explicit so the model would preserve anatomy.
3
Made motion direction more specific
I broke the animation into main action, character motion, camera motion, secondary motion, tone, and audio.
4
Added character-specific sound direction
I introduced a dedicated Bunny Sound input so each character’s audio could reflect its personality.
FINAL OUTPUT
One character system, multiple content directions
The revised workflows supported both polished campaign storytelling and more flexible student-life content. By separating reusable controls from fixed character rules, I could generate distinct visual directions while preserving Bunny Quest’s identity.
CHARACTER PERSONALITY SYSTEM
I started directing how each character behaves, not just how they look.
Here's some examples:
Brand Design Bunny 👑
MBTI: INFJ
Birthday: December 30
Personality:
Proud, calm, intelligent, smug, haughty, independent, and regal. She loves classical music and making things look pretty and organized.
Audio direction:
Soft, restrained hums and delicate squeaks; occasional smug “hmph” reactions. Elegant and composed, never overly excited.
Motion direction:
Deliberate, restrained, and confident. Subtle tilts, composed pauses, and selective reactions; never overly bouncy or frantic.
Game Bunny 👾
MBTI: ENFP
Birthday: June 8
Personaltiy:
Energetic, expressive, curious, playful, imaginative, easily excited. He loves games, gadgets, animation, making things move, discovering how things work
Audio direction:
Bright, playful electronic chirps and squeaks; quick excited “pip!” sounds, tiny game-like beeps, and energetic reaction sounds
Motion direction:
Quick, bouncy, reactive movement; frequent little shifts and excited turns; expressive timing; curious pauses followed by sudden bursts of energy
WHAT'S NEXT
From Figma Weave to Instagram!
The revised workflows will be used by my professor, C.J. Yeh to create and publish Bunny Quest content on social media. This moves the project beyond a design exercise and into an ongoing content system that can be tested through real use.
REFLECTION
Here's what I learned…
Working with AI means designing the system.
The real progress came from building reusable nodes and structure, not from chasing a single perfect image.
Prompting requires thinking like the model.
I learned to anticipate how AI interprets language and to avoid terms that triggered unwanted anatomy or actions.
Creative control still belongs to the designer.
AI could generate endless variations, but preserving the specific Bunny Quest characters required human judgment and direction.
BLOOPERS
Proof that the workflow did not always work(flow).
A small archive of cursed bunnies and questionable characters, broken prompts, and several thousand credits spent learning what not to do.
Want to see more?
Want to hear more about the bunnies, the node spaghetti, and the many times AI confidently misunderstood me? I’m happy to talk through the process. Reach me at akatsukokona[at]gmail[dot]com.