Full case study breakdown
01 Overview
Project overview, host, purpose, and verified deliverable

Featured portfolio case study
AI-Powered Virtual Podcast Production
Steel City Garage Podcast is presented here as a complete AI-assisted video production system that I developed, rather than simply as a finished video. It demonstrates my skills in creative direction, AI video production, character development, scriptwriting, visual storyboarding, prompt engineering, voiceover performance, AI voice conversion, virtual set design, image-to-video generation, video editing, character and environmental continuity, and production pipeline development.
Featured final video
Process snapshot
These five pages now sit at the front of the process flow so viewers can immediately understand the complete Steel City Garage system — from the overview and production challenge to the character/studio rules, three-frame storyboarding method, and the final performance-driven voice-to-video pipeline.
Full case study breakdown
Project overview, host, purpose, and verified deliverable

Full case study breakdown
Core production constraints, continuity risks, and system goal

Full case study breakdown
Reference roles for character identity, studio, props, and framing

Full case study breakdown
Directed start, movement, and settle beats for every segment

Full case study breakdown
Human-led acting, AI voice conversion, video generation, and review

My role
I directed the project from initial concept through final delivery. I developed the host character, designed the production environment, wrote and structured the episode, created three-frame storyboards for every segment, recorded the original voice performance, developed the AI voice-conversion workflow, generated the video material, and edited the completed production.
Project overview
Steel City Garage Podcast is a fictional financial-market podcast developed as a repeatable AI video content series. The show is hosted by Grant Keller, a blue-collar investor, father, and Little League coach from Pittsburgh. His character was designed to make financial discussions feel understandable, practical, and relatable to everyday viewers. I developed Grant’s personality, visual identity, background story, speaking style, wardrobe, mannerisms, and on-camera presentation. The project required the creation of a consistent virtual host, a reusable podcast studio, episodic scripts, visual storyboards, voice performances, and multiple AI-generated video segments that could be edited into a complete podcast episode.
Character development
Grant Keller was developed as more than a visual avatar. He was given a complete background, personality, profession, family life, regional identity, speaking style, and personal philosophy. His identity as a Pittsburgh-area father, former contractor, investor, and Little League coach influenced his wardrobe, dialogue, gestures, financial analogies, and overall presentation. The supplied character sheet was used as the master reference for maintaining his facial appearance, body type, hair, beard, clothing, and overall visual identity.
Virtual studio design
I designed the Steel City Garage podcast studio as a reusable virtual production environment. The set includes a custom podcast desk, microphone setup, branded neon sign, shelving, studio decorations, market-related visual elements, and controlled lighting. A detailed scene sheet was created to establish the studio layout, camera angles, object placement, scale, lighting, and visual continuity. This allowed every generated segment to appear as though it had been recorded in the same physical studio.

Virtual production design
Three-frame storyboard system
Before generating each video segment, I created a three-image storyboard consisting of a start frame, middle frame, and end frame. These storyboards established the character’s starting posture, facial expression, hand movement, speaking behavior, and final resting position. The final frame of each segment was designed to support visual continuity with the following clip.

Three-frame storyboard
Panel 1
Set the posture, lens framing, facial expression, and first beat of the line delivery so every segment begins with a controlled composition.
Panel 2
Plan the speaking gesture, eye-line, and resting pose so the final frame can hand off smoothly into the next generated segment without visual drift.
Production detail
The episode was divided into modular segments designed around the time limitations of AI video-generation platforms. Each section was written to fit within approximately 15 seconds while still sounding like part of a natural, continuous podcast conversation. The scripts included host introductions, caller questions, financial explanations, natural transitions, promotional messaging, episode introductions and conclusions, and calls to action. The modular structure allowed individual clips to be generated, corrected, replaced, and edited without rebuilding the entire episode.
Production detail
I performed and recorded the original dialogue for Grant Keller. The recorded performance established the timing, pacing, emotional delivery, inflection, pauses, and personality of the character. The performance was then processed through an AI voice-conversion workflow to create Grant’s final character voice while preserving the intention and rhythm of my original delivery. The underlying acting and voice performance were produced by me, with AI voice technology used as part of the character-production pipeline.
Production detail
Each segment was generated using detailed reference-based prompting. The prompting system controlled character identity, facial consistency, studio reconstruction, wardrobe continuity, camera framing, lighting, lip synchronization, gestures, speaker behavior, starting composition, ending composition, background objects, and audio routing. Master reference images were assigned specific purposes so the AI system could distinguish between character identity, studio design, starting position, props, and segment-specific storyboards. This helped prevent visual drift and created a more consistent final production.
Production detail
The individually generated clips were assembled into a complete podcast presentation. My post-production work included selecting the strongest generations, correcting pacing, trimming dialogue, creating transitions, synchronizing audio, balancing voice levels, adding the show introduction and conclusion, maintaining visual continuity, assembling the final episode, and preparing the completed video for web and social-media presentation.
Production pipeline
Step 1
Concept Development
Step 2
Character Design
Step 3
Studio and Scene Design
Step 4
Scriptwriting
Step 5
Three-Frame Storyboarding
Step 6
Voice Performance
Step 7
AI Voice Conversion
Step 8
Reference-Based Video Generation
Step 9
Editing and Final Delivery
Skills and technologies
Key project outcome
Steel City Garage demonstrates my ability to combine traditional creative-production skills with modern AI tools to build a consistent and repeatable virtual-content pipeline. The final result is not simply an isolated AI-generated video. It is a scalable production framework capable of supporting future podcast episodes, market updates, branded content, educational videos, and character-driven social-media campaigns.