A topic-based video editor
There are a lot of videos on the internet of complications of people talking on similar topics. See some examples below:
Motivational Video: https://www.youtube.com/watch?v=19sWsI7skpE
(The audio is being sliced together based on similar topics)
Birthday Videos: https://youtu.be/Za_buG87vGQ?t=344
(Clip ends at 7:27, but she articulates the point beautifully)
I think it would be useful to have a video editor that goes like this
- Add video
- Transcribe audio
- Convert groups of sentences into topics using NLP
- You can click on a topic and it will crop the rest of the video away from the main timeline
- Repeat the process for multiple videos.
I think it would save a large amount of time to achieve the same goal as what the editor is doing.
However, this idea is probably only fit for a very small business niche. More of a 1-time payment model than a saas. Even though birthdays are every day, a specific user's friend's birthday won't be every day.
Video editing unless done professionally is usually a 1 off.
Hey @adroitboss,
there are some video editors adjacent to this idea i.e. https://firstedit.io or https://typestudio.co or https://qlip.ai or https://scribe.ai check them out if you like!
I've also been working on something which is not an editor per-say but it lets you mark, and reuse bits from existing online audio/video content based on your reactions. It's called S-K-I-P https://skiptothegoodbit.com
Check an embeddable example here: https://sttgb.it/e/demo/jHi1vjjiR6iWfw3eDv61?onlythegoodbits
Cool demo! Do you think the reactions will happen at a point that would give the user enough context? I created a note-taking app that used Google transcriptions to take notes on youtube a while back. You would press a button to capture the last 30 seconds of the user's notes.
The problem became context. How do I know that the subject was entirely encapsulated within the last 30 seconds? Well, sometimes it wasn't. Ultimately what I was going to do was grab the title of the video, translate it into a question format, and then use a Q and A Machine learning model to answer the questions based on the transcriptions.
Then you just grab all of the timestamps associated with those transcriptions and those would be the "good bits". I wanted the extension to jump only to the parts of the video the user intended on watching when they clicked the title, and the title happens to be a really good indicator for that. Well, that and the thumbnail.