Adobe is buying video data for AI models

Adobe's playing catch-up in the AI video generation race. With rivals like OpenAI's Sora pushing boundaries, Adobe is scrambling to build its own text-to-video AI model.

What's going on here?

Adobe's buying up tons of video footage to train its AI model.

What does this mean?

Video tools are Adobe’s bread and butter (or maybe just butter, bread might be images and PDFs), and it’s serious about adding AI into these tools. They've shown their hands at AI image generation with Firefly, but video is the next frontier.

OpenAI’s Sora and other AI video models are putting pressure on Adobe to catch up but to keep on its promise of legally and commercially safe AI, Adobe can’t just scrape Youtube videos.

Adobe's past strategy relied on stock libraries, but they're going beyond that to build something competitive. It’s paying creators to submit short videos of common activities. Think videos of people doing everyday stuff, interacting with objects like smartphones, etc. The pay is about $3 per minute but can go as high as $7 in some cases.

Why should I care?

If you're building creative tools, this is a wake-up call. AI is rapidly changing how content is made, and staying relevant means adapting. AI image was so 2023, we have AI music now, and soon AI video. In the end, it’ll always be individual creativity as everyone will have access to the same tools.