FrogBase transforms multimedia content into navigable knowledge graphs, targeting both developers and non-technical users. It simplifies the workflow of downloading, transcribing, embedding, and indexing audio-visual data, enabling efficient searching and knowledge discovery.

How It Works

FrogBase orchestrates a pipeline leveraging yt_dlp for media downloads, OpenAI's Whisper for speech-to-text transcription, and Sentence Transformers for embedding text segments. These embeddings are then indexed using hnswlib for efficient similarity search. This integrated approach streamlines the process of creating searchable knowledge bases from diverse online media.

Quick Start & Requirements

• Install ffmpeg (e.g., sudo apt install ffmpeg).
• Install FrogBase: pip install frogbase.
• For UI: pip install streamlit and run streamlit run ui/01_🏠_Home.py.
• Requires Python.

Highlighted Details

• Supports downloading from YouTube, TikTok, Vimeo, and more via yt_dlp.
• Transcribes audio using OpenAI's Whisper.
• Embeds text segments using Sentence Transformers.
• Indexes and searches embeddings with hnswlib.
• Includes a Streamlit UI for a GUI experience.

Maintenance & Community

• Project previously known as whisper-ui.
• Links to Documentation (WIP), Issues & Discussions, and Discord are provided.

Licensing & Compatibility

• License: MIT.
• Permissive license suitable for commercial use and integration into closed-source projects.

Limitations & Caveats

This repository is a pre-release version and is known to be very unstable. Stable releases are available in 1.x versions.