(Mostly) Video + Audio Roundup
Winter is coming
Pika Labs vs. RunwayML
About six months ago, I built my first fake ‘AI generated’ beer commercial, just to see what was possible with RunwayML and have a timestamp for the state of the art. A month later, I built it again in Pika Labs and it looked a lot better. The tools have improved a lot since then, as has my understanding of how to use them. So yesterday, I rebuilt that commercial again, with both of those tools and compared the workflow and results. If you’d like to read the whole process, you can find it here.
Waveformer: Really Easy to Generate Music
If you just needed a snippet of music to launch that project you’re working on, or you just like killing time by playing around with a music generator, check out Waveformer. It’s based on Meta’s MusicGen open source model and it’s really easy to use. If you’re on the technical side, it also pairs with dev tools from Replicate.
Meta Again: Emu Video Generator
The research team at Meta is putting out some amazing stuff. I’ve posted about this one before, but I was mistaken about it being open to the public yet. However, the do have a configurable demo and you can at least play with a small sample of prompts and resulting video. If you’re curious about the future of generative video, check it out.
Pika Labs Raises 55 Million in Funding
Just like the never-ending feature race we’re seeing in the image generator space, I imagine that this means another huge leap forward. I’m probably showing my bias and rooting for the underdog, but watch out RunwayML, Pika is gonna eat your lunch.
Check it out.
On a side not, if you’re using Pika Labs in Discord, they’ve changed there icon. I spent about 5 minutes looking for it.
Update: I just read that Pika is going to make the tool web based. Get on the waitlist here: https://pika.art/waitlist
And of Course, Stable Diffusion for Video
It’s still mostly a research paper, but the quality is amazing. If you’re technically inclined, have a powerful graphics card, and can download from a github repo, then you can go check it out. Then rest of us will just have to wait a few months until someone creates a commercial version with a usable interface.
ChatGPT Frameworks to Save for Later
Okay, so technically neither audio nor video, but too good to pass up. I’m a big fan of Shelly Palmer and this set of frameworks to prompt ChatGPT with will be really useful when you have a real problem to solve. I’ve already copied it all into a Notion page for future reference. Check it out.
Well, since we’re on the subject of ChatGPT, you should really be aware of the ‘Make it More’ trend where users post a series of images asking Dalle-3 to ‘make it more ’
Generative AI Art: 2016 to the Present
Nettrice Gaskins has been experimenting with AI art for a very long time. In this illuminating essay, she really helped me understand how we all got to where we are today.
In 2019 people were mostly curious, not freaking out about AI art and I was committed to using Deep Dream Generator to produce and post at least one image per day for 365 days. I was experimenting and learning about the possibilities.I wanted to do things no one else was doing and it was getting noticed.
That’s it for today! I hope you’re getting a lot out of this newsletter. As always, send me a message. I’m always ready to keep learning if you have a request.