The transcript is mined for nouns and topic words, then matched against your local footage library. Relevant B-roll clips are placed on an overlay track at the right moments. No cloud, no stock search.
The Whisper transcript is parsed for noun phrases and topic keywords. These are the content words in the dialogue that represent things the speaker is likely talking about — products, places, actions, concepts. Common stop words and filler words are excluded.
The extracted keywords are then matched against the tags and filenames of media files in your configured asset library folder. Files whose names or tags include a matching keyword are candidates for placement at the moment in the transcript where the keyword appeared.
Your asset library is any folder of video files on your local drive or network storage. Point EditLint AI at the folder in settings and it indexes the filenames and any embedded metadata tags. The index is updated whenever you run an Add B-Roll operation.
The most effective libraries have descriptive filenames that include the subject of the clip: office_team_meeting.mp4 is far more matchable than clip_003.mp4. You can also add tags to your media files using your operating system or asset management software.
When EditLint AI can't find a matching clip in your library for a keyword, it marks that keyword as unmatched in the panel. Rather than silently skipping it, it shows you the keyword and suggests it as a search term for your next stock asset purchase or as a shot list item for your next shoot.
This turns the unmatched keywords into a useful asset gap analysis. Over time, building your library around the keywords that appear most often in your work means the match rate increases with every project.
The sequence is transcribed locally using Whisper. Noun phrases and content keywords are extracted from the transcript with their timestamps. Each keyword is the anchor for a potential B-roll placement at that moment in the timeline.
Each keyword is matched against your indexed asset library. Matches are scored by similarity between the keyword and the filename/tags. Results above the confidence threshold are presented as suggestions, with the best match shown first alongside alternates.
Each suggestion appears in the panel with the matched clip, a thumbnail if available, the confidence score and the placement timecode. Approve, reject or swap individual suggestions. Click Apply and the approved clips are placed on an overlay video track at the correct positions.
Stop manually hunting through folders for the right B-roll. Let EditLint AI connect what the speaker says to what you have in your library.