ClipLab

• 

  Idea List

• 

  Script Idea Detail Part 1

• 

  Script Idea Detail Part 2

• 

  Architecture

Inspiration

Content creators today are under massive pressure to publish consistently across multiple platforms. Burnout is rampant, and pre-production tasks like scripting, captioning, and trend research eat up valuable time. I wanted to build a tool that acts like an AI-powered creative team — fast, collaborative, and always on.

What it does

ClipLab is an AI studio crew that turns a creator’s idea into a ready-to-publish content pack. With just a 3-sentence brief, ClipLab automatically generates a 60-second script, suggests trending audio, and writes an engaging caption with hashtags — all tailored to TikTok, Reels, or Shorts.

How I built it

I used the multi-agent system powered by Google’s ADK and Gemini models. MongoDB stores idea sessions, while each agent — script, music, and caption — runs asynchronously as a task. I also integrated Google Gemini for voice input and designed a UI for creators to submit briefs quickly.

Challenges I ran into

• Tuning prompts to generate platform-native content while staying under token limits
  
• Matching music trends accurately with vague or abstract video ideas.
  
• Managing agent sequencing and state sharing across Google ADK.
• Deploying the Google ADK to the GCP Infrastructure.

Accomplishments that I'm proud of

• Built a working multi-agent system with real-time content generation
  
• Created a seamless voice-to-video pipeline using Gemini Models and services.
• Designed a modular system that scales with new agents (e.g. scheduling, analytics)
  
• Made the process fun and creator-first — not overly technical
  

What I learned

• How to manage complex agent workflows with Google ADK
  
• Best practices for content-specific prompt engineering
  
• That creators value speed, tone-matching, and trend relevance above all else
  
• How to build fast-moving async systems with Next.js
• Using Queue System with an open-source service for serverless applications called Quirrel .
• Use Pusher to generate real-time notification to update the content Creator once its completed.

What's next for ClipLab

• Add a post-scheduler agent to publish directly via Buffer or Meta APIs
  
• Introduce a shot_planner_agent to turn scripts into scene-by-scene breakdowns
  
• Expand trend_hunter_agent to support localized and multilingual content
  
• Launch a beta with real creators and gather UX feedback
  

Built With

• adk
• google
• next.js
• quirrel
• railway
• shadcn