Surely you have come across extremely long YouTube videos at some point in your life, at which point it would be nice if someone did a summary so you don’t have to watch every minute of it.
Many Youtubers infinitely lengthen the videos, since the longer it is, the more ads fit. That means that something that can be said in 2 minutes ends up being said in 15… bad thing.
Luckily, AI is here to lend a hand, as tools that summarize content into text are beginning to appear. We have already seen several that summarize texts, and today we have to talk about one capable of summarizing videos.
What summarize.tech does is analyze the text that is narrated within the video, it summarizes said text and shows us said summary every x minutes (depending on where the relevant parts are).
In this way, the summary becomes something like: the video talks about this, and at minute 5 it talks about this, at minute 8 it comments on this, and at 12 this other.
Here’s an example of how he summarized an 8-minute video of mine:
As you can see, the summary is not in video format, only in text, and in English, although it understands videos in other languages ​​well. When it comes across a script in Spanish, for example, it translates it, summarizes it, and displays the summary in English.
They present it as a good option for summarizing YouTube videos such as lectures, live events, or government meetings, into an easy-to-read summary with timestamps linking to relevant parts of the video.
The tool was built on top of OpenAI’s GPT-3, and used Gitpod, Next.js, Firebase Auth, Dagster, MySQL, and React-Bootstrap. They are now working on some additional features, such as a built-in player and neural voiced text to create short podcasts from long-form video.
There is a non-free version for $10 per month, to offset the cost of the GPT-3 API, which can be significant.
