🎙️ CSM-1B Gradio Interface 🎧

A user-friendly Gradio interface for Sesame's CSM-1B model that allows you to easily generate conversations and monologues using Conversational Speech Model technology.

✨ Features

• 🗣️ Multi-Speaker Conversations: Generate natural-sounding conversations between up to 10 speakers
• 🎭 Voice Cloning: Upload your own voice samples to personalize the generated speech
• 🔊 Built-in Voices: Use the included reference voices or generate random voices
• 📝 Simple JSON Input: Easily format your conversations with a simple JSON structure
• 🎚️ Advanced Controls: Fine-tune generation parameters like temperature and audio length
• 🌐 Web Interface: Intuitive UI powered by Gradio

🚀 Quick Start

Prerequisites

• Python 3.10 (recommended)
• CUDA-compatible GPU (for optimal performance)
• ffmpeg installed on your system

Installation

1. Clone this repository: bash git clone https://github.com/yourusername/c3-csm-gradio.git cd c3-csm-gradio

2. Install the required dependencies: bash pip install -r requirements.txt

3. Authenticate with Hugging Face (to access the model): bash huggingface-cli login

4. Launch the application: bash python app.py

5. Open your browser and navigate to http://localhost:7860

🧩 Usage Examples

Conversation Mode

Create conversations between multiple speakers using this JSON format:

json [ {"speaker_id": 0, "text": "This voice synthesis is amazing!"}, {"speaker_id": 1, "text": "I agree, it sounds so natural!"}, {"speaker_id": 2, "text": "And it's simple to customize voices too."} ]

Monologue Mode

Generate a speech from a single speaker:

json [ "Welcome to my presentation.", "Today we'll explore the future of AI speech synthesis.", "Let's begin with the fundamentals." ]

🐳 Docker Support

Using Pre-built Image

Pull and run the pre-built Docker image from GitHub Container Registry:

```bash

Pull the image

docker pull ghcr.io/comput3ai/c3-csm-gradio:latest

Run the container with your Hugging Face token

docker run -p 7860:7860 --gpus all -e HFTOKEN=yourhuggingface_token ghcr.io/comput3ai/c3-csm-gradio ```

Building Locally

Build and run the application using Docker:

```bash

Build the image

docker build -t csm-gradio .

Run the container with your Hugging Face token

docker run -p 7860:7860 --gpus all -e HFTOKEN=yourhuggingface_token csm-gradio ```

About HF_TOKEN

The HF_TOKEN environment variable is required for the container to authenticate with Hugging Face Hub and download the model files. You can obtain this token from your Hugging Face account settings.

⚙️ Advanced Configuration

• Temperature: Controls randomness (0.1-2.0, default: 0.9)
• Top-k: Limits token selection (1-100, default: 50)
• Max Audio Length: Maximum duration per utterance (1000-30000ms)
• Pause Duration: Silence between utterances (0-1000ms)

🔍 Implementation Details

This application is built on:

• CSM-1B Model: Sesame's Conversational Speech Model
• Llama-3.2-1B: For text processing
• Mimi: For audio codec operations
• Gradio: For the web interface

⚠️ Ethical Use Guidelines

This tool is provided for research, education, and legitimate creative purposes. Please:

• Do not use for impersonation without explicit consent
• Do not create misleading or deceptive content
• Follow all applicable laws and ethical guidelines regarding synthetic media

📄 License

The Gradio interface is licensed under the Apache 2.0 License. The CSM-1B model has its own license terms available at Hugging Face.

🙏 Acknowledgements

• Sesame AI Labs for creating and open-sourcing CSM-1B
• Hugging Face for hosting the model
• Gradio for the web interface framework