Newsfluencer

My Role

Design, character generation, and prompt engineering co-lead

I led the design and prototyping process, using Figma to create a simple, intuitive interface for journalists to generate AI-driven video content. I worked closely with Daniela and Hank to define character personas and conduct extensive prompt testing to fine-tune the LLM’s script generation capabilities.

Action

We began with an idea pursued by the studio group that worked on this project during the previous quarter: to create AI-generated influencers who would be the "face" of local newsrooms. This solution would leverage the influential power of influencer-centered content to help local journalists attract, engage, and inform their communities.

To build on the progress of the previous quarter’s group, I collaborated with Daniela and Hank on the development of six AI character profiles. Through iterative design, we refined the characters’ backstories and appearances, testing and adjusting to make them relatable and believable. These characters ranged from a 24-year-old TikTok influencer to a 60-year-old Midwestern farmer, representing diverse political views, life experiences, and communication styles to resonate with various local audiences.

Next, Daniela and I focused on refining the script generation, where AI would transform an article into an engaging, informative short-form video script. We faced challenges with consistency, as the models sometimes contradicted character personalities or produced inaccurate summaries. We refined the input prompts, testing variations until we found one that consistently generated authentic, engaging results. The prompt “Can you summarize this article into a fast-paced, 1-minute, attention-grabbing story in your style and personality?” became the key to maintaining character consistency.

We worked closely with Conor, Jason, and Cypher, who handled backend processes, ensuring seamless integration with the AI system for a smooth user experience. A key technical consideration was maintaining the characters' roles as narrators rather than influencers, which guided our backend development choices. This approach helped us avoid uncanny valley issues that often arise when AI tries to replicate human influencers' personality-driven content. The focus on narration rather than personality-driven content also simplified our technical architecture, as we could optimize for clear storytelling rather than complex personality simulation.

With character and content generation established, I shifted to designing the platform’s user interface. I led the design in Figma, focusing on simplicity and ease of use to ensure that journalists with minimal technical expertise could easily generate AI-driven video content. The goal was for users to select a character, paste an article, and generate a 60-second video with narration and visuals in a few clicks.

Throughout the project, we gathered feedback from classmates, industry professionals, and experts, including a guest from The Wall Street Journal’s AI team. This input helped refine the platform, ensuring it aligned with the needs of local newsrooms and was both innovative and practical.

Due to time constraints, we couldn’t fully complete the automation. As a result, our final deliverable was the Figma prototype, along with a video (created by Hank) demonstrating the platform’s potential output, which it would be able to generate at scale.

Ultimately, my contributions in character design, prompt refinement, and UI/UX design helped prototype a platform that empowers local newsrooms to quickly produce AI-driven video content. By focusing on simplicity and accessibility, we created a tool that would support the goal of expanding local news' reach while staying authentic to its audience.

Result

An open-source tool enabling newsrooms to create character-driven, 60-second video summaries of news articles using diverse AI-generated personas.

Paste article link in platform

Select the persona that aligns with your target audience

Persona hover state to view character summary

View the script generated from your article and persona, and easily regenerate or edit manually

Loading screen to allow for automated backend to generate short-form video

View final output with options to share to social media, regenerate, or start over entirely

Scroll to top

© 2026 Aaron Boorstein