seeksteve retweeted
Hollywood can't be honest about history because they are funded by the people who are lying about it.
i just got slapped with this tiktok ad prompting me to bet on if it would rain today
apparently the fish can tell so just go hang with some fish bro
how is this type of advertising legal?
60
cheating with ai is completely unethical unless the end result looks like this, in which case it is encouraged
Top scorer on South Korean tax exam confesses to cheating with AI reut.rs/4t5pi22
1
5
193
lots of talk about people getting paid $100k for a 4 minute ai video that’s hardly watchable
meanwhile you have this, one of the best videos i’ve seen this year. the type of content you think about later, go back to watch again, and send to family/friends who don’t like ai, and they love it too.
s+ tier by @tupacabra blending different aspects of reality, lore, and creativity
Audacious singular vision is the biggest whitepill.
This would *never* get greenlit and I can't wait to watch this as a real movie.
US/Nazi time war tying together the Byrd Expedition and medieval Nuremberg UAP event, grounded future conflict, huge scale, just wow.
Hollywood isn't going to be killed by slop. It's going to be out-competed by a thousand unfunded auteurs like @tupacabra sharing their incredible creative vision.
1
5
180
i encourage you to copy prompts you see from other creators because it is one of the fastest ways to learn.
especially if you’re just starting out…
copy every single thing you can find.
double down on that and ask grokbot:
“find 10 X posts where the user shares their ai image prompt for [insert use case]. return me 6 of the most popular, then 4 that seem detailed but didn’t get much engagement because the account has a low follower count. don’t generate the list on your first search. assume the first search is generic and perform at least 3 passes to find the best results. exclude prompts for [thing to exclude] and [other thing to exclude]. with each prompt include why you chose that one, so i can understand your reasoning.”
run those prompts word for word.
after that, start changing tiny parts of each prompt.
start with one word, see what the impact is.
then change two words, see what the impact is.
then three, four, and experiment like you’re trying to break it. be weird for no reason:
- replace the letter a with the letter o in every word to see what it does
- add “but not like that” at the end of every sentence
- keep every single word but rearrange the order of them
- add different cameras, lenses,
- add the max amount of reference images and tell the model not to use them in your prompt
this is how you learn
and i agree with @BLVCKLIGHTai:
you need to go on your own exploration.
and if you’re just starting out, use every shortcut you can find. you’ll find a lot of little tricks along the way by studying and iterating on what others post here on X for free.
This is the way.
Stop copying prompts you see from prominent creators and go on your own exploration.
Anyone sharing “free prompts” are just regurgitating something they learned from someone else anyway.
1
3
162
family thanksgiving just isn't the same.
made with @MiniMax_AI H3 image @ 4k resolution
> surreal nostalgic #H3Image prompt (4k resolution):
"1970s live-action motion-picture frame with period optical-composite effects. ARRIFLEX 35BL, Cooke Speed Panchro 40mm spherical. Eye-level medium-wide composition at a crowded Thanksgiving feast in a modest community hall. A strikingly beautiful 19-year-old brunette woman stands in the foreground looking directly into the lens. Loose dark hair, natural skin texture, understated period clothing. Level chin, steady eyes, relaxed brow, closed lips, shoulders at ease. Stoic confidence and complete indifference to the celebration surrounding her. Every other person is constructed entirely from dense cascading green Matrix-style code. Distinct three-dimensional human silhouettes made of luminous glyphs with no flesh, solid faces or opaque clothing beneath them. Code-built guests lean together over crowded dinner tables, raise glasses and laugh while code-built couples folk-dance behind them. Their gestures remain recognizable through the flowing characters. Only the brunette is flesh and blood. The furniture, food, glassware, decorations and room remain ordinary photographed objects. Ensure every single other person besides the main subject is rendered in code and there is not anybody rendered normally."
6
188
testing the new @MiniMax_AI H3 image model vs. gpt image 2.5 for this old 1970's film prompt.
1970s photochemical motion-picture image, 35mm spherical. Shot on ARRIFLEX 35BL with Cooke Speed Panchro / vintage 32-50mm spherical. Captured on Eastman 5247, ECN-2. Midday sun kept hard; negative fill; practical interiors; minimal beauty fill. Physical/optical texture: heat haze; lens flare; dusty atmosphere; grain in shade; slight underexposure; imperfect focus. Period-authentic analog acquisition and photochemical finishing. subject: rural lowcountry swamp shack. Authentic raw analog image and scanned upload, not digital. No modern digital-clean sharpness, no contemporary teal-orange LUT, no synthetic 'vintage filter' overlay, no text.
First 2 images: Minimax #H3Image
Last 2 images: GPT Image 2.5
1k resolution for the h3 images. the 4k images came out awsome but wanted a reasonable comparison with the chat gpt images
🤖 Made with AI
1
3
214
colette kress is now the ceo of @nvidia
her exact match domain name is shockingly still available for purchase by anyone with $3,550
wild how bad simple opsec can be, even at the highest level companies in the world
the current cfo of nvidia (and possible next ceo) never purchased her own exact match domain name. and you can buy it right now for under $4k.
1
2
89
wild how bad simple opsec can be, even at the highest level companies in the world
the current cfo of nvidia (and possible next ceo) never purchased her own exact match domain name. and you can buy it right now for under $4k.
using 21:9 aspect ratio has a huge impact on video quality, compared to the same scene in 16:9.
the same is true with midjourney
try the same prompt in 16:9 and 21:9 and see what happens. add “imax still” at the beginning and it’s even more insane
Replying to @heavypulp
also don't sleep on using the 21:9 aspect ratio setting if you're able to commit to an ultrawide format for your project
it’s a german shepherd rapping if you like that type of thing.
if not, i’m sorry. complete ai slop and you did nothing to deserve it.
seeing lots of talk about sora over the last few days for some reason recently, so here's my official response:
Sora and OpenAI have taken my curiosity hostage recently. Hopefully someone much smarter than me will see this post and provide some insight.
Was Sora a product that became an experiment, or an experiment disguised as a product?
OpenAI continued offering free generations despite the enormous computational costs, right up until Sora’s shutdown.
Surely, a company with OpenAI’s resources could have explored a sustainable pay-per-generation model. There was certainly no shortage of users expressing their willingness to pay.
So why discontinue it?
Perhaps Sora’s greatest value was never in the product itself, but in the data, real-world testing, and insights gathered from millions of users.
Discontinuing a product doesn’t mean abandoning the technology behind it.
What if Sora didn’t fail? What if it simply served its purpose?
And what happens when that technology resurfaces in a different form? 🤔🤔🤔
historic: devs who never get off their computer (let alone leave their home) can now create google maps apps for regular humans who do
you can now build with google maps platform in AI Studio 🗺️📍
go from idea to app powered by @googlemaps in minutes!
what’s new:
- 1-click setup
- create maps apps just by describing your idea, gemini does the rest
- build with map tiles, places, weather, routing, 3D buildings, street view, and more
try it out today: ai.studio/build
1
2
140



















