Runs Whisper over a directory of media files. For each mp4 and mp3 file, it generates a text transcript and a WebVTT caption file with FADGI metadata, using an optional CSV for Title and Media Identifier.
The CSV file should be set up with these headers: "File", "Title", "Media Identifier". Include the names of each reference file in the "File" column. (For best results, save the CSV as utf-8 encoded.) For example:
| File | Title | Media Identifier |
|---|---|---|
| filename.mp4 | media_title | uniqueID |
Script is customized for Emory Libraries; FADGI metadata is generated from the script or matched from the CSV as follows:
| Element | Source |
|---|---|
| Type | hard-coded |
| Language | auto-detected by Whisper or supplied by user |
| Responsible Party | hard-coded |
| Media Identifier | matched from CSV |
| Originating File | generated from media file |
| File Creator | hard-coded |
| File Creation Date | generated from script |
| Title | matched from CSV |
| Origin History | hard-coded |
| Local Usage Element: Software version | generated from script |
| Local Usage Element: Review history | hard-coded |