smol notebook lm

Convert text to speech using the ElevenLabs client. Ensure rawtext.md is updated, all environment keys are loaded, and outputs are available in .mp3, .txt, and .log formats.

You can see sample output here: https://github.com/smol-ai/temp

requirements

• make sure ffmpeg is installed
• brew install imagemagick

Setup

1. Install Dependencies:

   pip install -r requirements.txt

2. Set Environment Variables:

   export ELEVENLABS_API_KEY=your_api_key
   export OPENAI_API_KEY=your_api_key
   export CARTESIA_API_KEY=your_api_key # note that this thing generates a lot of tokens. we used up 52k cahracters just developing this.

3. Update rawtext.md:

   • Add or modify the text you want to convert to speech.

4. Run the Script:

   python main.py

Optional Video Generation

To generate a video from the audio and transcript, you can use the video.py script. This script uses the MoviePy library to combine the audio and image, and the OpenAI library to generate a default image using DALL·E.

Requirements

• Make sure you have the moviepy and openai libraries installed. You can install them using pip:

  pip install moviepy openai

Setup

1. Set Environment Variables:

   export OPENAI_API_KEY=your_api_key

2. Run the Script:

   python video.py

Outputs

• Video: final_video.mp4 file

Notes

• The video.py script assumes that the combined_dialogue.mp3 and dialogue_transcript.txt files are present in the same directory.
• The script generates a default image using DALL·E and resizes it to 1080x1080 pixels.
• The script combines the audio and image to create a video, and adds captions using the transcript.
• The final video is saved as final_video.mp4 in the same directory.

Outputs

• Audio: .mp3 files
• Transcript: .txt files
• Logs: .log files

Examples

• Sample MP3
• Transcript