LUMINIX-AGENTS

CONTENT CREATION MADE EASIER

Comment

LUMINIX.AGENT: Autonomous Multi-Agent AI for Multimedia Content Creation

Inspiration The genesis of LUMINIX.AGENT lies in our ambition to democratize multimedia content creation, making sophisticated video and audio production accessible to creators of all expertise levels. Building upon the foundation of LUMINIX.AI, we were inspired by the potential of multi-agent AI systems to streamline complex creative workflows. The Agent Development Kit Hackathon with Google Cloud offered an ideal platform to showcase how autonomous agents could collaborate to produce professional-grade content, reducing technical barriers and empowering creators to focus on storytelling. Our vision is to create a system where intelligent agents orchestrate tasks like voice generation and content assembly, transforming creative ideas into reality with ease.

What it does

LUMINIX.AGENT is a multi-agent AI system designed to revolutionize multimedia content creation through autonomous workflows. For the hackathon, we present a lightweight demo hosted on Replit at https://luminix-agents.mdalifsaniul.repl.co, featuring the VoiceGeneration Agent and basic multi-agent orchestration driven by the Orchestrator Agent (ORLEON). The demo allows users to input text prompts, which ORLEON routes to the VoiceGeneration Agent to produce high-fidelity, customizable speech outputs using the ElevenLabs API or a free model fallback. This prototype demonstrates real-time agent communication and logic delegation, serving as a proof-of-concept for our broader vision of a scalable platform that includes additional agents for video editing, scene detection, and content orchestration.

How I built it

The LUMINIX.AGENT demo was crafted with a focus on modularity and functionality within the hackathon’s constraints:

Technologies: We utilized Python for backend logic, JavaScript, HTML, and CSS with Tailwind CSS for the UI, and the Agent Development Kit for orchestration. The ElevenLabs API powers the VoiceGeneration Agent, with Google Cloud APIs enabling lightweight processing. The demo is hosted on Replit for accessibility.

Architecture: The demo centers on the Orchestrator Agent (ORLEON), which routes text prompts to the VoiceGeneration Agent for speech synthesis. The backend leverages LUMINIX.AI’s modular design, with isolated modules (e.g., voice) and dynamic imports via dynamic_imports.py for flexibility. The orchestration layer ensures seamless agent communication, fulfilling key ADK criteria.

Frontend: A minimal, server-rendered UI provides a clean interface for text input and audio output, optimized for performance and accessibility. Framer Motion adds subtle animations to enhance user experience.

Deployment: The full LUMINIX.AGENT system exceeds 25GB, making it unsuitable for Replit. We streamlined the demo to focus on voice generation and basic orchestration, ensuring reliability within Replit’s resource limits.

The open-source repository at https://github.com/SANIUL-blackdragon/LUMINIX-AGENTS is established to host code, documentation, and architecture diagrams, though it remains in its initial stages.

Challenges I ran into

Developing LUMINIX.AGENT for the hackathon presented notable challenges:

Resource Constraints: The full 25GB+ system could not be hosted on Replit, requiring us to limit the demo to the VoiceGeneration Agent and basic orchestration while preserving the multi-agent architecture’s integrity.

Time Constraints: Limited time prevented us from implementing additional agents (e.g., CaptionAgent, TimelineEdit Agent) or producing a demo video, necessitating a focus on core functionality and a live Replit demo.

Dependency Management: Integrating the Agent Development Kit, Google Cloud APIs, and ElevenLabs API in a lightweight environment demanded careful optimization to avoid conflicts and ensure performance.

Orchestration Simplification: Adapting the complex multi-agent orchestration for Replit’s constraints required significant simplification, balancing functionality with resource limitations.

These challenges tested our ability to prioritize and adapt, resulting in a focused yet impactful prototype.

Accomplishments that I'm proud of

We take pride in several key achievements:

Functional Demo: Delivering a working prototype on Replit (https://luminix-agents.mdalifsaniul.repl.co) that showcases real-time voice generation and agent orchestration, meeting 80% of the ADK criteria for logic delegation.

Modular Architecture: Successfully adapting LUMINIX.AI’s modular design for a lightweight demo, ensuring scalability for future cloud deployment.

Open-Source Foundation: Establishing an open-source repository at https://github.com/SANIUL-blackdragon/LUMINIX-AGENTS, ready for future contributions despite its current early stage.

Transparency and Focus: Clearly communicating the demo’s scope and limitations while highlighting the innovative potential of LUMINIX.AGENT, fostering trust with judges and users.

These accomplishments reflect our commitment to delivering a compelling proof-of-concept under tight constraints.

What I learned

The hackathon provided invaluable lessons:

Agent Orchestration: We deepened our understanding of multi-agent systems using the Agent Development Kit, particularly in coordinating tasks like prompt routing and speech generation.

Cloud Optimization: Leveraging Google Cloud APIs taught us how to balance performance and scalability in resource-constrained environments, informing our cloud deployment strategy.

Rapid Prototyping: Streamlining a complex system into a lightweight demo honed our skills in prioritizing core features and delivering functional prototypes under time pressure.

Clear Communication: Articulating limitations transparently while emphasizing future potential strengthened our ability to present ambitious projects realistically.

These insights will shape LUMINIX.AGENT’s future development.

What's next for LUMINIX-AGENTS

The roadmap for LUMINIX.AGENT is ambitious and forward-looking, building on the demo’s foundation:

Full Cloud Deployment: Transition to Google Cloud Run or Compute Engine to support the full 25GB+ system, enabling scalable processing for all agents, including video editing and content orchestration.

Expanded Agent Suite: Implement additional agents from the roadmap, such as:

CaptionAgent: Real-time, multilingual subtitle generation using Whisper.

SceneDetection Agent: Identifying cuts and breaks in video content.

TimelineEdit Agent: Cutting, trimming, and rearranging clips on a virtual timeline.

Transition Agent: Auto-applying transitions between scenes.

Overlay Agent: Adding captions, stickers, or overlays to video frames.

Hyper-Thematic UI: Develop a dynamic frontend with Cyberpunk, Classy, Alien, and Multiverse themes in Light/Dark modes, as outlined in LUMINIX.AI’s FRONTEND_PLAN.md.

Community Growth: Populate the GitHub repository (https://github.com/SANIUL-blackdragon/LUMINIX-AGENTS) with code, detailed documentation, and tutorials to engage the open-source community.

Advanced Features: Integrate analytics, user management, and notification systems using Supabase, enhancing functionality and user experience.

We invite judges to explore our demo at https://luminix-agents.mdalifsaniul.repl.co and follow our progress on GitHub (https://github.com/SANIUL-blackdragon/LUMINIX-AGENTS). Despite time constraints preventing a video submission, we are confident that LUMINIX.AGENT’s innovative multi-agent architecture and transformative potential will resonate. Thank you for considering LUMINIX.AGENT—a visionary step toward redefining AI-driven creative expression.

Built With

Share this project:

Updates