As you’ve probably heard, OpenAI is shutting down Sora.

The service, which has been stratospherically expensive for OpenAI to run, was always a weak consumer proposition.  You could create videos of Elvis riding a dinosaur, and anyone could make up to 30 videos a day for free.  The videos – which could be up to 15 seconds in length – cost OpenAI a fortune on a per-request basis.  Not only were they not profitable – not even close – they also didn’t build towards any kind of lasting commercial relationship.

Consumers tried it, and for a while it was fun to make videos of your grandmother riding in a tank or some celebrity (one their filters hadn’t caught yet) arm-wrestling the Pope.  But then the novelty wore off and you moved on to other things.

It was an expensive service to run.  We know because you can pay to use Sora through Azure and other services.  There, you’ll pay 10 cents a second.  Since the Sora app allowed people to make 30 videos a day, potentially anyone could run up a bill of $45 with no revenue for OpenAI.

Sam Altman expected brand partnerships to help shoulder the cost, but it’s difficult to understand his vision.  What does Disney gain from allowing you to create a video of Tinkerbell being tased and arrested?  Why would they want to risk their brand like that?

Turns out they don’t.  Disney had pledged $1 billion but has no canceled the deal, which probably either caused or rapidly accelerated Sora’s demise.

I Used It

If you watch a lot of YouTube, you may find yourself actually listening to a lot of YouTube.  I know I do.  Today I listened to some videos on

• Why Microsoft’s CoPilot is a flop
• What’s going on in the Iran war
• The role of jesters in the Middle Ages (turns out that being able to fart on command could get you a castle)

…and others.  Very few of these really needed visuals.  The makers are on the platform because that’s where the eyeballs are, but none of these videos actually required me to watch the screen.

However, you have to put something on the screen.  Historically this has been done using stock footage.  A lot of videos look like this:

• Intro graphics
• Some charts, images, or video bits that are pertinent to the topic
• And the rest is stock footage

That’s what I did for our new series on AI Companions.  In the next video (to be released soon!) there’s some charts showing how the tech works.  But probably 50% of the time, the video is me speaking over stock footage.

I tried using Adobe Firefly for this and it was a disaster.  I did use a stock footage site, and that was excellent for a lot of parts.  But I wanted videos that were very specific, and there simply was no stock footage available.

So I used Sora, and in some cases, Leonardo.  Leonardo’s videos were higher quality, but Sora’s were free.

The main drawback with Sora was that irritating watermark.  If you use Sora through some paid-for service, there’s no watermark, but if you just go to the Sora website or use the app, there is.  I decided that having a floating “Sora raindog308” watermark was OK for LowEndBoxTV but it isn’t for a lot of other uses.

So was I just being cheap?  Part of that is tech-driven.  I don’t mind making images on Leonardo because I’ll do 4 or 8 at a time and pick the best one.  Or on ChatGPT, where because I subscribe, it’s free.  But doing video generations is expensive (at least $1.50 each as we priced above).  Considering you often have to do 5 or 10 to select the best one, filling up a 15 minute video (which is 60 15s segments) would be very expensive.

I suspect the tech is not going away.  Just the free consumer-facing side is.  OpenAI will still allow people to use the API, which is available through Adobe, Leonardo, and others.  But the days of making free slop are gone.

(Or will be soon…the app is still up and working as of this writing).

raindog308

raindog308 is a longtime community LETizen, technical writer, and self-described techno polymath. With deep roots in the *nix world, he has a passion for systems both modern and vintage, ranging from Unix, Perl, Python, and Golang to shell scripting and mainframe-era operating systems like MVS. He’s equally comfortable with relational database systems, having spent years working with Oracle, PostgreSQL, and MySQL.

As an avid user of LowEndBox providers, raindog308 runs an empire of LEBs, from tiny boxes for VPNs, to mid-sized instances for application hosting, and heavyweight servers for data storage and complex databases. He brings both technical rigor and real-world experience to every piece he writes.

Beyond the command line, raindog308 has a life-long love of German Shepherd Dogs, high-quality knives, target shooting, theology, tabletop RPGs, playing guitar, and hiking in deep, quiet forests.

His goal with every article is to help users, from beginners to seasoned sysadmins, get more value, performance, and enjoyment out of their infrastructure.

You can find him daily in the forums at LowEndTalk under the handle @raindog308.