CreatorPilot

Turn YouTube ideas into research-backed, production-ready plans.

Comment

Inspiration

YouTube creators often have to act as a researcher, writer, producer, director, and editor at the same time. A simple video idea can quickly become overwhelming when they need to research a topic, structure the content, plan visuals, manage limited equipment, and fit everything into a realistic production schedule.

We wanted to build an AI agent that does more than generate a script. CreatorPilot acts as a production manager that turns a creator's idea into a realistic, actionable production plan.

What it does

CreatorPilot takes a YouTube video idea along with optional creator constraints such as budget, available time, crew size, equipment, location, and experience level.

The agent:

  1. Understands the creator's objective.
  2. Determines whether current external research is necessary.
  3. Uses Parallel Search when research is required.
  4. Analyzes the retrieved information.
  5. Creates a content strategy and video structure.
  6. Generates a practical shot and B-roll plan.
  7. Estimates production feasibility.
  8. Adapts the plan to the creator's actual constraints.
  9. Creates a production schedule.
  10. Identifies research and production risks.
  11. Produces a final production checklist.

The key idea is that CreatorPilot does not give every creator the same generic plan. It reasons about the creator's actual resources and adapts the production strategy accordingly.

How we built it

CreatorPilot uses Gemini with Google ADK and Google Cloud as the AI and agent infrastructure, with FastAPI as the backend. The React/Vite frontend provides the creator interface.

The agent uses the Parallel Search API as a runtime tool. When a topic requires current information, the agent decides whether to call Parallel Search, retrieves relevant sources, and uses those results when creating the production plan.

The application is deployed as a web application so creators can interact with the agent directly.

The core workflow is:

Creator Idea
→ Gemini / Google ADK Agent
→ Research Decision
→ Parallel Search when needed
→ Research Analysis
→ Content Strategy
→ Video Structure
→ Shot & B-roll Plan
→ Constraint Adaptation
→ Feasibility Analysis
→ Production Schedule
→ Risks & Checklist

What we learned

Building CreatorPilot showed us that an agent becomes much more useful when it can make decisions based on context and constraints rather than simply generating text.

We also learned the importance of tool usage: current information should be retrieved when needed instead of relying only on potentially outdated model knowledge.

Challenges

One of the main challenges was designing the workflow so that research is used only when it adds value. A fictional comedy skit, for example, does not need the same external research as a video about a rapidly changing technology topic.

Another challenge was making the production plan realistic. A solo creator with a smartphone and four hours available should not receive a production plan designed for a professional film crew.

We solved this by making creator constraints part of the agent's decision-making process, allowing the plan to adapt to the creator's actual situation.

What's next

Future versions could connect the production plan to publishing workflows, project history, collaboration, thumbnail planning, and deeper creator analytics.

Built With

Share this project:

Updates

Submission history