Animal-Soul

Our agent transforms animal-inspired voices into expressive, English-speaking characters using multimodal AI and real-time audio generation. The working prototype is stable, producing consistent dog- and parrot-style speech through controlled acoustic prompting and voice design. The system orchestrates LLM reasoning, voice synthesis, and browser-based interactions, demonstrating both technical complexity and creative innovation. This playful yet functional approach shows strong real-world potential for entertainment, accessibility, and interactive agents.

We built on existing foundational tools such as ElevenLabs TTS, OpenAI LLMs, and standard Next.js infrastructure, but no project-specific code or voice models existed before the hackathon.
All animal-inspired behaviors, acoustic prompt design, and the core “Animal-Soul” architecture were created during the event.
We developed the full real-time pipeline—LLM reasoning, voice generation, audio handling, and playback—from scratch.
The UI, interaction flow, and character system were entirely built during the 3-hour window.
No prior prototypes or custom ML models were reused; everything in the demo reflects new hackathon work.
Everything shown in the demo was developed within the 3-hour hackathon.