Finding highlights in long-form video content automatically - Sieve Blog
Today, we’re announcing the beta release of Highlights — an app built to find key moments in long-form video content like podcasts, news shows, movies, and more using natural language search.
Creators currently use products like Descript, Grabyo, or Kick to stream and edit long-form content. However, much of their editing time is still spent on manually splitting their raw media into highlights or other bite-sized snippets.
Finding these highlights automatically with machine learning is quite challenging. Highlights can be very different between different types of content. In sports, a highlight might be defined as an action (like a goal or play) while in a podcast or stream, highlights relate more to the transcript. Building a good solution for highlight generation must include some flexibility on what type of content to filter for.
Our app allows users to search in natural language for “highlight terms” like “most interesting” or “most action” and receive a set of matching clips. Try it out on your own content here or integrate via API.
