alt Hacker NewsWhenever I see the new releases around video generation (and image) generation models, I get goosebumps, because it just feels so fun to work with them. But then I remember that I spend upwards of $10k on inference generating well over 50k images for storyboards, training models; and probably creating almost an hour of video (I assume). Yeah, I get that things can be economic if you don't use the latest models (ran some case studies on this), but the latest models are the most fun to work with. It doesn't scale as well as "vibe coding" stuff together on the weekend. And when things work really well its almost as if you're seeing an zoopraxiscope come to life for the first time; and you just want to keep going.
I got a few offers to work with some startups in this space, but it also seems that many startups work on stuff that just doesn't seem to be very worthwhile (like creating masses of spam for YT or TikTok shorts), or even straight out morally/ethically wrong (cloning/deepfakes, etc). But seeing advances in this space; and coming from a filmmakers background, I might just end up being naturally drawn to this space on an engineering level and figuring something out along the way. As you can see I worked on a lot of stuff just for the fun of it, and documenting the process: https://edwin.genego.io/blog (but I stopped at the beginning of the year .... might.. just pick it up again.
Replies
MiniMax H3 is going to release weights. You can locally run it with definitely less than $10k (and possibly faster than Seedance's queue), and it's fun to train it for whatever you need.
I've been dabbling in this space on the application layer and have spent a few hundred myself experimenting with video models.
It can be really entertaining/addicting to build with them because you're essentially pulling the slot machine and having TikTok/Marvel/YouTube come out of it. I think that was the bet with Sora but the problem is mostly that the novelty wears off quick, and most people want to just consume content without typing in what content they want to see, or sifting through mass-generated spam "content" with nothing behind it to make it worthwhile (a lot of people engage with content parasocially)
Once the tools for creators to steer and integrate models in this space get better, it will explode. We've been working on what I think will be one of the first use case for integrating these models, because I think we're approaching a middle ground where they can be integrated in experiences to provide entertainment/engagement/fun experiences without feeling like slop.
Btw, I'm impressed with some of the AI content on your site but I think you might want to pare down the non-demo pages because it has a different impression me (can I trust that this text is true? / I'm reading a lot of words but not really learning about this person) than you might have intended). I'm a bit of a hypocrite here but also speaking from experience.
I went to Art Center for film. Loved it. But ended up writing software instead of shooting movies (while still also handling a lot of visual art direction, graphics work, UI, 3D animation, etc). Now I feel like we're starting to be roughly in the same boat as far as using prompts.
What bothers me is that every piece of content generated this way helps flood an already saturated market for content, while slowly degrading the expectations of what people see, to the point that no one will bother with shooting or animating anything anymore. Even if it's 50% worse, it's 90% cheaper, so the economics argur against producing any new physically made content. Simultaneously, it's cannibalizing all existing content. This points toward a feedback loop, like a snake eating its own tail. And even though Hollywood blockbusters have followed that pattern for a couple decades, it's demoralizing to me to see it enshrined as the future of film (or to hear from someone who makes films that it would be a preferred mode of creation).