Multicam Podcast Editor

• Multicam Podcast Editor
  • Dependencies
  • installation
  • Usage
  • Compatibility

If you use an app like streamyards, they provide the main podcast video, and then separate individual video files for each person attending your streamyard session.

I've created a tool to automatically do the following things:

1. Synchronize the videos to the main multipeople video. The individuals' files are always slightly off in their timing.
2. Chunk the file videos into 5 second segments and use the individuals' video if they are mostly the only one talking for those 5 seconds.
3. Use jump cutting to remove dead air
4. Transcribe the video using the individuals' videos
5. The ability to ask an LLM about your podcast/video transcription
6. Create shorts given a time range

Dependencies

• ffmpeg
• https://github.com/WyattBlue/auto-editor
• For chat, install ollama

installation

1. Install ffmpeg and ensure it's in your path
2. Install auto-editor as per their github page or by doing pip install auto-editor
3. run the following

git clone https://github.com/horvay/multicam-podcast-editor
cd multicam-podcast-editor
pip install -r requirements.txt

• note, you may want to make a conda environment first

Usage

This tool assumes you have 2 or more files: a video with all the audio and video from all the participants, and individual videos for each participant. You will need to follow this naming exactly:

• call the video with all the cameras and audios "main.mp4"
• call each of the individuals' videos "person1.mp4", "person2.mp4", etc, or have the word "webcam" in the name. (This makes it easy for streamyard users so they can just copy and paste the files as is)
• if you want to include screenshares, then copy those over and ensure they have the word "screen" in them.

Ensure these videos are in the inputfiles directory.

Run the following for multicam editing

python main.py --multicam

If you want to enhance the audio quality of the output for shorts or multicam, you can add -au.

or for transcribing

python main.py --transcribe

run the following to see all options

python main.py --help

For instance, if you wanted to do automatic multicam editing including jump cuts to remove deadair, transcribe the videos into a text file, and then ask for youtube titles about your videos, you could do the following:

python main.py --multicam --jump-cuts --transcribe --question "what are some good youtube titles for this podcast?"

This would give you the final edited video outputted to the directory, a transcript.txt file, and it would output the console the response to your question, using an ollama LLM (llama3.2 by default)

The following will create a short from the videos in the directory, starting at second 1705 until second 1812, it'll remove the deadair and will cut out seconds 50.1-67.0, 88.5-96.0, etc.

python main.py -s 1705 -ti 1812 --jump-cuts -c 50.1 67.0 -c 88.5 96.0 -c 9.5 103.0 

It will take quite some time to run but should give updates on it's progress.

Compatibility

• I've only run this on linux but it should work for other environments since it's python