https://t.co/Z5NLeDkAL7
One thing to note in this article is that the journalist says it's just "deceptive" or "misleading." The TRUTH of the matter is that Trump is LYING. It's not just simply misleading. Trump is lying about the figures. You need to say what it really is.
@kling_ai@googleai@pixverse_@veo_ai@hailuo_ai@seaart_ai@grok@invideoofficial@midjourney
This isn’t an exhaustive list, but these are some of the main players in AI. (Sorry if I have any wrong.)
With Sora AI ending soon, I am genuinely surprised at the lack of similar character generation and control anywhere else. Those that do have some semblance of working with characters make it so damn hard to work with, they don’t attract the general public as users. Is that by design? Why do you all make it so difficult? Why should the user have to figure out how to train a model when the software can do it on its own?
What made Sora great was that you could generate a character reference video, and seemingly on the back end, it automatically created needed references to train a model. Once you then used the character in a generated video, that created additional references to train the model to finalize the character. This was true for cameos of the user as well. All you had to do was refer to that stored character with the assigned name. That video generation process was also linked to the ChatGPT LLM for world building. The TikTok-like feed was familiar to users. Without any of this, none of these services will get ahead of the others.
We know that Sora lost money because it used so much power for computing without having any income from subscriptions or other revenue. Most of all these other platforms have paid levels. If that is still not enough, consider utilizing the user’s device for part of the computing power. Utilize ads for revenue within the feed of user videos but have them generated by the same process so they will be interesting to watch.
It’s unfortunate that Sora is ending. Many called it all AI slop, but I don’t see much difference between Sora’s content and the content of all the other platforms. I used it myself to generate a running story. The one thing that is different is that they trained their model from these videos, and they will still have that data. I believe Altman either didn’t plan or execute Sora properly, or Sora was never meant to stick around in the first place. I tend to believe the latter considering the potential investment from Disney to use their characters. They were as surprised as we were.
The end of the story is that the first AI to duplicate Sora’s process and make it automatic the same way, same kind of feed, connected to a LLM, and easy to use will dominate the market, simple as that. You can flame or bicker at me as much as you like, but it won’t change that fact.
Honestly, if computing power was such the problem, what about offloading some of that computing to the local computer? So many computers have high power processors (like Apple’s M1-M4, Nvidia, etc.), you could have the app utilize the user’s computer or phone to do some of the computing. We’ve seen it before. You really need to explore your options before just dumping the entire thing outright.
I cannot stress how huge a mistake this is, first for branding and company image of OpenAI, and second for the creation you’ve built, notwithstanding the times it broke. No one has ever made an AI app in the form of a social media app before. You literally don’t understand what you have here. Do you not see the importance? I agree that there should be a paid subscription version where characters and stories with audio can be created with accuracy and consistency with less restrictions, but I would also posit that to have the ability to create quick consistent character video content with audio for social media is equally important. Perhaps only very basic functions be available for free and a paid version with more features unlocked in a subscription type model be available. If I was the CEO, I would not be leaving that up to a competitor. You have a chance to be the pioneers here with all that comes with it if it’s executed properly. Don’t axe the Sora app but evolve it instead into two to three offerings. Look at the creativity that the app has sparked. I’d want to keep that going.
@Kling_ai So with the demise of Sora, Kling may be the closest to achieving what Sora had. While it’s nice you can get characters by defining the via Elements and then training your own model, here’s why that’s the wrong approach:
1. Sora generated the initial character video
2. Sora AUTOMATICALLY stores references from the initial character definition when the user tells it that they want it as a character, including audio, creating a voiceprint.
3. Once the first video is then generated from this character reference, it continues to add additional references for image and audio AUTOMATICALLY to train the model.
Do not expect the regular user to be a techie/computer geek who knows their way around computers and to know how to train AI models. Sora automated all the character building so all you have to do is have it place your character into the world created by your prompt. Sora also did the work of having objects referenced to train its model. People forget that was an option. That’s why the world building was more realistic. (All this is why it should have immediately had a paid option after it was introduced since all that data processing cost it a lot of money.) And then they were clever to put their video sharing option into a TikTok-like form which people are familiar with and is easy to use. Never assign an engineer to do UX/UI. Make it easy for the user, and you’ll have a better model, and the users will break down your door to use it. There’s the blueprint to be a great AI generator right there.
@Jim_Jordan Because Trump sabotaged it in his first term. Wow, you really have a bad memory. No wonder you don't remember the sexual abuse that happened at OSU.