This Video is Not Real
0 up · 0 down · 0 ratings
Description
this video is AI generated so I just reviewed Sora this was like a week and a half ago it was open ai's video generation model and it was really impressive uh was until Google showed up so they've just unveiled their latest and greatest probably inspired by Sora just getting announced but the scary thing is it's already noticeably better than Sora see Sora would struggle with some predictable things I talked about in my video things like physics and moving objects and vo already seem to be significantly better at that stuff specifically so Sora would often have dead giveaways just in the movement of things or in the physics of things moving around in a frame and vo would tackle the same prompt and make a much more convincing at least physical version of that movement and so with these starting to show up a lot of people are reacting like oh you know well these are just the best handpicked versions that Google is showing us maybe it's not that much better but I've had access to V2 for a couple of days now and I've been doing a lot of prompting a lot of back-to-back testing with Sora and yeah it's it's better than Sora now it's certainly not perfect there are definitely still dead giveaways in there of AI generated video especially with garbled text that still happens a lot also when you try to do a whole bunch of things in one scene or really really complicated physics things then it can still sort of break down and also this version that I'm testing doesn't have nearly as many controls as Sora it feels like they rushed it out to match the open AI announcement but there's no resolution controls or picking the length of the video for example but to me this really feels like the difference between using a little bit of YouTube data that they're getting away with versus owning YouTube and just just using all of it YouTube just added that button I think it's in Labs where creators can opt out of thirdparty scraping of videos for AI training but that's just third party stuff that doesn't say anything about Google not being able to use it but yeah it's clear that things are already evolving very quickly in the video generation space which as a video creator are are we cooked cuz
This short opens with the creator stating that the video itself is AI generated and discusses recent developments in AI video generation, comparing Google's VeO 2 model with OpenAI and Sora. The speaker notes that Sora performed well in certain predictable tasks, particularly with physics and moving objects, but VeO 2 delivers more convincing physical movement and fewer telltale signs of artificiality in many scenarios. As the video progresses, the creator evaluates limitations still present in VeO 2, such as garbled text and occasional breakdowns when handling complex scenes or heavy physics prompts. There is a critical observation that VeO 2 currently lacks some control features found in competing tools, like explicit resolution settings or video length controls, which makes it feel rushed to align with an OpenAI announcement. The speaker also reflects on YouTube data usage, highlighting a new opt-out option for third-party scraping in YouTube Labs, and discusses broader implications for how creator content is trained and monetized. Overall, the video situates VeO 2 within a rapidly evolving landscape of AI video generation, weighing immediate practicality against future potential and ethical considerations. The tone blends excitement about improvements with caution about overreliance on AI-created media and the need for thoughtful regulation and transparency as the technology advances.
Topics · technology · artificial intelligence · video creation · ai ethics
Questions answered
- What are the key strengths of VeO 2 compared to Sora according to the video?
- VeO 2 delivers more convincing physical movement and fewer AI giveaway artifacts in many scenarios, especially with physics and object movement, compared with Sora.