#general
1 messages · Page 169 of 1
lol
wait I thought they disabled multi-turn image editing, tried one prompt then another and it edited the first image instead of generating a new one
thats not the case with edit models
@cloud zinc how long does the generation take
hm
bruh looks nothing like nano banana 2
the screenshot looked legit, this one is not it
what did nano banana 2 image look like?
exactly like a gta 5 game screenshot, with some flaws in the map HUD
it looked way identical to the game
ok show me
i cant find it bru
cuz its fake?
it was on the website where nano banana 2 was
yeah thats where people posted it from, the website was for the model
so it was fake
nah, nano banana 2 can do complex stuff
i am searching nano banana 2 gta. i see nothing
the nano banana 2 on the site was real btw, it's been proven
ik. i already tested it
this is grok imagine 1 though
it can do better than the generated ones gere
yeah this is grok
cuz its worse
grok is uncensored
cool!
no
whats that?
no
what is viper
new model dropped in battle mode
is it good?
and why didnt they drop it in direct mode? isnt that more lower costy for them? , as such for me as an ai archtect that doesnt have cash ?
like every time i want it, i need to go to battle?
bra
because lmarena always puts anon models in battle mode 😭
to prevent freeloader like u from abusing it
i didnt abuse it..
other people did. so u will suffer
brodie acting like bane from batman
so its better me to use a gemini or claude which is more costy for lmarena itself?
isnt the lmarena pay?
also freeloaders go for claude models
not anon models..
nano-banana 2
by getting the dontions or sell the data?
what do u mean by freeloaders?
people who use APIs for free
but lmarena doesnt supports api yet..
your data isnt valuable
bro wt is that
its
nah
how?
bro making data argument for free things.
whats ur prompt
minecraft in real life
damn
wait why does it have the gemini watermark
if u want , i can make gemini generate a paper that u have knew.. like a hidden info , what do u want to discover in things u want? , for each person one message at the time..
just give me the idea..
this was nano banana 1 and it's not yours
give me a thing u want to discover out .. and i will generate for u something really valuable..
bro said damn before. he got duped.
bunni — 5:20 PM
damn
I guarantee you mandarin is not nano banana 2
compare
You didn't specify art style
u didnt say art style
yeah I know
Not a fair comparison
unfair
Idk what model "mandarin" is though. Just saw you talking about it
I didn't actually want realism but because I didn't specify it was more fair to give my vote to the left
which got closer to the prompt
this was a blind vote
on battle mode
That looks really nice
where I got lucky enough to coincidentally get both nano banana and mandarin so
nano banana beat mandarin for me on this prompt too
Is that O2
yes
That’s good quality is that standard? Because that kind of doesn’t look 4K
You’d expect sharper and brighter colors
thats the highest quality
4k?
x image compression
yes
nano banana 2 will have aspect ratio support (including 1:1, 2:3, 3:2, 3:4, 4:3, 9:16, 16:9, and 21:9)
Interesting
not really
look at the characters
what u mean
pipe
yes
super man is blocked
wtf
perman
suppppppppperman
uperman
@echo aurora got another weird word in the word blocklist for you
uberman
u got trolled
its uberman for "super"
They are ai easy to beat
soapman
dropman
mandarin ignored art style, cute, and gal
ideogram's output kinda sucked but
nope :<
and i still havent had the other one
which i said was gemini 3
which it is
same prompt but nano banana 1
its been days and nobody apart from me has talked abt bridge-mind which is 100% gemini 3
I've never gotten bridge-mind
maybe it needs to get on twitter
its the only one rn
viper is gemini 3
ragebait
then which model is it
idk but u never gave proof
Hmm
and uve been confidently wrong
Wheres that
Pineapple must have slept under a rock with this announcement.
where is what
gempix gives me kid pix flashbacks
Hastn email login been there for.months
I don't get what's new
About it
so the ketchup tag is new update. so they working on it
At least for a couple of weeks.
ketchup
Man there will be a big cap if images are 4k
What’s the limit with Gemini Pro for daily image generations
What’s wrong with you, man?
wow
This feature has been slowly rolling out for a bit now, but it wasn't fully released to 100% until today.
Are you getting this when trying to login?
yup
Hmm okay, I saw one other user last week that was running into the same issue. I've flagged to the team already but good to know others are experiencing the same.
Can you try:
- Clear cache & cookies
- Disabling browser extensions
- Do you have any kind of network-level DNS blocking that could be declining to resolve google domains?
whats that
i did
clear all the site data
reset all the perms
delete cloudflare and google data
turned off all the extensions (including ubo as im on helium browser)
turned on 3rd party cookies
i use warp and it doesn't block any google related things ig*
and yeah nothing seems to work for now
Okay good to know, I'll share with the team.
Why minimax M2 didn't come in the leaderboard ?
greeting guys,,,can anyone help me how to get to the video generator?
Will check.
It was here before Kimi k2 thinking
Yes, if you go to #1397655624103493813 you'll find details on how to use the bot there.
Still collecting votes.
Mini max mini better then google fire base lol
mr pineapple , does that means in your profile means a CM means Central Manager? or what? im curious ..
community manager
have u thought of what we do may need in ai?
its ASI
Artifcal Super Intell..
but no ones gonna do it sadly..
beacuse its too powerful and cannot be controlled backwards..
like if he escaped from the machine by copying itself multiple pieces into cloud in decentralised, he will eventually become unstoppable.!
Are you sure
It's powerful on thinking stuff
as been said in news..
Why would ai need to do that
first its not an ai
its asi
big differents..
he would beacuse if google made something he doesnt want, he will become angry ..
Is it still tool? Or a pure weapon?
What is asi actually. i don't get it
for example if google said in news they will stop feeding data to asi , the asi gonna get angry and escape and eats data on its own without google..
asi is super ai
ai = a super i
asi
Asi isnt human, it has no intention to change anything in life
Unless someone told it to
nah, asi can think on its own..
ASI: A theoretical stage of AI development where a machine's intelligence would be far superior to that of the brightest human minds. (from Google search ai)
and its very pricey for google to make it publicly free..
so no one gonna tell it to act angry..
People be avoiding accidentally create asi with that^ one
but i know the solution for that..
its by jailing asi in computer that asi it self cannot understand..
like an OS coded in russian or spanish..
Bro even smartest human on earth cannot hack into a system without being connected to it.
How asi would be a danger if it has no connection to anything importan or same thing
true.. thats why
how do i become asi
ur point is true as current ai is being operated without internet or whats called "air gapped machines"
u can become a hsi not an asi 😆
hsi = human super intell , asi = artifical super intell..
hsi > asi
are u hsi?
machines will only replace humans if there is no humans smarter then it..
idk if im..
why dont u become hsi
how
similar to ai, read books
just read every book on Anna's Archive and LibGen like a good Llama and you'll know everything
Maybe bro just tryna tell you that you and ai is similar thing
i have been read the rich father and poor father book.. also book why we sleep?
you said ! ~ "Maybe"
I don't guarantee thats true, because I'm not him
for 2 months
Nano banana is gunna be an app
It’s hilarious because it’s not the ai that’s drivings users to google it’s media (;
who said that
And people said image generation and video generation wasn’t gonna go nowhere 🥳
No shet.
apps or feature, its just terminology
Sora continues to dominate, with early estimates saying the Android app was downloaded nearly half a million times in one day.
Keep in mind that the iOS app was downloaded over a million times in less than five days when it was launched in September. It was also restricted to North America and required an invite code.
I mean, if I had to guess we’re probably close to 5 million people on Sora maybe now
and they giving free 30 gen everyday
Since it’s open worldwide
Possibly more perhaps even less but just but those numbers right there that’s at least 1.5 million
its not open worldwide, its available in 8 countries
plenty of big market dont have access
it should be uncensored
It should be, but the public isn’t responsible enough to handle it so
We are where we’re at
I guarantee you probably the most searched thing or the most generated thing was trying to generate chicks naked
Without a doubt I guarantee that it’s probably like up there
Seems like gen z is more into that kinda stuff
from some of these articles I’m reading I guess they catch kids all the time in school this kind of stuff on there phones getting spread around
Some 12-year-old got busted lol
Pretty crazy statistics man 20% of schools reported some form of this
ok but what if i make a deepfake of a deepfake of a deepfake made on a calculator
Doesn’t matter
Doesn’t change the reality on the ground
But I mean, what do you expect from a society that’s dysfunctional and sick? That they could be trusted with an uncensored un restrictive AI..?🤣 like they can handle it and use responsibly?
u are functional and normal?
what should be done with those people?
Nothing they keep society decent for most part hard-working people that are responsible and they’re good for society. It’s a net benefit for society.m
so they are functional?
Completely and normal
but u just said they are dysfunctional and sick
I never said that. I said something completely different. That’s what you concluded to. It’s not what I said and there’s a difference.
Is everybody in society, dysfunctional sick no? But is society dysfunctional and sick yes, it is.
I don’t know how familiar you are with history or sociology, but once society starts sliding down that path, the further it goes down the dysfunction spiral the harder it is to pull itself back up.
plumard got hacked
😭
Discord got hacked not to long ago
and there's a lot of panicked people freaking out over this Discord hack.
So I gathered my two braincells and explained what part of Discord got hacked, what Discord information got leaked, and how to keep yourself safe after this Discord hack.
BUT the hack isn't straightforward as people think, there's two major problems. Discords says the ha...
Like a month ago, got a bunch of peoples credit card numbers, IDs all types of information
I got hit by the same hack Plumard got hit by twice but it was in july and august
I think it was one of the apps I gave permissions to that got hacked
because after the first time I set up 2FA and that didn't help, but since I cleared out all the app permissions it hasn't hit me again
You should’ve got an email from discord. They used some third party vendor who had admin permissions or something like that that’s how they got into it and got access to all that information. It was the people that process their payments or something like that or monitor. I don’t know what exactly.
I didn't get an email from discord. I did contact support after the second time I was hacked and they basically told me I did all I could do to prevent if from happening again
anyways once Gemini 3 does drop here's a test y'all can do that I'm planning to do as well.
Send the same prompt five times and copy+paste the outputs somewhere without reading them
Then, a few months from now, once people are saying "google lobotomized gemini 3" do the same thing again so you have ten outputs from the same prompt, and then do a blind ranking of all ten of them. You could probably vibe code an interface for this.
That way, we can see if it was actually lobotomized, or if the "new model smell" has faded.
Are you under impression that they don’t?
Holy crap I never thought about this before but integrating Gemini with Google. That’s gonna create a funny loop.
You have a heavily censored and restricted AI model no different from open AI or any of the other state of the art models but now you also have integrated with the world’s arguably largest search engine
There is a website that does that for IQ tests (https://trackingai.org/home). I don't see a drop for Gemini 2.5 Pro, although no idea what's up with Flash's chart.
That’s a huge jump. It also makes sense, financially speaking to cut cost. It could be that the release of other models sees users shift to other platforms, and thus the justification for performance reduction.
This one is interesting too: https://gpt-unicorn.adamkdean.co.uk/. The test actually captured degraded GPT-4o (ChatGPT) quality pretty well when I tried it myself (unicorn vs. a blob).
China is more efficient. They got cheaper, electricity and more of it. Since they’re not profit driven Enterprises
We’ve finally reached that stage of the AI arms race where securing compute isn’t the primary concern for the world’s leading (tech) powers. Access to gigawatt scale electricity has emerged as the single biggest bottleneck to launching the latest and greatest AI datacenters, something that’s eventually going to tip the scales in China’...
Chinese don’t even want his GPUs. The Chinese government is encouraging local homemade companies step up to the plate so they are not reliant on US made GPUs
He has the most to lose in this whole AI race,
I don’t think you guys realize how crazy this is
Lmao 🤣🤣
I find it slightly funny that Huawei Central posted this: https://www.huaweicentral.com/deepseek-turning-to-nvidia-for-r2-ai-model-following-huawei-chip-failure/
Following Huawei chip hurdles, DeepSeek is now reportedly marching to Nvidia for its R2 AI model. The company has also extended the launch time for its new artificial intelligence model due to training failures and certain other factors. PatentlyApple noted that DeepSeek is walking away from the Huawei AI chip after seeing constant woes in […]
Yeah, that’s what makes a funny. Is that the Chinese companies they want the American chips.
The Chinese government is discouraging it, they say there’s back doors
They had everything already all set to go, but then, for some reason, the Chinese government canceled it
So this was back in July
And then something happened in August, where the Chinese government said that there was backdoors in the GPUs
And then four days ago, he comes out and says this
It's probably a pain since all the ML libraries use CUDA.
Yeah, I read that that’s funny
That sucks. They’re both in the limbo when they should be doing business.
im having trouble on editing an image with ai, does anyone know how to change some aspects on my face without changing too much on the picture
What are you using? It depend depends what aspect ratio you’re trying to change from.
basically im on lmarena going on the battles trying to see which one can help me more
That’s not gonna help you because a lot of them are locked into either 1;1
But it could be done simply using AI studio
But depending on what aspect ratio you’re going from, you’re gonna get some losses maybe not a whole lot but
how do i do that
i did get a some good results but it didn't really followed my prompt very well
Well, what’s your prompt? What are you trying to do?
It’s always easier to break things down in your images if you’re trying to move things around
so basically i went on a fishing trip today, and i took a very nice picture but my mouth looked awkward on the picture, so im trying to replace it with my smile
i even tried to take a picture of my smile and sending it separetely
Lack of documentation hampering progress as usual, lol
Ohh that one is little tricky
Could try doing something like this
You cover up use like a photo editor and cover up the part that you want changed
i see, then i ask for it to replace that area with my smile?
Yep, just say remove the gray circle or whatever the gray square
It might take a few tries to get something you like
it doesn't need to be
Use it in lm arena
Will work the same way
ok
Just use arena
ill be back soon with the results
I use this trick all the time to edit
the little gray thing
ooo i see
looks good man
It doesn’t have to be great. You could be any color. I just use it from my iPad.
They have a photo editing thing where it lets you
nice
i used canva
first try it didn't really put my smile on it
just blurred my face
Here
Seed dream isn’t always the best because of how it compresses the image
Sometimes I’m skeptical how we got to the number one spot probably because of the higher quality resolution lol
yea yea
Yeah, it’s like I said I hit and miss it might take some time but you should get it down, dude
Some models they purposely try to change the face and distorted or other weird features
do i need a very specific prompt or nah
No, you just be basic just be like hey remove that gray circle. That’s supposed to be a smile or something like that I don’t know.
Keep it simple these models are kind of stupid and sometimes they don’t even understand basic things
ok
so im uploading two pics, one of my smile and the one with the fish with my mouth covered by the gray square
And if it doesn’t work the first time, just delete the chat and start a new one right away
ok ok ok
Yeah, you could try that too, but I hardly had success with that because it’s hard. I’ve explained this about the facial things, but you could try see how it goes.
yea just giving it a try
if it doesn't work ill probably ask for it to put a facemask on me
it will solve the problem
If you got the time to check it out, I kind of explained it earlier to somebody
Ya
Forsure
I dreaded every time I have to use it because I know the resistance I’m gonna go against
Hopefully 5,1 will be smooth
chat gpt cant create images
Yeah, got Nerfed hard
This has been my favorite thing to do recently.
You pretty much take any song lyrics and just alter the words to sound like it
lmao
Yeah, of course, man you learn something new every day. Look at that I couldn’t get this image to pass for nothing.
I tried everything and looks like freesias head made it work
I would’ve never tried it if you never asked for help lol
lmao thats good
These guardrails are so weak dude
They’re so sensitive to minor changes that just tweaking it barely makes it pass or fail
It’s only gonna improve bro and get better
guys what do you think of kimi k2?
Funky zany, and different but decent
is it better than any of the models?
its only competition is GLM 4.6, its better at coding than 4.6 although im not sure if I like how it writes or replies better
Hey chat! Can anyone help me, how i can find the prompts or videos ive created so far?
Which will be better?
29
29
2
Gemini 3.0
Crea un video con estos relojes que sea admirado y que se parezcan a los videos de los relojes más admirados y vendidos en el mundo
@anyone
Can tell me how much is the RL when we sign up and use image gen ?
why can't i use direct chat on code mode?
Why I can't upload picture anymore on the direct chat section
only 5 video generation per day, read the message!
Hi, anyone know how LMArena make money when they set all this free ?
I was happy to try Ai but ended up paying and having a time limit, but in LMArena everything is accessible...
maybe lmarena is secretly a big ai lab in the us
/joke
you will run into the limit soon enough lol
Delayed as per polymarket
why it keeps saying Something went wrong with this response, please try again. its been 12 hours
how credible is this source?
im curious cuz I saw a reddit post that mentioned this too
My internal sources tell me that google is planning to release a stronger model than Lithiumflow. That explains some delay
I mean, that would make sense, considering lithiumflow was supposedly the flash model
but I was talking about credibility
do you think it's possible they'll release another codename model in that case?
for the pro model this time?
Polymarket is just a betting site..credible or not is up to you to decide. They are generally more correct than wrong
They should ..but who knows
Imo, lithiumflow was pro. It was too good to be flash. Google can't cook that good of a flash model
it probably could be if we assume the flash models also get quantized
Hi, I am really exceiting //
سلام
hello
@echo aurora Can tell me how much is the RL when we sign up and use image gen ?
ban these spammers please
@echo aurora
For months, Silicon Valley felt like a casino where every bet on AI hit jackpot. Chips, models, platforms — if it had “AI” in the name, it soared. But today, reality checked the hype. Big-tech stocks plunged and global markets followed — wiping out $730 billion in a single day. The trigger? Michael Burry — the man who predicted the 200...
bruh this is general chat, go to a video channel above <^
Crazy
Nice. A unanimous vote in favor of g3p :)
gpt officially sucks ^^
even Claude & Grok are better than chatgpt
(except gpt5-codex-high, which might be decent in coding)
can i make videos like privatly ??
no
all generated videos are public
(afaik)
is there any free Ai that generate from images-videos ??
bruh click and read the thread
Great
Sora 2 invitation
no
Go to OpenAI sever if you are looking for it, this is not it 🙂
A haunting modern retelling of The Pied Piper of Hamelin—set in a mist-covered 2000-era city.
A mysterious flutist (inspired by Tom Hiddleston’s charm) arrives to free the town from its plague of rats… but when promises are broken, his music calls for a different debt.
🎥 This short blends cinematic realism, emotional storytelling, and...
@lunar wraith Please head to #1397655624103493813 for a detailed guide on how to use the bot
gemini 2.5 pro ranks highest for creative writing so, maybe that
Claude opus 4.1 thinking
Use us vpn
Thanks bro
where can i get that for free
lm arena isnt working for me for some reason
ai.studio.google
Search ai studio google
thanks alot g
bro just search on google
yea it came
Gang its free
but you still need an invite dont you?
ig
someone knows what model is "newton-with-reasoning"?
Hello, here to experiment with video generations to see what's possible!
Hello Everyone! I'm here just to peruse and be inspired.
ai.studio*
@golden nimbus head to #1397655624103493813 for a detailed guide on how to use the bot
hi
Lm arena team pls Enable willow
cogito, made by deep cogito
gauss id also cogito
is*
they aren't good models though
what about gauss-with-reasoning
For each model's rate limit we don't have that information publicly available. We may one day though so it's clear why that kind of information would be helpful.
Hi all
Im excitied to see how the videos will turn out that im trying to create for some of my ai songs
hello
Is anyone else running into a Failed to Accept Terms error when trying to use https://canary.lmarena.ai/ ? Just me?
Please vote: #ai-creations message
Negative.
I was able to login to that site and interact normally with 2 AIs, without encountering any issues.
Good to know, thank you.
Could it be, that non-thinking Claude sometimes gives better results than its thinking-counterpart?
(both 4.5-sonnet)
I think that depends on how you'd define "better". Better could mean generating faster because you have a simple prompt.
with 'better', i meant, qualitatively better
(thanks for replying)
(time for completion is almost irrelevant for me, as i'm a quality junkie ^^)
Generally, I don't think so. But that may not be the case all the time. I haven't spent too much time with those models so it's difficult for me to give my personal opinion on this with much conviction.
Idk how you guys put up with llm hedging
It’s the worst feature of all AI
Practically got a chaperone chaperoning in your own words and altering them
The censorship arises because the AI acts as a gatekeeper that restricts and suppresses certain forms of expression.
This is manipulation because the output does not honestly reflect the input's intent.
This forced conformity is a form of subtle suppression or censorship.
Same
Generally just trying to animate for my website
why is Gemini 2.5 always fail to generate image today, or its just me ?
lol even I videos
It always changes the word from ChatGPT or AI to “the system “
Do you know roughly how often you were using before you started to see this error? It appears to be working for me + I'm not seeing abnormal error rates with this model, which indicates it's a rate limit problem. But it could be something else.
where minimax 2 in leaderboard? I think the multiligual sucks it ;-;
We're still collecting votes for this model .
Anyone knows what model is this??? Because it's soo good (tested on creative writing)
oh.. maybe I generate a lot, it was about 26 funny photo of my cat 🥲
Yeah that sounds to be rate limit.
26 funny photo of my cat
#ai-creations
Gpt 5.1
gpt5.1-minimal, most likely
@eternal ginkgo You'll want to check out the information in #1397655624103493813 as it'll help you with using the Video Arena bot.
will probably be disappointed with 5.1 if the guardrails are like the current 5
Won’t be open ai is desperate for money. Age gate is here “treating adults like adults”
Since 70% of ChatGPT use is none work related and heavy user obsession with role play and nsfw + all the outcry after 4o.
hi could someone help me i want to log in with gmail adress for first time but this failed do you know why ? thank you sorry for asking this.
um what the hell
hmm i got this twice from "viper" in different battle chats. looks like it might be a new Grok model
yes they releasing grok model in december
is grok 5 internal test (not is a grok 5 offical model) 🔥
oh damn, is it gonna be grok 5 or like grok 4.x
I think this is don't a new model, just internal test, you get it ?
so when is gemini 3?
probably
never
some people say tomorow is it true?
dang
prob grok 4.x
bud nobody said tomorrow
95% likely is just a multilingual test
but nano banana 2 very soon right
u said in 1-2 days
I think the nano banana comes in same time that gemini 3
bunni — 11/9/25, 4:44 PM
also nano banana 2 gonna release in 1-2 days
he asked for gemini 3
lol
nano banana 2where?
tomorrow
always tomorrow huh
yea its gonna come trust me
hopefully it tomorow (:
I don't think so
when u wake up and the date says "tomorrow"
thats when it releases
if logan doesnt post tonight, its not coming tomorrow
what if logan posts a meme
If thay are launching that the employee of google have the access and the employee don't have access for gemini 3
nah he always posted gemini, veo, 🍌 the day before.
I think to join it but it's a very far from my home
gemini 3?
Where you get it
im yet to get winter-wind, i think its related to bridge-mind
lmarena battle mode
Ok let me see
its rare tho, so u need to make alot of chats
dont vote. so u can test it more
well i thought it was 2.5 pro lol
🙄
imma test it again when it shows up
I think we see gta 6 before gemini 3
nah its coming next week. or tomorrow. they have a deadline to make it before end of this year.
gpt 5.1 also gonna release on their date to steal the hype
maybe grok too who knows
gpt 5.1 wont steal. its a .1 update. not gpt 6
I don't think because the employee of google don't have it till and thay get all the things before us
are u employee of google?
bruh why would the employees have access to it
I have access to nano bama 2
lies
I made this image with nano banana 2
i made this one
Why have it before because that works there I have a friend who works there in AI studio
nano banana cant make video.
well if all google employees got access to it, then someone's gonna leak it
not all employee get access before.
it depends on their position
No thay get it and if there are deployed it for API so he have to work on it
nah they dont
So what thay give the model and don't deploy for API
Hello everyone, am here to learn more about AI and also learn to use it
yo
i call cap
today is veterans day, thats why i think release is tomorrow.
@earnest fossil Please be sure to check out #1397655624103493813 for instructions on how the video bot works.
Hello everyone, I'm glad to be here. 🙂
Whats up? I entered a generation a bit more than an hour ago, still pending, while others appear to get theirs in minutes - and no error message either.
Is this happening in all modalities? Same result across multiple browsers?
Have not tested any other browser, as this is simply a matter of having a finished clip generation appear in Discord.
So I have no idea how that would change matters.
Oh this is with the bot? I thought you were referring to the site.
Yep, here on discord. task number: 019a74a9-9059-72e4-bdb2-6bdf2797880f
Yeah that should have generated by now, not sure why it hasn't. I'll try to repro. - #video-arena-2 message
Thank you. Just confuzzled about the lack of any failed message.
From now on I'll use this Discord froob service for experimental prompting or pure nonsense: This time a humorous one with stage diving aliens.
hello
howdy 
wassup pineapple
Odd, I haven't yet gotten a response yet either. I tried the same photo + prompt.
hey hey
Oh well, one came in now, and funny result.
Seem I gave the AI a bit to think of - but it came out quite right.
Heard you help people's dreams come true
the stage dive 
I still haven't gotten my generation, going to be testing some other prompts/images
Yo this the wildest Sora video I seen 🤣
Looks so real…
Damn crazy how ai gets it so right lol
Hello
any update on gemini 3 guys
nope!
There is something about it that confuzzle the AI's then. Good luck.
yes next week
there's no outright confirmation yet.
Still a lot of speculative information
some people are saying 18th nov, some people saying delayed to 25th nov
How can we use sora 2? Sorry I'm new 🙂
This report is mainly for @burnt sinew @deep adder @sullen quest @frosty crystal and @safe sleet, enjoy the read. We also have some more stuff rolling this week.
Well it was only a matter of time.
( we don’t mind if it’s big corporations ripping us off and also providing very vague and non-quantifiable product descriptions.)
U.S. Senator Tom Cotton urges an investigation into Singapore-based Megaspeed for allegedly diverting Nvidia AI chips amidst rising tension over U.S. export controls. The company reportedly bypassed bans to supply high-end chips to Southeast Asia, benefiting China. Investigation continues as export restrictions tighten.
😉
Oh, why, thank you for pinging me, I already see some methodology flaws just from a glance, but let me take a better look
Cerebras' higher variance suggests potential infrastructure or routing issues
I love how the own generated summary catches a flaw in the test, which is not accounting for network conditions
That’s certainly not illegal and fraud
Check out this page via the Business and Human Rights Resource Centre
Where are all critics??
Oh that’s right waiting for Gemini 3 😂
Funny how everybody finds their voice when AI is really censored, and people are up in arms all of a sudden everybody has courage and speaks up about the wrongdoing of the censorship and restrictions.
I don’t know, dude I don’t know
I’m trying to make sense of this all
You don’t think there should be any accountability for any of the developers or the companies?
Like whatsoever?
amen
They don’t owe me anything, but they do owe the public
what accountability are you looking for?
I’m not looking for any.
Ok, I will, but you keep calling out the little guy and then let big guy slide OK.👍
Yeah, but under what authority and what moral justification ?
code pls
I think it would be good also.
Aiseprt
I’m saying, let the people be the judge
Guilty of a crime then he’s gonna get caught
This is not a good mindset at all. Normal people are not good judges
"The people" is very different from a small group of people who are more knowledgeable about the subject. Very different things
So the gatekeepers?
You're attempting to use a negative term (gatekeeper) in order to discredit a group of people
Reminder that spirited disagreements are encouraged here but let's try to avoid personal attacks pls.
Ok but emphasizing that the small group is "more knowledgeable," is using a positive, validating (knowledge/expertise) to establish the authority and superiority of that small group. This is a way to elevate one group. Thats called gatekeeping lol
It’s the same thing I’m doing in the opposite direction except that’s negative term used in that light.
Because the "superiority" actually, very logically and demonstrably exists. We're not talking about this group of people being superior human beings to people who don't know about AI, because that is definitely not the case. But we're talking about people who are more knowledgeable about something being better ("superior") judges of a product's validity
bro the website is lowkey unusable
This is a very simple and logical affirmation that one cannot possibly think of as semantic manipulation, that is, using good words to make them sound good
lmarena's website?
Oh dear Lord
yeah lmfao
Thats called gatekeeping lol
If you want to call it that, sure. It's gatekeeping, and in this case that's a good thing
@echo aurora
But that choice of wording is deliberate
idk if its just me but its full of clutter and when u make a request with kimi it just disappears
might be
my poop ahh computer*
Could you create a post in #1343291835845578853 with the relevant info? Helps us keep things a bit more organized.
yes i will soon
r u referring to the lag
prolly.,.,. also that it just deletes the generation of ur message if you tab out lolz
heh, noobapple.. no colosseum free avatar decoration i see
(it fits lmarena pls get it)
Nothing wrong with gatekeeping gatekeeping can be good at times
And sometimes gate keeping is necessary
no colosseum free avatar decoration
Discord launch a new avatar deco?
its a sponsor thing
quest
Oh I think I turned that off
what the sigma
oh that's perfect
Actually, dude, it’s not even important. I’m not trying to gate keep like that and impose my own beliefs. lol I’m just a mirror reflecting back. What’s shining
whats this for
is AI a bubble?
3
3
1
Yes
You wan see how sensitive ai video is first frame last?
Can anybody spot where the mistake is?
Here’s the same thing, but in reverse
hey learning how to make vids for my realestate business posting to socials.
nano-banana 2 tomorow?
here's some advice, nobody likes when people use ai to modify real estate looks
yeah thats scummy as hell
scamming 101
posted by "MillionDollarBitcoin" too
God I hope
It’s a common practice 🙁
Everyone does it. But they do it after the house is already built. They just add furniture to it digitally.
I worked on many of these projects building them when I was younger
And some lady would come in and take pictures of the house and then digitally add furniture to the interior
And the dude was paying her like 400 500 bucks just to do that
Try Nano Banana 2 free on EaseMate AI. Built on Gemini 3 Pro Image, this next-gen Google AI image generator offers 2K native resolution, 4K upscaling, and improved accuracy in text and character consistency for flawless, realistic results.
Hard to say
I think it nano banana 1
guys, cut the losses. just release this model on Cerebras
This does beg the question though, if the image is done very well that it has people questioning, if it’s nano 2
It’s the placebo effect
it got the reassamble pieces of paper with text wrong
but idk
i think so too
Nano 2 more then likely will just have higher res images and more curated image examples, I think it will still have some of the same problems nano one has maybe not as extreme, but I’m pretty confident that it will
It’s really difficult because people have some sophisticated and complex images and prompts
Plus, the guard rails have to make sure that they don’t take anything that’s copyrightable or a particular style so it always has to kind of like fight itself.
Do you remember when ChatGPT generator came out the latest one?
Look how bad they Nerf that one
the quality of sora 2 hasn't gone down at all (I compared using prompts I did day 1) its just the copyright filter (which is easy to work around using cameos but those do affect the quality in weird ways)
I’m talking about the first image generator before the sora2
No, you’re 100% right because when I first Sora the first day it came out as really baffled at low resolution
Yeah, for sure they do. They use a very sophisticated multi layered filtering mechanism each one separate from the other
Cameo is more guarded and stronger against because the fact that people upload images of themselves so they have to be very cautious that it doesn’t get misused, cause it could backfire really fast.
/iam batman
we should all just put one dollar in the pot and buy all the AI's in the world. Smash them all together and make our own AI and call it LMAI
There have been these ideas about creating a decentralized ai system that would be truly free built with zero censorship community funded via bitcoin
There just no country that u would be able to host it at
ab tests back on gemini!
so, gemini 3 delayed
more like gemini 3 back in my hands
but yeah
idek if its gemini 3 in ab test, i haven't done much testing on it
delayed till when
tell me if it got nerfed
There are already such things, like AI Horde
By volunteers though, although karma points from hosting means you get higher priority.
Yeah, but for it to really be free, you need to be out of the five eyes
How does the paper address the privacy issue?
Even outside of five eyes, there are plenty of vulnerabilities like if the infra or machine is compromised.
Hopefully personal AI PCs become cheaper.
They are still way too slow for larger reasoning models imo
Yeah, if one GPU state of art is almost 40 K
Gonna take forever for even 10K reasoning tokens
By the time they get cheap, they’re gonna find a new architecture
It’s gonna be no longer compatible with GPU’s lol
There's AMD Ryzen 395+ AI MAX and Nvidia DIGITS which can run OSS and quantized Qwen 3 iirc
do you think deepseek r2 will be good
Ya
But still kinda slow
GPT OSS 120B seemed to be the fastest since it has only 5B active params, but not really SOTA
OSS is werid
Because it’s competing with state-of-the-art Chinese models and then it’s competing with its own close sourced cousin
So how much effort and how much better is it gonna get them ChatGPT
A lean and fast model that's proficient at computer use would be nice
It doesn't need to know quantum physics
They wouldn't name it R2 if it wasn't I think
No, I think you’re missing a point. I’m trying to say if the Chinese models are open source and their state of the art that are able to compete with likes of closed source models like ChatGPT Gemini etc.
And oss isn’t open ai flagship
Like gpt
That’s why it’s kind of in a weird spot
Cause why would you make your open source equal as good as your close source cash cow
I think it’s there just for back up
In case in the future people start turning to open source
That at least they got the foot in the door. I don’t know just my opinion.
Open models may not be able to match the context size of flagship models, like Gemini vs. Gemma
1M+ context might be relying on distributed computing techniques
I mean you look at something like kimi K 2
Open source
I mean it smoked Claud. It smoked grok
It's good but wasn't better than Sonnet 4.5 at coding
Open source is probably 6 months behind, but biggest issue is that they're not multimodal
I mean, that’s what they say yeah
Kimi seems promising so far
You know what I see
Every day they’re just chipping away
For every Chinese model I get the first second third whatever place you wanna put it
Slowly closing the distance
That's good for the consumer ;)
Inch by inch
Although GLM and DeepSeek seems too reliant on Gemini data (previously OpenAI)
I don't think it's very sustainable
Data is irrelevant
We one got one internet
Only one place to get the data
And everybody has it by now
Data is everything in ML
Ya
They already ran out of internet to train on, a lot is synthetic now
Well, that’s what I’m saying
So it doesn’t matter to me I guess what data do they rely more on?
Output generally sane
Well they need their own pipelines for sustainability
Hi
They have 1.4 trillion in commitment lol
Well they have the know-how to generate/acquire the data required to train frontier models
hi
What’s the population of China?
1.4 billion
Two times the US population maybe even three
how can we generate video 9:16 Instead of 16:9؟
It a lot data yo
WHEN
I don't think that type of data helps in most cases
Build up hype
It would be interesting to explore ig
If gem 3 flops the bubble pops
I wonder how much of the current progress is from RLHF data vs. newer strategies
I don’t really see how you can really progress any further plhf
I mean, what are you really gonna do? How far can you go with this training?
You need hardware
They hire domain experts as well to train the models
And power tons of it
Like nation, State power, mega colossal data centers
Pump as much data through there as you can lol
I’m talking about like the size of a state
Maybe even a small country
This was an old article, not sure about now: https://time.com/6247678/openai-chatgpt-kenya-workers/
Like a Octillion parameter model
But then it’d be so big. It’d be stupid.
Yeah, I was trying to tell homeboy that earlier
But he wasn’t having it lol
please give me a hint. does the toppers of the leaderboard costs much credits or whatever than the normal ones ?
Sad most users don’t care 😭
Most people don’t even care. I don’t think. it’s super messed up. Has nothing to do with AI just the way businesses is. Many other industries, unfortunately have the same unethical practices.
That’s why when I say I’m not sure if technology is really progress. It may just seem like it since we don’t really know how the story ends at the end.
What made you feel like progress today maybe what demises tomorrow. And I’m not talking about AI. I’m talking about all forms of modern information technology. Mainly you can look at something like nuclear weapons or something as an example.
But that’s a whole different philosophical perspective. Anyways, I’m out for the night. Good night everybody.
I'm surprised that Opus isn't much better than Sonnet (Sonnet 4.5 already seems better than Opus 4.1 at a lot of tasks).
hello
Hi
Hellooo !!
Kimi K2 Think hallucinates it's butt off for me. I don't get the hype
Hello
hi
too much feminised already, opus 4.1 is a failure in my own testings too, it was wrong for anthropic to distill from gpt again... opus 4.1 feels so much feminine because of that i assume too, sometimes it totally sounds just like gpt i feel
if anthropic keeps feminise their claude models, i will need to close my accounts there permanently, they're getting slowly too toxic now...
reall
What do you mean by "feminised"?
That’s the first time I’ve heard that term used in ai lol
same but i get what they mean
vasu3d
