X SpaceCadet

the first & only x space transcription tool that is able to identify speakers and summarize conversations

https://github.com/robbie-wasabi/xspacecadet

Features

Record, transcribe (with speaker identification), and chat directly with X Space transcripts.

• Record X Spaces: Bot joins the space and records the audio while noting the current speaker.
• Audio Processing: Download and process audio using twspace-dl.
• Transcription: Transcribe audio using Insanely Fast Whisper.
• Speaker Identification: Identify speakers in the transcript using captured frames and metadata.
• Chat with Transcript: OpenAI integration to chat with the transcript.

Work in Progress

Listening to people ramble on x spaces is often grueling but it is certain that there are diamonds in the rough. For this reason, I searched for tools that recorded spaces but couldn't find anything that was able to actually identify speakers and summarize conversations - so I built a solution myself. I've already dumped too much time into this and it suits my needs but I figured I'd share it in case anyone else finds it useful. Depending on how much interest there is, I'll continue to improve it.

1. Recording Termination: Currently, the only way to stop a recording is to kill the application. We're working on implementing a more graceful shutdown method.

2. Live Spaces Only: Only works for spaces in progress because the bot needs to join the space and identify speakers in real-time. Would love for this to work with completed/recorded spaces but might be tricky since the speaking animations used to identify speakers aren't as reliable.

3. User Experience: The current UX could be significantly improved. Streamlit was chosen to keep development time to a minimum.

4. Code Quality: The codebase is pretty shitty but its in an okay place to begin cleaning it up.

Installation

Prerequisites

• Python 3.7 but not higher than Python 3.11.10: insanely fast whisper has issues with newer versions of python.
• Chrome WebDriver: Required for Selenium to automate browser actions.
• ffmpeg: Needed for audio conversion.
• X API Key: Basic plan required for accessing X API endpoints.
• Hugging Face API Key: Required for accessing Hugging Face API endpoints.
• OpenAI API Key: Needed for chatbot.

Setup Steps

1. Clone the Repository:

   git clone https://github.com/robbie-wasabi/xspacecadet.git
   cd xspacecadet

2. Install Dependencies:

   # install python 3.11
   brew install [email protected]
   
   # create virtual environment
   python3.11 -m venv .venv
   source .venv/bin/activate
   
   # install dependencies to virtual environment
   python3.11 -m pip install -r requirements.txt

3. Set Up Environment Variables:

   Create a .env file in the root directory and add the following variables:

   • X_BEARER: Your X API bearer token.
   • X_COOKIE_FILE: Path to your X cookie file.
   • HF_TOKEN: Your Hugging Face API token.
   • OPENAI_API_KEY: Your OpenAI API key.

   Example .env file:

   X_BEARER=your_x_bearer_token
   X_COOKIE_FILE=./cookies.txt
   HF_TOKEN=your_hugging_face_token
   OPENAI_API_KEY=your_openai_api_key

   Note: You can obtain these tokens from:

   • X API Bearer Token: X Developer Portal
   • X Cookie File: Use the Get cookies.txt LOCALLY Chrome extension.
   • Hugging Face Token: Hugging Face Tokens
   • OpenAI API Key: OpenAI API Keys

Usage

XSpaceCadet can be used via a shitty command-line interface or the Streamlit web application.

Command-Line Interface (CLI)

Recording a Space

To record a Space:

python3.11 main.py record <space_id> [cookie_file] [options]

Example:

python3.11 main.py record https://x.com/i/spaces/AAAAAAAAAAAAA ./cookies.txt

Transcribing and Identifying Speakers

To transcribe the recorded audio and identify speakers:

python3.11 main.py transcribe <space_id>

Example:

python3.11 main.py transcribe AAAAAAAAAAAAA

Fetching Space Metadata

To fetch metadata for a space:

python3.11 main.py fetch-metadata <space_id>

Example:

python3.11 main.py fetch-metadata AAAAAAAAAAAAA

Streamlit Web App

Running the App

Start the Streamlit application:

python3.11 -m streamlit run app.py

Using the App

1. Open your web browser and navigate to http://localhost:8501.

2. Configure your tokens and settings in the sidebar:

   • Hugging Face Token
   • X Bearer Token
   • Path to X Cookie File
   • OpenAI API Key

3. Use the Record Space tab to start recording an X Space by entering the Space ID or URL.

4. Use the Transcribe tab to process recorded spaces, generate transcripts, and summaries.

Contributing

Contributions are encouraged (please) just open an issue or submit a pr.

License

project is licensed under the MIT License.

Contact

discord: robbie_wasabi