the council

• 

  Landing page

• 

Inspiration

Council.ai was inspired by the fundamental human need for diverse perspectives and wisdom when facing life's challenges. We recognized that while individual AI assistants are powerful, they can be limited by a single perspective. By combining multiple AI advisors with distinct personalities and expertise, we could create a more nuanced and comprehensive advisory experience. The project draws inspiration from historical "councils of elders" and the idea that complex decisions benefit from multiple viewpoints.

What it does

Council.ai creates a personalized advisory panel by:

Analyzing user responses to generate a tailored set of advisors matched to their needs Creating visual representations of each advisor using AI image generation Giving each advisor a unique voice through advanced voice synthesis Enabling interactive conversations where users can discuss their challenges with the entire panel Maintaining consistent personalities and perspectives for each advisor throughout the conversation

How we built it

We developed Council.ai using a modern tech stack designed for scalability and real-time interaction:

Built the frontend using Next.js with TypeScript for type safety and better development experience Implemented backend logic using Next.js API Routes for seamless full-stack integration Leveraged Supabase for user authentication and data persistence Orchestrated multiple AI services:

Used Anthropic Claude for advisor generation and core reasoning Integrated GPT-4 for natural conversation flow Implemented ElevenLabs for voice synthesis and character voice design Utilized Fal AI flux-lora for generating visual representations of advisors Added Whisper API for audio transcription

Challenges we ran into

Voice Synthesis Complexity: Navigating ElevenLabs' copyright restrictions required creative solutions, leading us to use Claude for generating voice characteristics instead of direct historical figure references Personality Consistency: Ensuring each advisor maintained consistent perspectives and speaking styles across conversations required careful prompt engineering Real-time Performance: Managing multiple API calls for text, voice, and image generation while maintaining responsive user experience Integration Complexity: Coordinating multiple AI services to work together seamlessly while handling potential failures gracefully

Accomplishments that we're proud of

Created a truly multi-modal AI experience combining text, voice, and visual elements Developed a creative workaround for voice copyright issues using Claude's descriptive capabilities Built a scalable architecture that can handle multiple concurrent users and conversations Successfully maintained distinct personalities for each advisor across different interaction modes

What we learned

Advanced prompt engineering techniques for maintaining consistent AI personalities Strategies for orchestrating multiple AI services in real-time applications Methods for handling API rate limits and failures gracefully Techniques for optimizing performance with multiple concurrent API calls Creative approaches to working within AI service limitations and restrictions

What's next

Expanding the advisor generation system to include more diverse perspectives and expertise Implementing group conversations where multiple users can interact with the council Adding memory systems to allow advisors to remember and reference past conversations Developing more sophisticated personality matching algorithms Creating mobile applications for iOS and Android

Built With

• elevenlabs
• fal.ai
• nextjs
• openai
• supabase