FlowSpeech

**Project Description:** **What it is & The Problem it Solves:** FlowSpeech is a free, context-aware AI text-to-speech studio that brings stories to life. We noticed that most TTS tools sound robotic and lack emotional depth. FlowSpeech solves this by understanding the context of your text—automatically applying the right emotions (like joy, sadness, or excitement) and even assigning different voices to different characters in a dialogue. It’s perfect for creators who want to turn PDFs, EPUBs, or scripts into engaging audiobooks and podcasts without hours of manual editing. **How I Made It:** I built FlowSpeech by integrating advanced LLMs with a custom neural audio synthesis engine. The core challenge was creating a "Director Agent" that reads the text first to understand the mood and identify speakers. I then pipelined this metadata into our audio model to ensure the voice changes naturally based on the scene's context. **Tech Stack:** * **Frontend:** Next.js, React, Tailwind CSS * **Backend:** Python, FastAPI * **AI/ML:** PyTorch, Custom VITS Architecture, LLMs for semantic analysis * **Infrastructure:** Cloud GPUs for real-time inference