This tool was inspired by watching Catan gameplay on YouTube. Most content creators spend lots of time editing their videos, however smaller or newer creators often just upload the raw footage, leaving in all the downtime of essentially watching the clock tick down.
The Auto Video Editor tool extracts audio from an input video file and sets up an audio detection mechanism to highlight areas where there's vocal activity. The following image shows an example of automatically detected audio events (note each event is padded by 0.60 seconds):
After identifying these segments, the tool provides a configurable padding buffer to each segment in order to retain some continuity of context between cuts. By splicing these segments together, the tool automatically removes areas without any audible activity.
The following video shows an example of the tool's capabilities, using a video taken from BuddyCatan's channel:
The original video duration was approximately 49 minutes, and the tool condensed this to a more concise 32 minutes. In essence, this eliminated around 17 minutes of inactive (or uninteresting) content, roughly a third of the original video!
This tool relies on ffmpeg (https://www.ffmpeg.org/), and has only been tested on Windows 10.
This website provides step-by-step instructions on how to install ffmpeg on windows: https://www.geeksforgeeks.org/how-to-install-ffmpeg-on-windows/
To install this tool, you will need to do the following:
- Create a Python 3.10 (or greater) environment (typically using Conda)
- Download the content into a project folder
- Activate your python environment and install the required dependencies with the following command:
pip install -r requirements.txt
This tool has some fairly basic functionality, however I have plans to make additonal features.
To use the tool with all the default options, run the following command from your python project directory:
python auto-video-edit.py /path/to/input_video.mp4 /path/to/output_video.mp4
The output file does not have to exist; the tool will generate the video. To modify how the tool operates, use the following options:
| Option | Description |
|---|---|
| -v, --version | Print the current version of the tool. |
| -p, --padding [VALUE] | Amount of time of a detected audio event (on both sides) to pad for context. (default = 0.60 s) |
| -d, --decimate [VALUE] | Integer value to downsample audio. If left undefined, the program will attempt to automatically determine a suitable decimation value based on the end frequency. |
| -t, --threshold [VALUE] | Value (in dB) the signal must be above the noise. (default = 5 dB) |
| -n, --nice-audio | Flag to indicate if you'd like the tool to filter the output audio in the 150-10000 Hz range (this is useful if the video was recorded with lower quality microphones) |
| --example-image | Flag to indicate if you'd like the tool to generate an example image of the audio and at least one detected event. |
| -sf, --start-freq [VALUE] | The start frequency of audio events of interest. This is typically the narrator's vocal range (default = 200 Hz) |
| -ef, --end-freq [VALUE] | The end frequency of audio events of interest. This is typically the narrator's vocal range (default = 3000 Hz) |
| -bg, --band-gap [VALUE] | Buffer around vocal range to find out-of-band energy. (default = 50 Hz) |
| -li, --lead-in [VALUE] | Amount of time (in seconds) to ensure the tool does not cut in the beginning of the video. (default=None) |
| -lo, --lead-out [VALUE] | Amount of time (in seconds) to ensure the tool does not cut at the end of the video. (default=None) |
| -c, --config [VALUE] | Unused parameter in V0.0.X This will be used in the future to define configuration parameters for additional features |