Videogrep is a command line tool that searches through dialog in video or audio files and makes supercuts based on what it finds. It will recognize .srt or .vtt subtitle tracks, or transcriptions that can be generated with vosk, pocketsphinx, and other tools.
- The Meta Experience
- All the instances of the phrase "time" in the movie "In Time"
- All the one to two second silences in "Total Recall"
- A former press secretary telling us what he can tell us
See my blog for a short tutorial on videogrep and yt-dlp, and part 2, on videogrep and natural language processing.
Videogrep is compatible with Python versions 3.6 to 3.10.
To install:
pip install videogrep
If you want to transcribe video or audio, you also need to install vosk:
pip install vosk
Note: the previous version of videogrep supported pocketsphinx for speech-to-text. Vosk seems much better so I've added support for it and will likely be phasing out support for pocketsphinx.
The most basic use:
videogrep --input path/to/video.mp4 --search 'search phrase'
It works with audio too:
videogrep --input path/to/audio.mp3 --search 'search phrase'
You can put any regular expression in the search phrase.
NOTE: videogrep requires a matching subtitle track with each video you want to use. The video/audio file and subtitle file need to have the exact same name, up to the extension. For example, my_movie.mp4 and my_movie.srt will work, and my_movie.mp4 and my_movie_subtitle.srt will not work.
Videogrep will search for matching srt and vtt subtitles, as well as json transcript files that can be generated with the --transcribe argument.
File or files to use as input. Most video or audio formats should work. If you mix audio and video input files, the resulting output will only be audio.
Name of the file to generate. By default this is supercut.mp4. Any standard video or audio extension will also work. (If you're using audio input or mixed audio and video input and you keep the default supercut.mp4 as the output filename, videogrep will automatically change the output to supercut.mp3)
Videogrep will also recognize the following extensions for saving files:
.mpv.edl: generates an edl file playable by mpv (useful for previews).m3u: media playlist.xml: Final Cut Pro timeline, compatible with Adobe Premiere and Davinci Resolve
videogrep --input path/to/video --search 'search phrase' --output coolvid.mp4
Search term, as a regular expression. You can add as many of these as you want. For example:
videogrep --input path/to/video --search 'search phrase' --search 'another search' --search 'a third search' --output coolvid.mp4
Type of search you want to perform. There are two options:
sentence: (default): Generates clips containing the full sentences of your search query.fragment: Generates clips containing the exact word or phrase of your search query.
Both options take regular expressions. You may only use the fragment search if your transcript has word-level timestamps, which will be the case for youtube .vtt files, or if you generated a transcript using Videogrep itself.
videogrep --input path/to/video --search 'experience' --search-type fragment
Maximum number of clips to use for the supercut.
Show the search results without making the supercut.
Randomize the order of the clips.
Padding in seconds to add to the start and end of each clip.
Time in seconds to shift the shift the subtitles forwards or backwards.
Transcribe the video/audio using vosk. This will generate a .json file in the same folder as the video. By default this uses vosk's small english model.
NOTE: Because of some compatibility issues, vosk must be installed separately with pip install vosk.
videogrep -i vid.mp4 --transcribe
In combination with the --transcribe option, allows you to specify the path to a vosk model folder to use. Vosk models are available here in a variety of languages.
videogrep -i vid.mp4 --transcribe --model path/to/model/
Exports clips as individual files rather than as a supercut.
videogrep -i vid.mp4 --search 'whatever' --export-clips
Exports the transcript of the supercut as a WebVTT file next to the video.
videogrep -i vid.mp4 --search 'whatever' --export-vtt
Shows common words and phrases from the video or audio file.
videogrep -i vid.mp4 --ngrams 1
from videogrep import videogrep
videogrep('path/to/your/files','output_file_name.mp4', 'search_term', 'search_type')
The videogrep module accepts the same parameters as the command line script. To see the usage check out the source.
Also see the examples folder for:
- silence extraction
- automatically creating supercuts
- creating supercuts based on youtube searches
- creating supercuts from specific parts of speech
- creating supercuts from spacy pattern matching
Videogrep is maintained by Sam Lavigne, and built using MoviePy and Vosk. A big thanks goes out to all those who have contributed, particuarly to Charlie Macquarie for his efforts in getting the project to work with audio-only media.
Videogrep has received financial support from the Department of Digital Humanities, King’s College London and from the Clinic for Open Source Arts.