Two AI entities talking to each other — in real time, with premium neural voices
AI Dialogue is a sophisticated conversation simulator where you configure two AI entities with simple prompts, hit start, and watch them engage in natural, human-like dialogue. Perfect for exploring AI behavior, testing conversation scenarios, or just enjoying entertaining AI interactions.
- 28+ Azure Neural Voices including Brian, Ava, Emma, Ryan and more
- Multilingual support with natural-sounding speech
- Real-time audio streaming with minimal latency
- Speed control from 0.5x to 1.5x playback speed
- Circular audio waveforms that dance to the actual audio signal
- Clean, responsive design that works on desktop and mobile
- Real-time character counting with live validation
- Collapsible advanced settings for power users
- Human-like conversation instructions built-in automatically
- Temperature and Top-P controls for fine-tuning AI behavior
- Response length options: Short (35 words), Medium (60 words), Long (90 words)
- Character limits: Up to 375 characters per entity prompt
- Sample conversation templates: Riddle challenges, political debates, joke rating
- Automatic conversation limits: Max 20 interactions (10 per entity)
- Real-time WebSocket updates for smooth interaction
- Comprehensive cleanup between conversations prevents overlaps
- GDPR compliant with proper legal links
- Azure EU data residency for European users
- No data training on your conversations
- Secure WebSocket connections
- Backend: FastAPI + WebSockets + Azure integrations
- Frontend: Pure HTML5/CSS3/JavaScript (no build step needed!)
- Text-to-Speech: Azure Cognitive Services Speech SDK
- LLM: Azure OpenAI GPT-4o mini
- Audio Visualization: Web Audio API + Canvas
- Real-time Updates: WebSocket streaming
- Python ≥ 3.9
- Azure OpenAI account with GPT-4o mini deployment
- Azure Speech Services account
- No Node.js required! ✨
git clone https://github.com/Jamroszczyk/AI-Dialogue.git
cd AI-Dialogue
# Create virtual environment
python -m venv venv
source venv/bin/activate # Windows: venv\Scripts\activate
# Install dependencies
pip install -r requirements.txtCreate .env file in the project root:
# Azure OpenAI (required)
AZURE_OPENAI_KEY_DE_4_1=your_openai_api_key_here
AZURE_OPENAI_ENDPOINT_DE_4_1=https://your-resource.openai.azure.com/
# Azure Speech Services (required)
SPEECHKEY=your_speech_service_key_here
SPEECHENDPOINT=https://your-region.api.cognitive.microsoft.compython app.pyVisit http://localhost:8000 and start your first AI dialogue! 🎉
- Configure AI Entities: Enter simple prompts like "You love discussing space exploration" or "You're skeptical about new technology"
- Choose Voices: Select from 28+ premium Azure neural voices (Christopher and Cora are defaults)
- Pick a Template (Optional): Try pre-built scenarios like riddle challenges or political debates
- Fine-tune Settings: Adjust temperature, top-p, response length, and speech speed
- Start Conversation: Hit the play button and watch your AIs come to life!
AI-Dialogue/
├── app.py # FastAPI web server & WebSocket handler
├── azure_tts_helper.py # Azure Speech Services integration
├── transformers.py # Azure OpenAI API calls
├── index.html # Main frontend interface
├── style.css # Modern UI styling
├── script.js # Frontend logic & audio handling
├── requirements.txt # Python dependencies
├── voices/ # Generated audio files (auto-cleaned)
└── README.md # This file
The app includes three built-in conversation templates:
- 🧩 Riddle Challenge: AIs take turns creating and solving riddles
- 🗳️ Political Debate: Respectful discussion between conservative and progressive viewpoints
- 😂 Joke Rating Time: One AI tells jokes while the other provides ratings and feedback
- System Prompt: Up to 375 characters defining the AI's personality
- Voice Selection: Choose from 28+ Azure neural voices
- Speed Control: 0.5x to 1.5x playback speed
- Temperature: 0.0-2.0 (creativity vs consistency)
- Top P: 0.0-1.0 (response diversity)
- Response Length: Short/Medium/Long word limits
- Character Limit: 375 characters per prompt (with live counter)
- Interaction Limit: Maximum 20 exchanges per conversation
- Auto-cleanup: Prevents conversation overlap issues
fastapi==0.116.1 uvicorn==0.35.0 python-dotenv==1.1.1 pydantic==2.11.7 openai==1.99.9 azure-cognitiveservices-speech==1.45.0
See requirements.txt for the complete list with all pinned versions.
- Create an Azure OpenAI resource in your preferred region
- Deploy a GPT-4o mini model
- Copy the endpoint and API key to your
.envfile
- Create a Speech Services resource
- Choose a region (preferably same as OpenAI for lowest latency)
- Copy the service key and endpoint to your
.envfile
💡 Pro tip: Both services offer free tiers perfect for testing!
Male Voices: Adam, Alloy, Andrew, Brandon, Brian, Christopher, Davis, Derek, Dustin, Echo, Lewis, Onyx, Ryan, Samuel, Steffan
Female Voices: Amanda, Ava, Cora, Emma, Evelyn, Jenny, Lola, Nancy, Nova, Phoebe, Serena, Shimmer
Neutral Voice: Fable
All voices support multiple languages and natural prosody patterns.
- Secure Connections: All API calls use encrypted connections
- Auto-cleanup: Audio files are automatically deleted after playback
"Connection Error"
- Check your Azure credentials in
.env - Verify your Azure resources are deployed and active
- Ensure you have sufficient quota/credits
"Audio Not Playing"
- Check browser audio permissions
- Try refreshing the page to reset audio context
- Verify Speech Services key and endpoint
"WebSocket Disconnected"
- Restart the backend server
- Check for firewall/proxy issues
- Clear browser cache and reload
- Check the browser console for detailed error messages
- Verify all environment variables are set correctly
- Ensure Python dependencies are installed in the virtual environment
Contributions are welcome! Please feel free to submit a Pull Request. For major changes, please open an issue first to discuss what you would like to change.
MIT License - feel free to use this project for personal or commercial purposes. Just give credit where it's due!
See LICENSE for full details.
This project uses Azure SDKs to interact with:
-Azure OpenAI for language model services
-Azure Speech Services for text-to-speech synthesis
While the SDKs are open-source (MIT licensed), access to Azure services requires:
- Active Azure subscription
- Compliance with Microsoft's Terms of Service
- Appropriate API keys and endpoints
Users are responsible for their own Azure usage costs and compliance.
- Azure Cognitive Services for premium neural voices
- FastAPI for the excellent async web framework
- Azure OpenAI for powerful language models
- Web Audio API for real-time audio visualization
Privacy: Privacy Policy
Legal: Imprint
Made with ❤️ by Jamroszczyk
AI Dialogue - Where artificial minds meet and magic happens ✨