meetscribe 0.1.4

An intelligent audio-to-transcript chatbot powered by Whisper, PyAnnote, FAISS, and LLMs.

MeetScribe

MeetScribe is an intelligent AI-powered tool that converts spoken meeting audio into accurate, speaker-labeled transcripts and allows users to chat with those transcripts using powerful LLMs like GPT-4o.

Powered by OpenAI Whisper, PyAnnote for speaker diarization, and LangChain RAG pipelines.

---

Features

• 🎙️ Speaker Diarization using PyAnnote
• ✍️ Transcription using OpenAI's Whisper
• 🧠 Question Answering using GPT-4o (or any OpenAI model)
• 🔍 Customizable Embeddings – Use HuggingFace or OpenAI Embeddings
• 📜 Chat History – Multi-turn conversation memory
• 🖥️ Streamlit UI – Easy-to-use local or web-based interface

---

Quick Start

1. Clone & Install (for Streamlit)

git clone https://github.com/jagadale1234/meetscribe.git
cd meetscribe
pip install -r requirements.txt

For CLI

pip install meetscribe

Requirements

• Hugging Face token (for PyAnnote)

• OpenAI API key (for embeddings + LLM)

Usage (CLI)

meetscribe \
  --audio path/to/meeting.wav \
  --openai-key sk-xxx \
  --hf-token hf_yyy

Usage (scripts and notebooks)

from meetscribe import process_audio, ask_question

transcript, qa_chain = process_audio(
    wav_path="meeting.wav",
    openai_key="sk-xxx",
    hf_token="hf-yyy",
    embedding_model="OpenAI"
)

print(transcript)

answer = ask_question(qa_chain, "What was the deadline for the new design?")
print("Answer:", answer)

Usage (Streamlit)

Check Github (https://github.com/jagadale1234/meetscribe.git)