Case Study · Logline AI

Co-founded an AI startup that built professional pitch decks from a single sentence.

In Hollywood, building a pitch deck takes a filmmaker one to six months, hundreds to thousands of dollars in design fees, and a real chance the result still misses. I co-founded Logline AI to compress that to thirty seconds.

Role

Co-Founder

Stage

MVP · USC Spark Award

Tech

GPT · Stable Diffusion · Midjourney

Timeline

2023 - 2025

Logline AI alpha. A user enters a one-sentence logline. The platform generates story, characters, world, visuals, and a web-based pitch deck.

The Problem

I have produced eight feature films and a dozen-plus commercials over a fifteen-year career. The pitch deck is the most underrated bottleneck in the entire film industry. It is the thing every executive asks for, the thing every filmmaker dreads making, and the thing that decides whether a project moves forward or dies on a hard drive.

A professional pitch deck takes one to six months to build. It eats hundreds to thousands of dollars in graphic-designer fees, with no guarantee the designer is reliable or the result is on-brief. By the time the deck is ready, the executive has often moved on. Speed and quality are at war. Independent filmmakers usually lose.

Generative AI changed the math. In 2023 the tools to fix this finally existed, but no one had assembled them into a workflow a filmmaker could actually use. So we did.

What We Built

Logline AI was a web platform that turned a single sentence into a complete pitch deck. The user typed a logline. The platform generated a fully realized creative world (story, characters, tone, look, world rules, filmmaker statement, even episode breakdowns), then generated original artwork tied to that world, then assembled everything into a polished Web 2.0 pitch deck with the option to export as PDF.

First draft in roughly thirty seconds. From there the filmmaker could edit, rewrite, regenerate, and direct, treating the AI output as a foundation rather than a finish line.

Logline AI generated pitch deck screen Logline AI generated pitch deck screen Logline AI generated pitch deck screen
Generated pitch deck output from the alpha platform.

The AI Pipeline

The product was an orchestration problem more than a single-model problem. We chained multiple generative systems into one workflow:

Narrative Generation
ChatGPT API. From the user's logline, generates a cohesive story, characters, tone, look, world, meaning, and filmmaker statement.
Visual Prompt Engineering
A second ChatGPT pass that turns each narrative element into image prompts designed for diffusion models.
Image Generation
Stable Diffusion and Midjourney APIs for character art, environments, mood boards, and key art. Tuned for film-industry visual grammar.
Deck Assembly
ReactJS / NextJS front end. Outputs a Web 2.0 pitch deck site with PDF export, deployed via Vercel.
User Iteration
An editor layer that let the filmmaker rewrite copy, swap visuals, regenerate sections, and curate the output toward their own vision.
Roadmap
Planned next: music, video clips, and interactive elements for an enterprise tier with full Web 3.0 functionality.

What Made This Hard

Tonal Consistency Across a Deck
A pitch deck only works if every element feels like the same project. Generating thirty good images is easy. Generating thirty images that look like they belong to the same film is the hard part. We had to thread the visual prompts together so the model treated the entire deck as one creative world, not thirty separate ones.
Speed vs. Quality on Image Generation
Stable Diffusion in 2023 was great at certain things and notoriously bad at others: groups of people, complex landscapes, hands. We engineered around the weaknesses by constraining what the deck would ask the model to generate in the first place, and by making the editor strong enough that filmmakers could direct toward what worked.
A Tool That Threatens Its Buyer
Many of the people whose work Logline AI would replace were the same people whose feedback we needed. Navigating that, positioning the product as a force multiplier for filmmakers rather than a replacement for crew, was as much a creative problem as a technical one.
Building Inside a Moving Industry
Mid-build, the 2023 WGA and SAG-AFTRA strikes reshaped the industry's posture on AI overnight. We had to keep building while staying credible with the same creative community we were trying to serve.

The lesson I took out of Logline AI: in AI media, the product is the pipeline. Picking the right model matters less than orchestrating models cleanly, designing prompts for the specific creative output you want, and building an editor strong enough that the human stays in charge. Everything I do now in AI production, including the work I lead at Sawhorse and the Microsoft AI Persona pipeline, traces back to what we built here.

Conversations That Opened Up

Within twelve months, Logline AI was in front of decision-makers across the industry:

01Demoed to professionals at Netflix, A24, Allen Media Group, and Wonder Street Management
02Featured on the Making Movies Is Hard Podcast
03Beta-tested with USC filmmakers and producers
04Selected as a Spark Award recipient by the USC Center for Generative AI & Society
05Keynote at ShowCAIS 2024: "Democratizing Hollywood: Envisioning a Future with Singular AI-Driven Film Creation"
USC Center for Generative AI & Society Spark Awards Launch Reception
Spark Awards Launch Reception. USC Center for Generative AI & Society, February 2024.

Recognition

Spark Award Recipient
USC Center for Generative AI & Society · AI for Media & Storytelling (AIMS) Initiative
2023 - 2024
Featured Project
Making Movies Is Hard Podcast
2024
Keynote: Democratizing Hollywood
ShowCAIS · USC Center for AI in Society
2024

What Happened

Logline AI received institutional funding through the USC Spark Award and generated strong interest from across the entertainment industry. Development paused in 2025 due to a co-founder's medical leave. The technology and the team's learnings live on in Stephen's current AI production work, including the pipelines built for Microsoft and the AI strategy he leads at Sawhorse Productions.

The Team

Stephen Gibler
Co-Founder
Charles Sims
Strategic Advisor · ex-CTO, UTA & LA Clippers
Paul Gibler
Founding Engineer · ex-Meta
Patrick McAnneny
Founding Engineer · ex-HubSpot
Emilia Doda
Developer
Celina Louissaint
Developer
Liam Spaeth
Advisor · Microsoft / BlueMetal Architects
Dr. Ziyaad Bhorat
Advisor · USC AI & Society / Mozilla