FreeTitle AI Video Production Agents
What it does
FreeTitle is a fully automated AI video production system powered by the Gemini. From a single prompt, it generates cinematic trailers through a multi-agent pipeline — writing scripts, designing characters, creating storyboards, generating video, self-validate results, and editing the final cut.
What’s more is what we build on top of that foundation: a Gemini-enabled multi-agent orchestration system purpose-built for video production, with a cinematic stylizer at its core. We do not aim to offer an “AI slop” machine that mass-produces hollow content. Instead, we take a user’s idea and deeply interpret the business goal and creative intent behind it, then apply domain-driven cinematic taste to create the signature “soul” of a directorial world—through audiovisual language, emotional atmosphere, narrative attitude, aesthetic identity, and editorial rhythm—pushing visual boundaries and delivering style and coherence that base models can reach only when expertly directed and precisely conditioned.
Challenges and Accomplishments
Production workflows are hard to automate. Video production requires creative judgment — what to do next, when to retry. Gemini acts as the director: reasoning about which agents to invoke and adapting the plan as the project evolves.
AI generation quality is unpredictable. Outputs can be random and unusable. Our agents close this loop autonomously — Gemini visually inspects every generated asset and regenerates when needed, turning unreliable generation into a self-correcting pipeline.
Visual consistency across many assets. Characters and environments must look consistent across scenes. A context management system feeds each agent relevant references and prior outputs, grounding every generation in the project's visual identity.
Producing dozens of assets is slow. Internal orchestrators within each agent plan batches, spawn parallel sub-agents per asset, and handle retries — while maintaining consistency across the set.
Fixed pipelines waste time. Our orchestrator analyzes dependencies and runs agents in parallel when there are no conflicts — cutting production time without sacrificing coherence.
Coordinating agents without information loss. Each agent operates with isolated context relevant to its task, reducing noise and keeping reasoning focused.
Models Used
- Gemini 3 Pro — the reasoning “brain” for orchestration, scriptwriting, and editing decisions, with multimodal understanding capabilities to analyze generated visuals, validate quality, detect issues, and drive intelligent regeneration and editing.
- Gemini 3 Pro Image — generates characters, storyboard frames, props, and environments with reference-guided visual consistency.
- Veo 3.1 — image-to-video generation to produce final motion shots from curated frames.
Inspiration
Having worked as an AI Engineer at a Hollywood movie studio, I've seen a real tension in the creative field around AI. For many, the resistance is a defense of something sacred: human taste, lived experience, and the hard-won craft of turning emotion into cinema. I share that feeling as I've been an independent filmmaker too, and I'm also proud of what humans can do that machines can't.
But I've also watched a different tragedy play out, quietly and repeatedly. Talented filmmakers with unique voices, and brands with real stories, simply cannot bring their vision to life because of resource constraints. Money, time, crew, access, production infrastructure. The gap between "I can see the film in my head" and "I can ship it" is brutal.
Generative AI changes that equation. It unlocks the ability to create cinematic content with far fewer constraints. However, the problem is that most AI creation today still feels like wrestling. Tools are fragmented, pipelines are messy, output can be random and failure rates are high, and prompts reward technical perfection over artistic intent. Creators end up acting like part-time engineers, spending their energy on wrangling workflows instead of making creative decisions.
That's why FreeTitle exists. The goal is to free creators from the tooling and let them operate where they're strongest: vision, taste, story, direction, and letting agents handle the pipelines. This lowers the barrier for people with creative vision who never had formal film training, and it gives brands and small businesses a path to premium storytelling they couldn't afford before. It also helps filmmakers move faster in pre-visualization, prove concepts, and win the resources needed for real production.
This is the stand we take: AI should not replace human creativity. It should remove the bottlenecks that keep human creativity from reaching the screen, and democratize cinematic production so more voices can create, iterate, and ship without gatekeepers or massive budgets.
This project was created independently as an individual entry for this hackathon competition.
Log in or sign up for Devpost to join the conversation.