I tried OpenAI’s Sora 2 and I’m not exaggerating when I say I was blown away. In this video, I give you a first look at what Sora 2 can actually do when you feed it text + visual input (like a photo or a “cameo” of your face) and it generates a full video output that looks insanely realistic. The reason it’s breaking the internet is simple: the videos are physically accurate, the dialogue can be synchronized with visuals, and it’s way more controllable than older video generators. The second reason is access—right now it’s invite-only, and I show you the app experience that feels like a TikTok-style feed for AI videos. I walk you through my screen, generate a video live, and show a bunch of examples I made—funny ones (like Sam Altman stealing my In-N-Out) and more serious ones (like me “teaching” multimodal systems in a YouTube-style clip). I also show the remix feature, where you can insert your cameo into someone else’s video, and I’m honest about the rough edges: guardrail warnings, random errors, and moments where audio/video sync isn’t perfect yet. Finally, I talk about the scary part—deepfake risk and misuse—because any powerful tech can be used for good or bad, and we need to see how OpenAI handles safety as this scales.