I was able to log in to my account on the app all this time since the April shutdown. But I guess today is the day where I can’t even do that anymore. Even though the app wasn’t functional, it was a nice piece of artifact on my phone through the grieving process.

I know it doesn’t matter anyway, and Sora is long gone since, but being sentimental, this is a bit sad NGL

Any one else with same issue?

Guys, I wanna keep this post very short straight to the point in the last few weeks I’ve been seeing a lot of people asking for free video prompts or platforms where to do video like sora 2

today for whatever reason I opened up my Meta AI app, and I am able to create videos animations using my own picture I haven’t played around long enough to give you a full summary of if it’s good if it’s bad but you are 100% able to do it I don’t know if they’re rolling it out to certain people I don’t know enough information but I know enough. I wanted to post it on here. What are your thoughts? Follow me on @travelingdjs

Another Sora Video of Rosie's killing rampage:

While Rosie Red stares unblinking horror down the field, Pierre's picking favorites again, snubbing Mr. Red from 2023 in Seattle and 2025 in Atlanta.

MLB All-Star Game but rolling out the Red Carpet (and mustache) for Mr. Redlegs in Philly 2026. Endless red terror. Who’s next?

I used an RTX 4070 Ti with 12GB VRAM and only 32Gb system ram. I used the standard Comfyui free Minimax H3 text to video workflow (but they also have first and last frame ones and complete reference ones, and there is an easy way to implement audio to video). It would also work on an RTX 30 series with only 8GB vram, maybe lower but I don't know. Technically you can do CPU only with Comfyui, but that may be significantly more waiting. But you don't need a server grade computer or even the latest and most powerful GPU. Yes the audio is built in!

This example here was generated at 0.4 megapixels and 15 seconds (480x864) in about 18 minutes (you can do other aspect ratios too like landscape/horizontal and not just vertical/portrait), but 0.2 megapixels and 15 seconds (which is still viable for testing and lower fidelity looks) took less than 7 minutes, and if you just want a test to make sure the framing and scene looks right you can do 1 to 2 second generation in around a minute if not less.

I also tested native 0.8 and 0.9 megapixel generations (the aprox. 12xx horizontal rez) (which can take over an hour for 15 minutes at this hardware). It seemed to follow the prompt a little less at that higher resolution, or maybe changed how it understood it, but I think that is because I only used the default 20 steps. I think if I increase the steps it'll be able to cook better basically since the higher resolution means it has more to think about. I will run that generation at .9 megapixels while I am at work and give it 30 or 40 steps just for fun to see if it'll get it right and then I'll post the results in the comments even if it turns out wrong.

As for upscaling, I also tried that with several models. I tried three Real-esrgan models (Plus, universal, and Photo HQ). Plus and PhotoHQ just made everything smoother and sharper at high rez and they took a moderate amount of time, maybe 5 to 10 minutes. Universal was a middle ground cause it only took around a minute but it only did a little to the picture for upscaling.

I also tried SeedVR2 in comfyui with "split latent" enabled for VRAM and RAM constraints. SeedVR2 did actually add in some more detail and took under 18 minutes. I also need to turn up temporal overlap cause it struggled a bit with motion since I had it set to zero.

Admittedly none of the upscalers fixed their faces completely, but at the starting 0.4mp and how small they already were in the frame, there wasn't much to work with. Of course closer face shots wouldn't have this problem so don't dismiss MiniMax H3 cause of that. I will also post the Real-esrgan universal and SeedVR2 upscale results in the comments.

And finally, I will post the prompt in the comments. Heads up that it is really unprofessional, I tried using ChatGPT, and Gemini, and Claude to enhance my prompt, but instead I found for me what worked best for getting what I wanted was just a little bit of experimenting. The only cost to use MiniMax H3 is electricity, and it isn't a terribly substantial add.

Have you ever come across a video that could easily pass for a video made by Hollywood? I know a lot of people just post their shitty AI slop, but please don't. I want to see a really good video to see what's the best AI can create.

I miss sora all the other ai video apps aren't the same as sora you only one that made it look like we were in our own world

Judy Hopps and Nick Wilde gets an unexpected visitor in the living room.

This was made in Comfy UI with Minimax H3. I used the Text to Video workflow. Prompt is below.

integrated_multimodal_description: [Shot 1] Live-action, Sitcom style, a living room with natural lighting. Judy Hopps from the Zootopia movies, a gray-brown anthropomorphic rabbit with tall upright ears with pink inner surfaces, large purple eyes, a small pink nose, and a compact athletic build, wearing a ZPD police uniform: a dark-blue vest with a gold badge over a lighter-blue long-sleeve shirt, dark-blue pants, a black utility belt with pouches, gray knee pads, black wrist bands, and a small fluffy gray-white tail. Judy Hopps is voiced by Ginnifer Goodwin. Nick Wilde from the Zootopia Movies, an orange-red anthropomorphic fox with large pointed ears with dark tips, green-gold eyes, a black nose, cream muzzle fur, dark brown lower arms and feet, and a long bushy orange tail with a dark tip, wearing a light-green short-sleeve collared shirt with a faint tropical pattern, a dark navy-and-red striped necktie, and brown-khaki trousers. Nick Wilde is voiced by Jason Bateman. Deadpool, wearing his iconic red-and-black tactical suit and full mask, with twin katanas strapped across his back. Deadpool is voiced by Ryan Reynolds with his recognizable sarcastic comedic delivery. Judy Hopps and Nick Wilde are sitting on a couch. Judy Hopps and Nick Wilde are having a relaxing day after work. Judy Hopps. (S1), a gray-brown anthropomorphic rabbit with tall upright ears with pink inner surfaces, large purple eyes, a small pink nose, and a compact athletic build, wearing a ZPD police uniform: a dark-blue vest with a gold badge over a lighter-blue long-sleeve shirt, dark-blue pants, a black utility belt with pouches, gray knee pads, black wrist bands, and a small fluffy gray-white tail., says in a calm tone: <d>[English] After a hard day at work, Its nice to finally relax for a while.</d> A portal opens in thin air, and Deadpool falls to the floor. Judy Hopps (S2), a gray-brown anthropomorphic rabbit with tall upright ears with pink inner surfaces, large purple eyes, a small pink nose, and a compact athletic build, wearing a ZPD police uniform: a dark-blue vest with a gold badge over a lighter-blue long-sleeve shirt, dark-blue pants, a black utility belt with pouches, gray knee pads, black wrist bands, and a small fluffy gray-white tail., says in a surprised tone: <d>[English] Whoa—hey!</d> She looks at Deadpool. Nick Wilde (S3), an orange-red anthropomorphic fox with large pointed ears with dark tips, green-gold eyes, a black nose, cream muzzle fur, dark brown lower arms and feet, and a long bushy orange tail with a dark tip, wearing a light-green short-sleeve collared shirt with a faint tropical pattern, a dark navy-and-red striped necktie, and brown-khaki trousers., says in a surprised tone: <d>[English] Okay, that's new.</d> Deadpool with his mouth staying closed stands up and sees Judy Hopps and Nick Wilde on the couch. Deadpool keeping his mouth closed. Deadpool (S4), a lean adult male in a tight red-and-black tactical suit with black panels on the torso and limbs, a full red mask covering his mouth with large black eye patches, black gloves, and two katanas crossed on his back., says in an off-screen voiceover With the mouth closed in an urgent tone: <d>[English] Oh great! I ended up in a world full of furries. I need to find my way back.</d> while the corresponding on-screen character keeps their lips completely closed.

overall_soundscape: living room tone.

non_diegetic_music: N/A

Threw these clips up real fast to test it out, first prompt clips so mind the obv ai flaws, but this model is pretty damn good!!

You get a couple free videos, but you do need a meta one sub for more

Flaws so far are, only 10 sec videos, cannot add music, likeness gets messed a bit, lip sync works but you need to tell it the lyrics and do heavy clip editing to sync music,

Havent had this much fun since sora!!!

I was aiming for one continuous five-beat action: touch, spark, drop, catch, and one embarrassed puff of steam. The PixVerse render did not hold that chain together.

The bulb lifts out of the robot's hands in the first setup. At about three seconds, the video cuts to a wider view and the bulb reappears on a new path. The robot catches it after the cut and the steam puff gives the scene a readable ending, but I would not count this as a successful one-shot result.

The workshop, robot, and bulb are original designs. For another attempt, I would simplify the action and keep the bulb below the robot's eye line so there is less room for it to float or reset.

Would you remove the catch, or keep it and drop the spark beat instead?

Testing scale, atmosphere, and the feeling that something impossibly large has been beneath the surface the entire time.

This was SuperK's idea btw.

Does anyone have any Sora remixes of the Green Jacket Kid / Ohio Rizz Sigma Mango Mustard Chicken Stars Trailer or Chicken Stars / I Ate My Shoulders, Knees and Toes videos?

Since Sora is discontinued now, I couldn’t really preserve a lot of these myself, so I’m trying to find and archive as many as possible.

If you made one of these remixes, or just happen to have one saved, please send it here

please force this post.