I assumed Sora would launch with basic features and really restrictive limits. I was wrong. Their team used those 10 months of development to build something polished.
First, the usage limits exceeded expectations. You get 50 video generations monthly with a $20 ChatGPT Plus subscription – that’s generous considering the subscription was already worth it just for ChatGPT. Video generation isn’t cheap. Even with their optimized Sora turbo model, each 5-second 480p video likely costs OpenAI at least 25 cents to generate.
The $200 tier with unlimited generations is clearly subsidizing the $20 tier. There’s no way someone on the expensive plan generates 10x more videos to justify the price difference.
But the interface truly sets Sora apart. This is the cleanest, most intuitive video generation interface I’ve seen. It feels like an Apple product – beautiful and functional. Beyond basic video generation at different resolutions and lengths, you get style presets, video looping, start frame uploads, and batch generation controls.
The storyboard interface provides even more control – you can blend videos, extend them, add keyframes, and remix elements. It’s a complete video creation studio, not just a model that spits out clips.
The model itself performs well but isn’t dramatically better than alternatives like China’s open source Hunyuan. The interface makes up for this by making the whole experience feel premium.
Initially, new signups were blocked due to server overload and generations took forever. Things have stabilized now – you’ll typically get a 5-second 480p video in a couple minutes.
For more context on Sora’s development and initial announcement, check out my previous coverage here: https://adam.holter.com/openai-plans-12-days-of-ai-releases-including-sora-and-new-reasoning-model/
