Story2Audio: Multimodal Story Generator

This project is a microservice-based NLP application that generates creative stories from a given topic and narrates them using text-to-speech. It includes:

• FastAPI REST endpoint (/generate)
• gRPC endpoint for high-performance interservice calls
• Text generation using Databricks/dolly v2
• TTS audio synthesis
• Tests for both REST and gRPC interfaces
• Gradio UI + Postman testing
• Dockerized for easy deployment

---

Features

• gRPC + FastAPI microservices
• Async concurrency and parallel request handling
• Error handling for invalid topics and internal issues
• Streaming and downloadable audio output
• Unit tests for REST and gRPC interfaces
• Frontend UI with Gradio
• Dockerized Deployment

---

Requirements

pip install -r requirements.txt

Key packages:

• transformers
• torch
• fastapi
• uvicorn
• gradio
• grpcio
• grpcio-tools
• pydantic
• requests
• edge-tts
• uuid
• asyncio

---

Usage

FastAPI

uvicorn app.main:app --reload

gRPC

python grpc_server.py

Run Tests

python tests/test_fastapi.py
python tests/test_grpc.py

Or run both via:

run_tests.bat

Gradio UI

python gradio_app.py

---

API Reference

POST /generate (FastAPI)

{
  "topic": "Apocalypse"
}

Response:

{
  "status": "success",
  "story": "...",
  "audio_file": "static/xyz.mp3"
}

gRPC Method: Generate(StoryRequest)

message StoryRequest {
  string topic = 1;
}
message StoryReply {
  string story = 1;
  string audio_file = 2;
}

---

Docker

Build

docker build -t storygen-app .

Run

docker run -p 8000:8000 -p 50051:50051 storygen-app

---

Sample UI (Gradio)

The Gradio interface allows:

• Topic input
• Inline audio streaming
• Downloadable MP3 file

---

Credits

Created by Talha Syed — BS Artificial Intelligence (2025) Course: Fundamentals of NLP (AI4001 / CS4063)

---

License

This project is academic-only and for educational use.