Vibes.ai

All-in-one web creation tool to create and edit AI generated images and videos.

timeline

Jan 2026 – May 2026

role

Led design strategy, delivered end to end design, refined front-end code for craft

scope

Redesigned an internal tool and launched to GA. Post GA, improved the product for global launch.

tools

Figma, Claude Code

Context

GA launch: an internal creation tool with 3 weeks
to become a public product

Vibes.ai started as an invite-only tool for employees and contracted creators, built quickly to feed AI content into the Vibes mobile app and the Meta AI mobile feed. I had 3 weeks to redesign it for public GA, and 3 months to take it global.

Previous design: Internal tool was retired before I made this portfolio, so I restored
the original code locally to capture these screens.

Vision for Consumption

Let people play with the ingredients of a video

CharacterPink horse
ObjectGreen jacket
SceneSurreal pink clouds
CharacterLady with white hair

Vision for Creation

Let creators remix and re-use those ingredients

Remix a character into a new scene, or a scene around a new character, and
carry them across creations.

Remix scene
Original video
Remix character

Creator Feedback

Partnered with a panel of 30 creators familiar with AI creation tools for continuous feedback throughout the process.

GA Launch Prioritization

Prioritized by creator need, weighed against eng and design cost

With three weeks until GA, I worked closely with eng to align on what we could build. To move faster, I often built the front end while eng built the back end, rather than handing over Figma specs and re-updating them on every change.

GA Launch
Post GA
Use ingredients in prompt
Can add a single ingredient
Can add multiple ingredients
Create your own ingredients
Can only use provided ingredients
Create and edit my own ingredients
Remix and re-use ingredients
Can re-use your own ingredients
Can re-use your own and remix others
Rich AI video editing capabilities
Editing such as animate images, add music/voiceover, edit via prompt
Robust AI editing, change camera angles, easy ingredient swap
Agentic creation
Need to prompt to create
Co-create with an agent (prototyped but not shipped)

IA Change

Project-based creation: Easier to group/organize assets.

Instead of generating assets first then organizing them into a project, everything starts with a project. Within a project, you can create assets, edit, and publish.

Previous

Prompt → Generate media

Updated

Create project → Prompt → Generate media

Design Decision

Making it easier to keep characters consistent

Creators had two ways to hold a character steady, both imprecise. We added a third with real control over characters, scenes, and styles.

Less control More control

Describe the character every time

Write the character into each prompt and rely on the model to infer it. Leads to high variability in the outcome.

Attach a reference per generation

Upload an object reference image alongside each prompt. Reduces variability but still lacks control in details.

Persistent, named ingredients

Characters, scenes, and styles become reusable ingredients that live in the project and library.

Introducing Ingredients

Maintain characters, scenes, and styles across your creations

For MVP, creators couldn't create ingredients from scratch. Instead they selected from a Meta-curated library or turned an asset from their own project into a reusable ingredient. Enough to test whether reusable ingredients changed how people worked.

Post GA Addition

Allow creators to turn their assets into ingredients

Creators wanted to turn their own images into ingredients, but extraction depended on the original prompt. Supporting it needed a different pipeline (visual segmentation and analysis), so I designed within the constraint: a flow to make ingredients from platform-made assets while we built upload support.

Easier Way to Edit

Generate & edit in one place

You can now refine, animate, and add music without leaving the tool. The work that used to mean exporting to Premiere now happens where the video was made.

Timeline Interaction Alts

Exploring different timeline layouts

I explored layouts that kept edit controls and assets on separate surfaces, with edits opening in a dialog. But the dialog covered the timeline layers, and creators need to see those layers while they add new ones. So instead of a dialog, I consolidated controls and assets into a single adaptive panel that changes with what you select, keeping the timeline visible throughout.

Previous Design

Edit tools in a floating bar next to preview, assets in the left panel

Final Design

Controls and assets in a single adaptive panel that changes depending on what you select

Timeline Final Design

Consolidated asset preview and edit controls in one place

Vibes timeline view — adaptive edit panel on the left, preview and timeline tracks on the right

Thinking About the Future

Agentic creation: from prompting to co-creating with agents

Post-GA, I started rethinking how vibes.ai should evolve. Today you prompt for an asset. The next step is an agentic collaborator: you drop in what inspires you, and it helps you find the story.

Concept for agentic creation — inspiration panel on the left, ingredient suggestions for characters, scenes, and styles on the right

Result

3 weeks
internal → public
Global expansion
H1 2026
+50%
like rate vs UGC

Vibes.ai launched globally in H1 2026. Content created with Vibes.ai outperformed off-platform content in the Vibes mobile app, driving 50% higher like rates than UGC. In June, a reorg shifted company priorities and the project was paused.

← Home Next project →