Taler - Real-time Presentation Speech Tutor

Taler is an AI-powered speech coaching application designed to help users improve their public speaking skills. It provides real-time feedback on pacing, filler words, and overall delivery, acting as a personal speech tutor.

Features

• Real-time Speech Analysis: Analyzes your speech as you talk.
• Pace Monitoring: Calculates Words Per Minute (WPM) to ensure you are speaking at an optimal speed (120-150 WPM).
• Filler Word Detection: Identifies and counts common filler words (e.g., "um", "uh", "like", "you know").
• AI Coaching Persona: Receive feedback from "Gandalf the Grey", offering wisdom and guidance on your speech in a unique and engaging way.
• Transcription: Converts your speech to text using advanced speech-to-text models.
• ElevenLabs Integration: Leverages ElevenLabs for high-quality transcription and agent interactions.
• Local Processing: Utilizes local models (MLX Whisper, Qwen) for privacy and low latency (in the local agent version).

Project Structure

• local_agent/: Contains the local version of the speech tutor using MLX Whisper and a local LLM.
  • local_agent.py: The main script for the local real-time speech coach.
• eleven_test.py: A script demonstrating integration with the ElevenLabs API for transcription and agent feedback.
• voice_samples/: Directory for storing audio samples for testing.

Getting Started

Prerequisites

• Python 3.12+
• uv (recommended for dependency management) or pip
• ElevenLabs API Key (for eleven_test.py)
• Mac with Apple Silicon (for local_agent using MLX)

Installation

1. Clone the repository:

   git clone <repository-url>
   cd taler

2. Install dependencies:

   uv sync
   # or
   pip install -r requirements.txt

Usage

Local Agent (Gandalf)

Run the local real-time coach:

python local_agent/local_agent.py

Speak into your microphone, and the agent will analyze your speech and provide audio feedback.

ElevenLabs Integration Test

To test the ElevenLabs integration:

1. Set your API key:

   export ELEVENLABS_API_KEY="your_api_key_here"

2. Run the test script:

   python eleven_test.py

Technologies

• Python: Core programming language.
• ElevenLabs API: For cloud-based speech-to-text and conversational agents.
• MLX / MLX Whisper: For efficient local machine learning on Apple Silicon.
• SoundDevice: For audio recording and playback.
• Mutagen: For audio file metadata analysis.

License

[Add License Information Here]