TIMBRE! — Advanced AI Audio & Text Toolkit 🚀

TIMBRE! is a high-performance, modular AI application designed to bridge the gap between speech and text. Built with a "Privacy-First" and "Efficiency-First" mindset, Timbre leverages state-of-the-art (SOTA) models like Google Gemini 2.5 Flash and OpenAI Whisper to provide a seamless multilingual experience.

🚀 Live App

Timbre: https://timbreai.streamlit.app

Video Demo: cooming soon

• landing page

🌟 Key Features

Timbre is organized into four core processing services, accessible via an intuitive "Electric Cyan" themed dashboard:

• Speech to Text (STT): High-accuracy transcription using the OpenAI Whisper (Small) model. Optimized for low-latency performance on cloud environments.
  • Audio → Text → Grammar enhancement
  • screenshot
• Text to Speech (TTS): Generate natural-sounding neural voices in 100+ languages and accents. Features real-time streaming and adjustable speech rate/pitch and vocie selection.
  • Audio → Text
  • screenshot
• 🔄 Speech to Speech (STS):
  • Same-language conversion: Transform your voice into different neural personalities.
    • Audio → Text → Grammar enhancement → Audio.
    • screenshot
  • Cross-language translation: Speak in one language and hear the output in another, preserving clarity and context.
    • Audio → Text → Translation → Grammar enhancement → Audio.
    • screenshot
• Text Polisher:
  • Translator: Professional-grade translation that preserves formatting.
    • Text → Translation
    • screenshot
  • Enhancer: Refine grammar and syntax or completely transform the tone (Professional, Creative, Technical, etc.) using Gemini 2.5 Flash.
    • Text → Grammar or Style enhancement
    • screenshot

🛠️ The Tech Stack

Component	Technology
Frontend	Streamlit - (Custom UX/UI with Dark Mode & High Contrast)
Intelligence	Google Gemini 2.5 Flash - (Advanced reasoning, translation & text enhancement)
Transcription	OpenAI Whisper – (Small model optimized for low-RAM cloud environments)
Neural Voices	Edge-TTS – (Microsoft Azure Neural Voices)
Backend Logic	Asynchronous Python - (Asyncio), Object-Oriented Programming - (OOP)
Infrastructure	FFmpeg - (Audio processing engine), Streamlit Community Cloud

🏗️ Modular Architecture

Timbre-web/
├── .streamlit/
│   └──config.toml             # Custom Theme
├── data/
│   └── voices.json            # 100+ Neural voice metadata
├── timbre_module/             # Core Logic Package
│   ├── __init__.py            # Package initialization
│   ├── polisher.py            # Gemini 2.5 Flash Integration
│   ├── stt.py                 # Whisper STT Logic
│   └── tts.py                 # Edge-TTS Async Logic
├── scripts/                   # Automation
│   └── voice_extractor.py     # Automated data pipeline for voices
├── app.py                     # Streamlit Main Interface
├── .gitignore
├── requirements.txt           # Python Dependencies
├── packages.txt               # OS-level dependencies (ffmpeg)
└── LICENSE.txt                 

---

⚙️ Local Installation

1️⃣ Clone the repository

git clone https://github.com/thegksingh/Timbre-web.git
cd Timbre-web

2️⃣ Create a virtual environment

python -m venv venv

3️⃣ Activate the virtual environment

On macOS/Linux:

source venv/bin/activate

On Windows:

.\venv\Scripts\activate

4️⃣ Install dependencies

pip install -r requirements.txt

5️⃣ Set up your API Key

echo "GOOGLE_API_KEY=your_actual_key_goes_here" > .env

6️⃣ Run the app:

streamlit run app.py

👤 About the Author

Gaurav Kumar Singh B.Sc (Hons.) Botany Student at Banaras Hindu University (BHU) I am a self-taught developer passionate about building cross-functional AI tools. My journey started with CS50P (Intro to Programming with Python) and continues through CS50x. Timbre represents my ability to apply complex computer science concepts—like asynchronous programming, data parsing, and LLM integration—to real-world problems.

📜 License

Distributed under the MIT License. See LICENSE for more information.