Hacker News Reader

Anything technically interesting you can share about any of these?

How do you decide on which parts to extract?

What models do you use for the features you listed?

Sure, here’s what I can share: we use Veo 3, and Sora 2 is coming soon. For captions, we rely on FFmpeg combined with our own speech-to-text and dynamic subtitle API.

Reply on news.ycombinator.com