-
-
CutLine minimized for a full-screen view
Inspiration
Filmmaking involves a lot more than writing a good story. A single scene can create unexpected problems with locations, vehicles, logistics, permissions, or other production requirements.
We wanted to build something that could catch those problems before production starts, instead of discovering them when they become expensive.
That idea became CutLine.
What it does
CutLine is an AI agent for film pre-production.
You give it a screenplay, and it analyzes the story to find production requirements that may need further investigation. It then researches the real world, gathers evidence, evaluates the findings against production rules, and gives you an actionable recommendation.
Instead of just saying "this might be difficult," CutLine explains why, shows the evidence, and suggests what could be done about it.
How we built it
We built CutLine around Gemini and Vertex AI for screenplay understanding and reasoning, with Parallel handling real-world research.
The agent identifies what needs to be investigated, collects relevant information, and passes the findings through deterministic rules before producing the final recommendation.
We wanted the LLM to handle the reasoning while keeping important decisions consistent, explainable, and evidence-based.
Challenges we ran into
The biggest challenge was connecting a creative screenplay with messy real-world information.
A few words in a script can require a surprising amount of research to determine whether something is actually practical or expensive to produce.
We also had to make sure the agent didn't turn uncertain information into confident conclusions, which is why evidence and deterministic rules became an important part of the system.
Accomplishments that we're proud of
We're proud that CutLine goes beyond simply asking an LLM questions about a screenplay.
It can identify what needs to be researched, investigate it, evaluate the evidence, and turn that research into a production decision.
Most importantly, we built something we could actually use ourselves rather than just a proof-of-concept that looks good in a demo.
What we learned
We learned that building a good agent isn't just about making an LLM more autonomous.
The real challenge is figuring out when the model should reason, when it should research, and when deterministic logic should take over.
We also learned that AI becomes much more useful when it doesn't just generate an answer, but helps someone make a better decision.
What's next for CutLine
We want CutLine to become a much more complete AI production partner.
The next step is expanding the types of production risks it can identify and the decisions it can help with, while making the recommendations even more accurate and useful for real filmmakers.
Our goal is simple:
Find the problems before the cameras roll.
Log in or sign up for Devpost to join the conversation.