Steady Improvements in Generative AI
Generative video continues to advance quickly. It seems like every week there’s another startup that releases a tool that allows you to prompt a video. The quality is also improving.
Many months ago, I tried one of the tools on a photo of a mural from one of the kid’s bedrooms. It created a movie that was a horrific nightmare of barnyard animals losing limbs and becoming disfigured. Fast forward to today and tools like hailuo AI are able to to get some context from animal limbs and mouths and put together something coherent. I wasn’t more scared than looking at the original input video.
Within six months, some of these tools will be able to extract the characters and create stories. It’s going to be wild.
Same with audio generation…
I had played around with Suno.AI many months ago and was astounded with the quality. Today, it’s taken a leap forward and we’re still on the cusp of the next versions. I can happily listen to a four-minute song without my ears bleeding.
What’s amazing with encountering these new tools is the knowledge that they’re only getting better — and quickly. For mass adoption of these advancing generative tools, the workflow needs to be made much easier:
- YouTube could automatically recut content as shorts
- Music could be generated to match the video
- Title cards, including meme-based ones, e.g. “two minutes later…”, and their content could be generated on the fly
These can also be part of posts on Facebook, Instagram, X, etc. We will benefit from hearing something no one else has ever heard or seen before.
