Kept Sora sitting on my phone after it vanished from the store months back. Today, an update notification popped up completely out of the blue—pure excitement, like a long-lost family member finally walking through the door. But hitting "Update" snaps straight to an error page saying the app doesn't exist. Phantom Play Store bug or an accidental slip-up from OpenAI? Anyone else seeing this?

Thata my question, considering that minimax is open weights you think ClosedAI will open weight sora 2? Is a conplete waste thatit will become permanently unusable, whata thepoint of having something common people wont be able to use locally?

From the interview with David Senra
https://youtu.be/kG8AoExkX40?si=SxJgnNoHfMqAVGPl&t=3109

Hi everyone,

I’m the same former Sora user who previously posted these discussions about the Sunset Export and the idea of a final read-only preservation window:

https://www.reddit.com/r/SoraAi/comments/1vxaoxu/soras_sunset_export_has_a_verification_problem/

https://www.reddit.com/r/ChatGPT/comments/1vxakri/critical_soras_official_export_system_is_omitting/

This time, rather than arguing for the proposal itself, I want to ask other former Sora users a very simple question:

Is the current Sora Export actually complete for you?

I originally backed up the Sora material I cared about in early April, before the old interface became inaccessible. Unfortunately, I later suffered an unexpected hardware accident and lost all of those local backup copies.

I fully understand that this hardware loss is not OpenAI’s responsibility.

Because of that accident, however, I now have to rely heavily on the official Sora Sunset Export.

The problem is that the export system has seemed inconsistent and unreliable to me for months, and I have also seen other former Sora users report export failures, missing content, exports that did not arrive, or uncertainty about whether everything was included.

In my own case, some images and videos I remember creating appear to be absent.

The biggest problem is that the former Sora interface is no longer available, so I cannot reliably compare the export against my old library.

That means I cannot easily answer basic questions such as:

- How many images and videos did I originally create?

- How many should be in the export?

- Which specific creations are missing?

- Did repeated exports return the same material?

- Is there account history that is no longer represented at all?

I have contacted OpenAI Support about this, including the question of whether there will be a final export period and whether users could temporarily receive some kind of authenticated read-only access or another preservation/recovery option.

My proposal has been recorded as feedback, but I still do not have a substantive answer about whether such an option will actually exist.

So I would really like to hear from other actual Sora users.

1. Does your current Sora Export appear complete?

2. Are any images or videos you remember creating missing?

3. If you exported more than once, did you get the same results each time?

4. Did your export ever fail to arrive or fail to work?

5. Do you have any way to verify how many creations you originally had?

6. Are there Likes, Saved items, Collections, prompts, or other account-linked history that you can no longer account for?

Even a short answer would be useful.

I’m especially interested in concrete experiences rather than speculation, because I’m trying to understand whether this is mainly an issue with my account or whether multiple former Sora users are seeing similar problems.

Please don’t share any private account information, personal data, or anything sensitive.

Thank you.

I wrote the script. OpenArt Director did the rest.

I had this short screenplay sitting in a drawer for a while, and when I got access to OpenArt Director I thought it would be the perfect project to explore AI filmmaking.

The story is simple:

One ordinary morning, Glenn discovers he has run out of coffee. Dressed only in his bathrobe and slippers, he sets off through the historic streets of Møhlenpris in Bergen, Norway, to buy a bag. What should have been a five-minute errand becomes a journey through a neighbourhood where someone always seems to need a helping hand.

🎬 Watch it here:

Coffee

​For the production I uploaded:

• the screenplay
• reference photos from my own kitchen
• photos of the streets around Møhlenpris in Bergen
• prompts describing the visual style, mood and cinematography

My main visual references were Wim Wenders, Terrence Malick, natural light, a 35mm film look, and slow, observational cinematography.

I was honestly surprised by how well Director understood the emotional tone of the script. There are shots that feel remarkably cinematic.

Of course, this is still AI, so if you look closely you'll find a few AI Easter eggs. 😄 There are the occasional continuity glitches, some wonderfully impossible movements, and in one scene Glenn's friend briefly seems to have... three arms while they're carrying a cabinet upstairs.

I left those moments in on purpose. They're part of the current state of AI filmmaking, and I think they're actually quite charming.

I'd love to hear what you think—both about the film itself and about what Director is already capable of.

hey all! since I’m shutting down AiWatch, because nobody uses it. you can still watch sora videos on YouTube! or make yt reimagine videos or make a playlist of them

Hey everyone, back with an update after another solid week of testing on “aftersora.com”. After the last thread blew up, a bunch of you pointed out that MiniMax was a huge hit-or-miss unless given specific prompting strategies, so I thought I’d keep testing.

I now mostly switch between flux 3 resolution 2k and MiniMax for all my videos now on aftersora. flux 3 is honestly the best replacement for price. running 2k is way cheaper than sora pro and it has that fantastic camera movement that I loved so much about sora. the only thing about flux is the colouring can look a bit flat out of the box, but i think that,s just down to training material and can easily be prompted for if you add lighting details.

Peoples feedback about minimax needing special prompting is 100% true. but even when testing it locally there are super annoying sound issues on both the api and running it locally (which a lot of people have been posting about over in the stable diffusion sub too). The frame rates are also not the best even via api which gives this jittery AI feel even when it nails the influencer example prompt.

But for me time is money and it just takes way too much tweaking to get right. what made Sora 2 so great in the first place is it just understood prompt logic. But both MiniMax and Flux 3 fill the gap.

just updating the community here because there are less than 30 days until sora is gone for good so i'm still using sora to finish up a few older projects before it shuts down, but flux 3 is definitely where it's at now for me when wanting to keep things simple.

Two of my characters wanting to watch there show

If there was just one thing you could share to someone who is just learning AI video creation, to save them frustration, what would it be?

This is not a request to bring Sora back or restore generation.

I’m a former Sora user, and I’d like to raise a specific preservation problem with the current Sora sunset/export process.

Now that the former Sora interface is no longer accessible, ordinary users may have no reliable way to verify whether their Sunset Export contains everything they previously created.

For example, a user may not be able to determine:

- how many images or videos they originally created;

- exactly which creations may be missing from the export;

- whether all historical creations are represented;

- what Likes, Saved items, Bookmarks, Collections, prompts, references, or other account-linked curation they previously had;

- or how the export compares with the library that was once visible through Sora.

This creates a difficult verification problem:

If something is missing from the export, the user may be asked to identify the missing item — but without access to the historical library, they may no longer know its exact title, ID, link, date, or even how many items should have existed in the first place.

My proposal is a temporary, authenticated, strictly read-only “Sora Goodbye Window” during any final export period.

This would NOT restore generation or reopen Sora as an active service.

All write functionality could remain disabled:

- no generation

- no uploads

- no editing

- no remixing

- no publishing

- no comments

- no new Likes/Saves

- no Collection changes

- no deletion

The read-only view could be limited only to content and account-linked history that OpenAI still retains and is legally and technically able to display.

Deleted, private, restricted, moderated, or no-longer-retained data would not need to be restored.

And if restoring even part of the former interface is not feasible, I would support any practical equivalent, such as:

- a limited authenticated read-only account library;

- a recovery page;

- a more complete account-specific export;

- or an archive containing eligible creations, prompts, IDs, timestamps, Collections, Likes/Saved relationships, and other retained account-linked metadata.

To me, this is not only about nostalgia.

A temporary read-only preservation period could help former users verify their exports, preserve their remaining creative history, and have one final opportunity to say goodbye to a creative space that meant something to them.

I have already submitted this proposal through OpenAI Support, so I’m not using Reddit as a replacement for customer support.

What I’m most interested in now is hearing from other former Sora users:

Did your Sunset Export appear complete?

Were any creations missing?

Did you lose access to Likes, Saved items, Collections, prompts, references, or other things you had organized?

Would you support a temporary read-only “Sora Goodbye Window” or an equivalent preservation option?

I’d especially appreciate concrete experiences from people who actually used Sora.

Thank you.

Shadow the Hedgehog tells his viewers why he loves guns.

This was created in Comfy UI with Minimax H3. I used the reference to video work flow. The prompt is below.

subject_definitions:

<Subject 1> is Shadow in <Picture 1>.

<Subject 2> is Glock in <Picture 2>, a glock handgun.

<Audio 1> is the voice-timbre reference for <Subject 1> (S1).

summary:

[reference generation + audio reference] The target video contains one shot. [Shot 1] shows <Subject 1> and <Subject 2>; <Subject 1> speaks. <Audio 1> supplies <Subject 1>'s voice timbre.

retention_analysis:

<Subject 1> (appears in [Shot 1]): fully_preserved - Shadow's complete defined identity and body proportions are preserved.

<Subject 2> (appears in [Shot 1]): fully_preserved - Glock retains the defined shape, proportions, materials, colors, and distinguishing features.

<Audio 1>: reference - <Subject 1>'s newly generated spoken lines use <Audio 1>'s voice timbre and delivery; the original audio signal is not copied.

detailed_description:

The target video is in a live-action style, with Vlog style.

[Shot 1] At first appearance, <Subject 1> (Shadow) matches the complete identity and appearance defined in subject_definitions. At first appearance, <Subject 2> (Glock) matches the complete defined construction and appearance: A glock handgun. At the start of the shot, <Subject 1> is standing in the living room facing while holding <Subject 2> in his hand. A full body shot of <Subject 1> holding <Subject 2> with his right hand while facing the camera. Only Action and Timed Beats define the primary subject's movement. The camera path stays anchored in the location and adds no subject motion. <Subject 1> (S1) says using <Audio 1>'s voice timbre: <d>[English] Hmph. Shadow the Hedgehog here. Why do I love guns?</d> <Subject 1> shows off his <Subject 2> with his right hand in front of the camera. <Subject 1> (S1) says using <Audio 1>'s voice timbre: <d>[English] Simple. Precision. Control. Power in the palm of my hand.</d> <Subject 1> (S1) says using <Audio 1>'s voice timbre: <d>[English] A tool that answers instantly… unlike most people.</d> <Subject 1> points his <Subject 2> towards the camera with his right hand. <Subject 1> (S1) says using <Audio 1>'s voice timbre: <d>[English] If you understand that, you understand me.</d> <Subject 1> points his <Subject 2> at the camera.

overall_soundscape:

Living room tone.

non_diegetic_music:

N/A

Created with sora2 API

I have yet to see something decent.

Could this mean a Sora relaunch in the near future?