It’s been a minute since I’ve seen the latest in AI video and holy sh*t!
Check out these examples (slightly NSFW):
These were made with ComfyUI and AnimateDiff and a few other programs available to the public (upscalers, post-processing, etc).
Honestly, it looks better than Runway Gen-2’s outputs with their $237 million in funding. The open-source AI community is amazing.
You have to know your way around ComfyUI and a few other programs to get this result but it is still quicker than firing up Blender or Cinema4d for ad concepting.
I used to spend hours a day on Blender creating 3d motion graphics to use in ad concepts. I have been an advocate that AI is an alternative for that workflow.
As we know, the rate of AI advancement is maddening. I haven’t dived into ComfyUI, I use Automatic1111 for AI image and video workflows.
If you want to learn how to do this I’d start with this tutorial on YouTube:
Want to make AI video ads in minutes? Check out the platform I’m building:
Sovran.ai updates this week:
shifted text on 9x16 higher
added email notification when video is complete
slight redesigns
Roadmap:
save last input fields
4x5, 16x9 aspect ratio options for Meta feed and YouTube placements
improve visual hooks to increase hook rate
improve scripts to increase hold rate and conversion rate
I am iterating on the video outputs to increase hook rate and hold rate. Ultimately, running on Meta/TikTok will be the judge.
We have 39 users already and it’s in free open beta for a limited time.
- Manson