Most AI tools at the moment are over-hyped which is why I try to take a pragmatic approach. I'm most interested in ones that can augment and shorten mundane workflows – like slicing and dicing content, which is something we have to do a lot in developer relations.
I gave Opus.pro a test run this week, which is an AI tool that can cut long-form videos into 15 short-form clips "instantly" (it takes a few minutes). It's also free (keep in mind you're sending your video data / transcript to OpenAI, I'm positive).
I fed it my 50-minute money talk which is something I would love to slice and dice. Here's how it went.
The Good
- It's fast – relative to combing through a video manually for sections to extract.
- It picked out parts I agree could be good for short-form content.
- I also like how it overlays the transcription and even picks out emojis.
- Once you create an account, you can edit the clips.
The Bad
It is designed for social apps like Instagram and YT shorts. That means it doesn't really work when your source video is a presentation with slides and text, or not in the vertical portrait format.
You can tell it to "fit" the video and then you'll get that effect where it's a small box with the background blurred.
Here's a clip it produced (notice how it cut the audio too at the beginning 😑):
The Ugly
The headlines it generates are TOTAL clickbait and I wouldn't use them 🤣 (The sad truth is that these would likely work).
Verdict
There's potential but it's not there yet – especially with regards to technical content and videos like demos or conference talks. Out of the 15 options, there wasn't one that could "just work" out of the box.
So no, the promise of "one click is all you need" and seeing a 10X productivity increase wasn't fulfilled.
This is the biggest problem with AI tool marketing right now: they are high on their own farts.
Remember: marketing is making a promise to your audience – if you can't keep it, you lose trust. Opus could improve its marketing (and retention) by simply being honest about the style and format of the content it works best in and not over-hyping it. As a dev tool, if you had content like podcast interviews, panels, or other talking heads-style video content, you'd probably have much better luck.
That said, this could be useful to pick out what sections might resonate the best and then you'll need to manually create the visuals.
This is still early days and most stuff is just going to fall way short of expectations BUT to be fair, this is how it starts. We need these little experiments to see what use cases people want and then they can be enhanced.
Like with this tool, an evolved version (or a different, more specialized tool) could be tuned for conference talks or technical demos. Imagine it understanding code snippets and reformatting them. This is the kind of innovation I would expect to see in the next year or two.
Have a lovely day,
Kamran
