That’s a great prompt. I like using somewhat vague prompts and letting the AI fill it in. You get interesting and unexpected results.
Midjourney’s prompt following isn’t that strong (it works fundamentally differently than the engines you guys are using for the most part) but it has the spirit of a good one even if it’s not an old timey blurry photograph
This one in the same batch is kind of funny. It got the old timey photo right this time, but not the sort of cryptid mythology part
And this one isn’t even an attempt at a photograph but I thought the creature design was cool.
Is anybody else seeing this for Darren’s post now?
works for me without any login or registration. may be a country legal issue, like how apparently imgur doesn’t work in great Britain now? Where are you from?
Well, I was born in Connecticut but now I live in Virginia.
They still show up in Firefox (apparently because they are cached) but don’t work in Chrome or Edge now. I’ve ran into Imgbb censoring images before. They seem to have an AI looking at them.
I think you aren’t looking at the image close enough?
Sorry, that was unclear. I see the giant hay monster. It’s just that he’s posing with the men happily rather than being sort of hidden and blurry in the background somewhere. Compare it to the other image where the shadow monster is lurking, half hidden, looking at a person who is unaware of it.
The old timey shot looks more like a carnival novelty than a cryptid photo. Which is funny, because the one guy has a rifle so it looks like it’s designed to be a hunting photo, like a hunter is showing off a giant buck he killed, except the cryptid is posing with them like they’re posing for a fun group photo. Apparently the cryptid was part of the hunting party. Maybe they teamed up to hunt down an even weirder cryptid.
This is absolutely the worst result I’ve gotten out of Flash Omni. I get 3 free 10 second videos every day out of gemini in addition to all my flow credits so I often spend them on silly things like making videos of my friends’ and family’s pets.
The prompt for this one was: “let’s draw [cat 1] as a major league baseball pitcher and [cat 2] as the batter. [Cat 1] focuses and throws a hard pitch, but [cat 2] hits it out of the park for a home run.” and the result is all kinds of terrible. Though I must give it credit – the character preservation system is fantastic here. Those look exactly like humanoid renditions of her cats. It’s only the physics and action (and shot to shot consistency) that’s terrible, the rest of the scene (stadium, characters) is excellent.
Let’s see:
1: Wrong stitching pattern on the ball in flight.
2: The ball bounces before it’s hit (I think that’d be a legal hit, but it means the pitcher is pretty terrible).
3: The ball spontaneously shrinks just before it’s hit.
4: The ball spontaneously vanishes just after it’s hit.
5: The batter runs for the pitcher’s mound, instead of first base.
6: Everything is slowed way down.
I wonder if maybe there’s an overdose of cricket in the training data? That might account for both the bounced pitch and the direction of running.
Oh, and also:
7: The catcher is also a cat, but everyone else (fielders, ump, fans) are all humans. Why just one other cat?
8: In the close-up of the hit, the catcher is replaced by a human catcher in the dugout, and the ump is gone. In the running shot, both are gone.
I actually made a mistake and I should confess - I accidentally wrote “battery” when I meant to write “batter” during the prompt. I figured… it understood what I meant since it displayed her correctly as the batter, but then I thought maybe there was some sports-adjacent use of the term that might’ve confused it. I thought - perhaps “battery” may be why cat 2 is charging the mound at the end, intending to do the pitcher harm.
You’re right that it looks more like cricket. The slow motion part I think is forgivable/an aesthetic choice – if it’s creating a cinematic sports scene then slow motion is common in that visual language.
But yes, temporal consistency and awareness of the orientation of characters between cuts are definitely flash omni’s weak points right now. On the strong point, the character preservation is genuinely great for the named/specified characters (possibly top notch) and it really does integrate multiple sources of information – you can do video to video edits and then use image references for what you want to insert and you could insert your own soundtrack too (I haven’t tested this) which would mean that you’re using 3 input sources in addition to the prompt to shape the output video.
I think it was pretty clearly undercooked because they forced it to release for Google I/O because that’s exactly the sort of bad managerial decision that rules the tech world, but I think it’ll be really good in 6 months. If it’s cheaper than Veo 3.1 (and I’m not sure if it is or not – I suspect its tendency to cut scenes into multiple shots is a cost saving measure) it may become the default frontier video generator.
sometimes the AI gets it wrong.
Huh, I wonder if bowstrings are something that the AIs are bad at in general. One of my D&D character portraits had half the bowstring, from the top to the arrow but not from the arrow to the bottom.
Take another look.
That’s what I see 12 hours after @Darren_Garrison’s post: imgbb placeholders talking about content upload failure when I’m trying to see / download something, not upload anything.
I’m on Win-11 & Edge in the USA. I do not have an imgbb account.
FWIW, I get the same failure image if I extract the urls from his post and paste them directly into a fresh browser tab’s address bar, thereby bypassing Discourse. So whatever is wrong, it’s not the board’s fault.
FWIW I see them in firefox, which I viewed them 12 hours ago in, but having opened up a fresh Brave window and a fresh edge window, I get the error messages. So it may be a Chromium thing.
Edit: No, that’s not right – I opened up a fresh firefox profile with no cache and it also gets the error messages. So maybe it worked briefly, my firefox cached the working version, and then it stopped working.
You would think in 2026 image hosting would be a solved problem.
Ah, OK, that wasn’t as clear on the smaller version, but opening it in a new window did make it obvious.
And the post where I showed my archer with the half-bowstring.
Like I mentioned, I think an AI at Imagebb looked at the images, saw nudity, and decided to delete them. Once before I had an image be refused there (that didn’t have anything that looked like nudity, but I forget what it was).
Not your fault of course, but it would sure be nice if ImageBB would render a replacement image that was an accurate error message, not a completely irrelevant and misleading error message. Idjits. This too, meaningful error messages however delivered, ought to be a solved problem in 2026.