AI-powered creativity in
social media marketing
From scattered inspiration to structured creative direction
It's where
data sparks creativity
To drive Viral Moment, an AI startup's next stage of business expansion, we designed Spark — an add-on module that transforms social listening data into actionable creative direction for marketing teams.
I was the lead designer driving the vision and design of Spark’s end-to-end experience.
Duration: 3 month
Team: 7 members
Viral Moment is a leading video-first social listening platform. It utilize computer vision technology to read videos, detect what’s trending, and use semantic AI to break down audience conversations
However, it currently stops at analytics, leaving a gap between insights and actionable output
Fragmented Inspiration, Unclear Drivers
Team members collect scattered inspirations but struggle to decode which creative elements truly drive performance.
Team members save video inspirations in their own way. When it comes time to share and analyze, it’s messy and requires long discussions about what actually works.
Shifting Trends, Opaque Algorithms
Trends change rapidly, and black-box algorithms make it hard for marketers to anticipate and act on what’s next.
We usually spend about a week revising a proposal, yet the final creative output from the team often doesn’t align with our expectations.
Misaligned Expectation, Slow Decisions
Different campaign vision slows approvals, delays buy-in, and leaves teams with less time for creative decision-making.
We have to manually scrape trending data and analyze why it’s popular — so much of it still relies heavily on experience.
In co-creation brainstorming sessions, together, we generated 30+ raw ideas
User testing
Modular Output Structure in timeline format
Users strongly validated the three-pillar structure: storyline, visual, and audio. It reflects how they naturally plan proposals.
Beyond that, they suggested a tighter connection between storyline and visual elements. They envisioned something closer to a storyboard, where narrative beats align directly with visual cues, helping teams translate strategy into execution.
Collaborative space
Our second idea was a whiteboard-like collaborative canvas designed to solve the problem of misaligned goals and expectations.
However, Users expressed concerns about potential overlap with existing collaboration tools, which might add extra burden.
Considering this feedback and the scope of an MVP, we decided to focus on the most critical feature — collecting and aligning team inspirations in one place while leaving advanced collaboration functions for future iterations when we can integrate this product to third-party communciation tools
Mobile inspiration gathering
The third idea focused on a mobile tool to quickly capture inspirations when user scrolling social media in their own time. Users could tag the specific parts they liked in a video to use later as references. Feedback was very positive: marketers loved the convenience of mobile for saving ideas.
Building on feedback from the other two ideas, we unified the tagging system into the three validated pillars. However, some users worried the tagging process might become tedious. To address this, we iterated a quick voice/text note feature, allowing users to capture thoughts instantly without excessive clicks.
The Role of AI in
Shaping Ideas
Multiple AI models work together to deliver outputs that stay true to the marketer’s vision.
01
Establish Campaign Context
AI reads the campaign brief, brand assets, documents and audience details. This information is stored as a knowledge base, forming the foundation for all later recommendations and data sourcing. By grounding the system in each brand’s context, we ensure that any inspiration, trend, or creative suggestion stays aligned with campaign objectives rather than generic outputs.
NLP
RAG
02
Trend Retrieval & Data Gathering
Based on all information provided, AI automatically decomposes them into relevant search keywords and hashtags. Using API calls to major social platforms, it then retrieves trending videos that match the campaign context.
Along with the videos, the system collects engagement metrics, audience sentiment, and platform-specific signals, creating a live data feed of what’s resonating in the market
NLP
API integration
03
User Tagging & AI Enrichment
On mobile, users can tag specific elements within a video they find inspiring. These tags are standardized into the three creative pillars: storyline, visual, and audio. Behind the scenes, AI captures these annotations, links them to the video’s extracted features (visual, audio, storyline), and stores them as structured metadata. By combining user judgment with machine-extracted signals, the system learns which creative elements resonate most and enrich context of the compaign
Reinforced learning
Computer Vision
Speech to text
04
Element Analysis
AI goes deeper into the storyline, script, and visual composition of each video. Using natural language understanding and computer vision, it dissects the narrative arc, maps them against visual and audio features such as color palettes/soundtrack.
These structured elements are then cross-referenced with user tags and notes, allowing the AI to reason about why certain parts of a video resonate.
Computer Vision
LLM Reasoning
05
Creative output and copilot
AI retrieves all the processed information — campaign goals, trend data, video features, and user tags — and synthesizes them into a structured creative output.
Spark also introduces an AI Co-Pilot that allows marketers to converse with the system directly to make edit on the output
LLM
Final Design
Spark reads videos, analyzes trends, and augments team inspiration — helping marketers generate precise, curated and data-backed creative ideas.
AI generated Creative Direction
Provide user a structured, data-backed creative strategy. The interface presents a canvas-style workspace where teams see curated recommendations for storyline, visual motifs, and audio choices tailored to campaign goal
Users can also see the AI-analyzed video sources, with clear explanations of why specific elements and storylines drive virality — turning raw inspiration into actionable reference.
Seamlessly collecting inspiration
On mobile, users can easily save inspirations. They can highlight and tag standout moments, such as a hook, visual, or audio cue, that caught their attention. These annotations are then fed into Spark’s AI engine, enriching the system’s ability to generate more accurate output
All inspirations are stored in shared campaign folders, ensuring the entire team has visibility into the same pool of ideas. This keeps everyone aligned, avoids duplication, and makes collaboration around a campaign faster and more seamless.
Inspiration vault
A centralized library where teams can collect, organize, and revisit inspirations without the chaos of scattered links and screenshots. User can easily filter, review inspirations, allowing team members to add insights or rationale, ensuring that inspirations become actionable knowledge assets.
AI also summarizes each video’s key elements and surfaces insights — explaining why certain hooks, visuals, or sounds resonate — so inspirations become actionable guidance, not just references.
Trend analyzer
Acts as Spark’s social listening engine, helping teams stay ahead of shifting conversations. Users can query trending topics, hashtags, and competitor activities directly related to their campaign goals.
Any trending video can be saved directly into the Inspiration Vault, keeping research and creative references seamlessly connected.
Conversational assistant
The Co-Pilot lets marketers refine ideas through natural conversations, asking why certain elements were suggested, requesting alternatives, or adjusting tone and style. Designed as a conversational side panel with editable cards, it turns AI into a transparent, collaborative partner that adapts in real time to user input.
