How to pull out only the parts of a long video about one topic
Upload the long recording to Vizard Agent and describe the topic you want kept. Vizard Agent transcribes the whole thing, filters the transcript for segments about that subject, reads the lines around each match to find where the passage really begins and ends, then joins them with crossfades into one video.
What do you type into Vizard Agent?
Describe the theme in plain sentences. Vizard Agent reads the transcript for meaning, so a paragraph explaining what you want kept will find more of the right material than a list of terms, and it will not miss passages that make your point without ever using your words.
Prompt
What does Vizard Agent actually do?
It searches the transcript for meaning, then reads the lines on either side of each match, because a passage almost never starts exactly where the keyword lands. That is the difference between a passage that makes sense and one that begins mid-sentence, and it is what separates this from a keyword search. In the session behind this article it:
- Analysed the video's format and streams, and checked how the transcription tool worked and what it cost before running it.
- Transcribed the full forty-one minutes and inspected the transcript's structure on a sample segment.
- Filtered the segments about the requested theme and read the surrounding segments at every hit: seven separate passages, each examined in context.
- Cut all seven passages concurrently rather than one after another.
- Calculated the crossfade offsets, merged the pieces and extracted frames at each transition to check the joins looked smooth.
- Delivered a 1920x1080, 30fps file of 425.97 seconds: seven minutes out of forty-one, in the recording's own shape.
When does this not work well?
The transcript is the whole search index, so anything the speaker did not say out loud is invisible to this method. Vizard Agent reads meaning and context, and it still cannot find a moment that exists only in the picture. A topic that runs throughout has no edges, and poor audio gives a rough transcript and a rough selection.
Common questions
Does it search for exact words? No. Vizard Agent reads the transcript for meaning, so it finds passages that make your point in different words.
Will the cuts sound abrupt? No. Vizard Agent reads around each match for natural boundaries and joins the pieces with crossfades.
Can I get the passages separately? Yes. Say so and Vizard Agent delivers each one as its own file.