AI Director
Client
Decohere(YC-Backed)
Decohere: AI Director is a Video Generating and Editing Tool, powered by AI that transforms simple text into compelling short videos, all within capabilities of current technology.
Role
Product Manager
UX Research Lead
Product Designer
Team
Product Manager
UX Researcher/Designer
Software Engineers
Skills
Semi-structured Interview, Survey Design, Contextual Inquiry, Feature Prioritization, Prototyping, Competitive Analysis
Context
Decohere's current product, Turbo, which generates text-to-4-second clips in real-time, attracts many new users. However, while many users try it out, retention is low.
Project Goal
Impact
100%
On-time Delivery
Designed and delivered the live demo of the AI Director within 5 months, leading to client approval and a scheduled product launch in Q4 2024.
83%
Reduction in task completion time
Reduced the time to create a 30-second video from 60 minutes to 10 minutes in user testing.
90%
User satisfaction rate
In user testing with 26 participants (six paid users and 20 new users), 90% indicated their willingness to use the AI Director in the future.
Project Process
Discover
Define
Develop
Deliver
Discover
Research Question
What are the gaps between current user experience and users’ expectations of AI video creation?
Sub-questions:
Who are the users?
What is the AI video creation process?
How is the current user experience?
What do users expect and desire to do with AI-driven video generating and editing products?
Research Process
Our research methodology encompasses surveys, in-depth interviews, contextual inquiry, netnography, and competitive analysis, to thoroughly understand user experience and market dynamics.
End-user
Key Findings
Survey(N=2,000)
The survey primarily focused on understanding who the users are, gaining insights into the AI video creation process, evaluating the current user experience, and identifying users' expectations and desires. This approach enables targeted improvements to enhance user retention and satisfaction.
Contextual Inquiry(N=6)
Contextual inquiry allowed us to observe users interacting with the AI video generation tool in their natural environment, capturing a holistic understanding of their interactions and challenges. By recruiting three users familiar with AI technologies but new to Decohere, we ensured diverse perspectives and comprehensive insights into the user journey with the tool.
In-depth Interview(N=4)
In-depth interviews provided insights into users' perspectives and expectations regarding AI video editing tools, uncovering gaps between desired experiences and current offerings. These interviews yielded rich qualitative data that informed the development of a user-centric AI video editing tool.
Netnography
To complement these methods, we also performed Netnography by analyzing user-generated content about AI video making on TikTok and YouTube. This analysis helped us understand the AI video creation process, the usage of AI-generated videos, users' current experiences, and their expectations for such AI tools.
Users want to make videos instead of animated images.
To create videos, users are seeking:
general help on prompting to create storylines and high-quality outputs.
consistently styled clips to compile short stories effectively.
ability to connect and make transitions among clips for seamless storytelling.
2. Users seek simplification and efficiency, but the current process is time-consuming and involves redundant steps when switching between platforms.
3. Users desire more control over specific video elements than AI automation.
For example, selective motion and camera angles, to fine-tune their videos.
Market
Key Findings
Competitive Analysis(N=4)
Competitive Analysis involved an in-depth UX audit of the current Decohere product, assessing its user experience characteristics and comparing them with competitors to identify strengths, shortcomings, and trends in the AI video generation and editing space. This analysis helped us develop a specific product strategy and prioritize our design approach by determining the market gap Decohere fills and how it can be positioned against competitors.
Strengths of Decohere are Quick Response & Extensive Previews
Quick Response: Decohere excels in the real-time generation of images, providing a quick response compared to competitors.
Extensive Previews: The product offers an extensive range of previews, which is a significant advantage over other tools.
2. The current market has a limited number of AI video-generating tools.
Decohere and its direct competitors are primarily GIF generators.
The market can be mapped along two key dimensions: image output vs. video output and efficiency vs. detailed control.
Technology
Key Findings
Expert Interview(N=2)
We also consulted with the engineers and CTO at Decohere to understand the technology restrictions. By seeking the experts' perspectives, we aimed to shape our product design with feasibility considerations in mind, ensuring that our proposed solutions are both innovative and technically achievable. This collaboration allowed us to align our design goals with the practical capabilities of Decohere’s technology.
Current technology cannot create high-quality videos longer than 10 seconds or effectively depict serial object movements.
Advanced controls, such as precise camera motion, are not yet available but are achievable with future developments.
It is feasible to implement basic video editing features to help users arrange and edit clips.
The model struggles to produce multiple outputs with consistent styles or elements due to limited control over randomness.
Define
Insights
Target User
User Need
Users want (a) simplified AI video creation VS (b) more detailed controls than AI automation.
Business Goal
This product should fill a unique niche by offering an AI tool that enhances the company's strength of efficiency.
Technical feasibility
It is feasible to produce videos by generating and arranging animated frames into a coherent sequence.
Product Value Proposition
Develop
“How can we leverage current LLM and GenAI technology to help users create videos more quickly and effortlessly?”
User flow
Currently, users create image previews from text and then generate the image into a 4-second clip(GIF).
When using the current AI model to create videos, it first generates individual animated frames(GIF) from text and then compile them into one video.
Brainstorm & Prioritization
Low-fi
Low-fi Wireframes
Based on the high-level user flow, we brainstormed the necessary features for each step in user flow and prioritized them based on importance and estimated efforts.
For the MVP, we focused on the most essential features required for product usability. Other necessary features that require more development effort are scheduled for Phase 1. Lastly, we deprioritized features that heavily rely on the technology capability advancement to enable quick delivery.
For all the features in the MVP and Phase 1, we began grouping them by different levels to facilitate the alignment of the information architecture.
We brainstormed and evaluated 10 variants based on user experience, business goals, and technical efforts, narrowing it down to two solutions. Finally, we collaborated with stakeholders to make the final decision collectively.
We chose Solution 2 because it better meets both business and user needs, despite requiring slightly more development effort.
Solution 2 allows users to easily create, sequence, and edit individual animated frames into cohesive video creations. In comparison, Solution 1 primarily focuses on creating and editing single animated frames.
Deliver
Final Design Overview
AI Script Assistant
The AI Script Assistant quickly generates video scene descriptions that capture users' vision and adapt AI language to produce high-quality scene video outputs. It can one-click assign text prompts to the scene card, making the process faster and easier.
Effortless Video Creation
Users can effortlessly add scenes, input text, choose a starting image from a wide range of options, and let the AI technology do the magic. The text becomes a video in seconds, allowing users to create as many scenes as they want.
Generate Multiple Scene Videos at Once
Users can select multiple scenes and let the product handle the rest. With one click, videos for selected scenes are generated automatically.
Add Transitions
Users can add a professional touch with transitions. With just a few clicks, the video flows smoothly from one scene to the next.
“How did we get there?”
Project Level
Low fi to High-fi
Edit project name and export are the two main features placed on the project navigation bar.
Overview Level
There are four main features on the overview bar: scene navigation, batch generate, add transition, and play video.
Scene navigation helps users quickly find and edit scenes while providing an overview of the scene content. Batch generate allows users to create multiple scenes with a single click. Add transition enables smooth transitions between scenes. Play video lets users view the video result while editing.
Scene Edit Level
The scene edit card includes four main features: moving and deleting scenes, inputting text prompts, selecting images, and generating clips.
In our design, we prioritized the information hierarchy by making the clip view the largest element, as it is the primary focus for users. The text input and image selection areas follow in importance. The text input space accommodates approximately 50 words, which optimizes the generation results
Script Level
We designed the landing page to educate users about the capabilities of the AI script assistant. It provides three action shortcuts, offering examples of how to interact with the AI script assistant.
Drawing inspiration from exemplars like Webflow and Notion, we recognize that users are familiar with interacting with ChatGPT as a pure conversational-based idea generator. To streamline the process, we added a one-click button for assigning scene descriptions, speeding up the workflow.
The script assistant utilizes a pop-up AI chatbot, enabling users to freely share their story ideas and requirements.
Usability Testing Finding
Iteration
The Batch Generate feature, designed to create multiple scenes with a single click, is confusing. Users often misinterpret its purpose and are unsure how to interact with it.
Problem 1: Naming
The name “Batch Generate" does not accurately reflect the feature's functionality. Users interpret it as combining clips into one video.
Problem 2: Entrance Point
Users lack expectations regarding the batch generate function before opening it. Additionally, they struggle to comprehend the supportive text in the pop-up, which gives the impression of completing a task rather than an editing step.
Problem 3: User Flow
The placement of the button is too close to the export option, leading users to perceive it as the final step rather than a middle step. Furthermore, its distance from the generate button in the scene card confuses users about their functionality.
Design Changes
Before
The Batch Generate feature is difficult to understand, as users struggle to grasp the meaning without prior experience with '“batch” in other products to indicate what it does.
After
The button now clearly indicates that it will generate videos for the selected scenes, helping users understand its function. Additionally, we used the same icon as the single scene generate button to provide users with a familiar reference.
1. Revised naming to "Generate [X] Scene Videos"
Before
After
2. Repositioned the entrance point closer to the editing area.
Before
After
3. Revised the user flow to require users to select scenes before pressing the generate button.
“How did we enable a faster hand-off and developing process?”
Product Requirements Document
As the product manager, I developed a comprehensive PRD (Product Requirements Document) that included essential information for development, such as basic product details and functional requirements.
This document served as a detailed guide for the development team, outlining high-level epics and specific features with thorough descriptions. Each feature was accompanied by user stories, notes, acceptance criteria, priority levels, and design elements. Additionally, I indicated the necessity of each feature for the MVP (Minimum Viable Product). To ensure clarity and a more comprehensive understanding of the product functionalities, I also provided interactive prototypes. This structured and detailed approach streamlined communication and facilitated a more efficient development process.
Design System
Additionally, we converted the existing color and typography system from code into a design system in Figma to streamline the design process and minimize confusion during development. To expedite future engineering efforts, we also developed a design system for commonly used icons and components, ensuring the engineering team can easily reuse them in future projects.
Video
We created this promotional video to attract new and existing users to explore our new product. It can also serve as a simple and efficient tutorial.
Demo
The engineers developed a Minimum Marketable Viable Product (MMVP) demo for us to introduce and test live with users. Feel free to try the demo by clicking the link here: AI Director Demo
AI Director Demo
Pic: guests trying the demo on showcase day:)
Takeaways
Effective Stakeholder Management
One of the most significant lessons from this project was the importance of effective stakeholder management. Coordinating with various stakeholders, including the startup founders, engineers, designers, and end-users, required clear communication and a deep understanding of each party's needs and expectations. Regular updates, transparent decision-making processes, and active listening were crucial in building trust and ensuring alignment across the board. This experience underscored the value of fostering strong relationships and maintaining open lines of communication to navigate complex project dynamics successfully.
Comprehensive Product Development
Another key takeaway was the application of my product management skills to drive the development of this new product. I ensured that we met user needs, aligned with business goals, and adhered to technological feasibility while integrating responsible AI practices. This involved balancing user experience, business objectives, and ethical considerations, leading to a well-rounded and impactful product. By carefully navigating these aspects, I was able to significantly influence the product’s direction and success, demonstrating the importance of a holistic and responsible approach to product development.