#general
1 messages · Page 128 of 1
Sound like AI help in a snut-shell.
I asked it to elaborate on something it mentioned in the last message, but it started explaining something completely different with the same name
IMHO, they are designed that way, to trap users to buy pro version, hoping for better results.
I would get Pro if the results weren't so bad in the app
2.5 Pro is limited to 5 messages so I assume that's what I'd get if I get the subscription
But it's terrible compared to AI Studio
Work the opposite to that for me, when a service cannot give even a decent answer, generation or AI clip in regular version - I do not see any reason for premium.
Can't rely on that. The app itself is another nutshell, limiting the ability of the models, to run exactly in "corporate safety" guidelines..
I use LMArena versions because they are outside the shell
And their privacy policy is terrible
Indeed, I had not tried some of the ones LMarena got, which brought me here to try.
"We'll train on your conversations unless you disable activity"??
But if you disable activity none of the conversations are saved
Terrible
Ask Gemini App to edit a pic, most of the time your uploaded pic will considered as "minor," even if it's a 50 years old person..
I don't really use pictures, I just want consistent quality and conversation history without privacy issues
Oh they do train from conversations and your images also videos.
I've seen it several times with the SeaArt engine, it fail miserably the first time, get the scene almost right on the second and on third get things just right.
Continue to do it will often result in worse clips - why? Because the AI assume you're not happy with what you request - and start to overact or overreact the result.
Confront it with facts, and in the end, it will reply "I am just a language model."
SeaArt got Deepseek as auto prompt, it's quite good an 'seeing' what the image show - but not always.
And a few times get creative and write an entire novel about it. 😹
🚮
I have at least gotten my privacy issues fixed on the phone and tab - while not on this PC.
So, are they helping? No. They are designed to "lead" users to go inside their "lines of policy." When they failed to deliver a user's request, they'll simply apologize.
<- Got a non google phone, and have de-googled my tab entirely - which only could be done by going radical.
Even Chinese models have better privacy when they are hosted on providers with no logging
Yeah, I'm on Wechat, limited monitoring is no issue for me.
But listening on my mic and logging everything I do is way to much.
You know the Gemini app is bad when even OpenAI has a better privacy policy
Yeah, as said above, I'm not hysterical about my privacy.
But Google have gone too far.
Yeah stuff like voice + video is a no-no
It's almost like they think they're the FBI.
Fried an Barbequed Intelligence
Sautéed intelligence.
I spent hours these last days to chat with them until they do not have answer and simply reply with "I am just a language model." Or "I'm sorry, it's my fault." Only GPT will add "Would you like me to get other options?"
You're already on their radar at the first day you get your Android phone.
Yeah, exactly.
Why the videos has no voice
Hi
It's going to be random if your video gen has sound or not. Not all models have audio capabilities, and it's going to be random which models you're sampled.
How to use elevenlabs for
Free
Information on how to use Video Arena can be found here: #1397655624103493813
I want to create a replica of someone's voice.
Hello there. I am still having so much trouble trying to find my messages. The bot has not DM me since last night and I’m worried that I am doing something wrong with every single prompt. I’m doing. I’m using video arena number one, and the vibe has not DM me at all. Every single time I try to post a prompt with an image, gets completely disappeared becausethe server goes super fast, and my data is literally lost within the flood of prompts and videos.
I'm not sure what we're offering here is what you're looking for.
If you go into the search bar, select mentions, and type in your username you should see all of the generations you've made.
Well, I also need to know if my prompts are even going through. Because the last video that I see, the most recent one, is from last night, and I made a generation prompt from an image I took this morning and I still have not received anything. I need to know if my prompts are even going through or if they’re displaying an error message, but I just can’t find it
Hello
Forgive me, I have a little knowledge of how this works. I wish it was a part of the website and not on the server with thousands of other people. Because then I could have full control over what models I want to use.
It looks like you're not prompting the bot correctly when trying to use with image upload. You'll want to type out /image-to-video if you're looking to use image to video creations. /video is just for text to video.
Well, that’s what I’m doing. That’s exactly what I’m doing. When I tap on an image to video, it brings me up my photo library and then I select the image. And then I tap my prompt now, I remember the last time I tried to do it it said that I reached my daily limit. I thought that whole thing resets at midnight the next day because it’s technically the next morning, but I’m starting to wonder if it’sabout the first time you generate a video and that that’s when it resent.
I think I'm confused as I'm seeing you've been generating videos fine. #video-arena-1 message Are you using a different device now and experiencing something different?
I thought that whole thing resets at midnight the next day
It's going to be 24 hours later
I’m only seeing the most recent video of last night. So, problem is, I am visually impaired, and I’m using the Apple screen reader built-in called voiceover, and tends to jump around places and won’t get me the full list of messages. Even without it, I just have such a hard time finding my stuff. Every time I look in direct messages with the bot, I can only see my prompt from last night, and nothing from the day
It’s literally showing me videos from last night only and nothing’s coming through today. The bot has not let me know that any sort of video is ready since I started using it today.
I’m probably doing something wrong, but I have no idea
I'm only seeing generations from yesterday as well; however, I'm not seeing attempted generations either. I believe you're at your 5 videos per 24 hours limit.
If you try to generate today, what happens?
I reached my limit it said. So I started using it for the first time 2 o’clock yesterday, so I guess that’s when the daily limit resets? For some reason, I was thinking that it was set at midnight and will allow me to continue the next morning or something.
Okay yeah that's the problem then.
Yeah it should reset after 24 hours from your first generation from the previous day.
I didn’t realize it was that limited to only five videos. Dang. Thank you very much for everything, and I’m so sorry, I probably acted so freaking dumb.
hello ned video veo 3
It's no problem at all, no need to apologize in the slightest. It's a bit confusing so completely understandable why you'd have these questions.
Check out #1397655624103493813 for more info on how to use the bot.
why cant i find gpt 5 codex
on lmarena?
real
🤦♂️
💔
Can someone spam ping me once they add stuff to LMArena
hello everyone, I'm here to learn and test the tools here
We post updates to the #announcements channel so be sure to keep an eye on there
yes
why cant i find gpt 5 codex
hello I want to experiment
Go ahead then.
Hi everyone,
I could use some help reviewing my voice prompt. I’m trying to lock in a consistent voice for my character, but I’m running into issues with variation between generations. Could someone take a look at my prompt and give me feedback on what might be causing inconsistencies or where I could tighten it up? Thanks in advance!
Sure.
You'll never get consistent voice unless you provide a sample to the video generator.
Yep 🙁 — but I mean more generally: am I setting up my prompt correctly, and are there opportunities to adjust it with tips? Also, could there be any conflicts in how I’ve written it.
Well, we won't know what's wrong unless you provide it.
It's okay to paste prompts in here.
Show us the prompt. Might be better in #share-prompts
Hi man
Hi man
Hru all of you?
Gregy McGreggerson.
Are you real Greg or just another Greg greg?
Greetings.
Real greg
My old account is hacked
For context: his name is different because that's another account he's using.
Apparently, like he mentioned, his Discord account somehow got hacked.
Yup
He will pay dont worry
I know. I've read here earlier. I just saw many Greg's today.
imagen-4.0-fast-generate-001
Is the worst model ever - when speak about contextual thing.
Ah, alrighty.
I hadn't noticed that he mentioned it before.
gulp
Jep, i shared it in #share-prompts. Thx!
I already contacted discord but he will pay with the same price of mine account
I suppose it might be.
Though, I'm not so sure.
Brother is Shakespeare.
Real
Them shalt deal with thine consequences of confiscating mine account.
I know
In fact I want to wait
But I will act if nothing will happen
Not for me but for Discord security
Makes sense.
I found it better than the 3.0
However, Discord is usually pretty good at getting things done pretty fast.
Problem that he tries to make text even though he can't
Kinda
Have you reported?
They're far better than Road blocks in terms of moderation.
Yes man
Road blocks almost never does anything about reports.
Idk I've never reported no one (except zooph*le servers)
And how fast did it take for them to get banned?
What about Supercell bro 💀
So we mass reported like in 25 and the server was shout down in 1 day after continous limitations
Not sure, considering I don't know the guy or thing.
No is a video game company
Well, that's certainly far better than Roadblocks.
It takes a week for those guys to deal with anything.
Idk but now i am obsessed with this my pfp lol
I can definitely tell.
What you can say with certain?
how old r you
Well, you wouldn't have made it your PFP if you didn't like it.
So I'm pretty certain.
Yeah
An apple a day keeps the Diddy away.
an Apple ❌️
A kid ✅️
Or adult.
Hey, how old are you now?
About the same, 17 years ago..
I'll get punished for my last banana pic..
And I'll be the punisher.
And i will be the teacher pet
Yours was much more menacing..
Dropped mega banana in a crowded market? Oh no..
I'm gonna show you what a gigantic banana really looks like.
I've read "made in China" label there..
Fr
Coze
I didn't feel like removing the watermark
They will not obey that
🤫🧏
🗣️ 🔥 🔥 🔥 💦 💦 💦
Looks like even my 120 years old gramma can catch him easily..
Yesterday he put a cat in his pizza recipe, and now...
Fr
Italian cat 🤌🇮🇹🍝🍕
I am contaminated
Years ago, before Flux, it was my daily experiment to make a picture of kitten or puppy in places they shoudn't be. Kitten in ramen bowl, puppy in candy jar, etc. Lol..
Ah.
Not real
Banana pizza?
No, dont do it
Don't offend my object for italian rituals
Greg will not allow Pineapple pizza 🤣
Fr
Nah, it's obviously a banana pie.
With some strawberry jam.
Obvious as in obvious
War problems: WWIII
Italian probelms: pineapple on pizza
Yes.
Good choice man
And broken spaghetti
🔪🩸
And ketchup on pasta
But no problem with kitten pizza
And today the cat even in banana, in taco, and packed in toy wrap. Oh my ...
Facts 🤓👆
People are complaining about Banani and Soda are not really works well in combining objects in LMArena. I've tested them several times today. And they are just fine for me. How's in Dreamina and Coze?
(It's meant to be read as a joke.)
I am not understanding lol
Heh.
hey people, hi
Greetings, Trap.
Hi, Ulla..
Salt and vi–
Don't
I'm busy sweeping the floor
Ah, right, right.
HI to all!
question since its so much traffic how can i view the video that ai generate me without having to scroll to find it, and how do i know when its done
It'll notify you in your DMs once it's finished.
Greetings.
Oh Ok thank yo so much
My pleasure.
Hello guys.
Beginner beginning
Hi
Greetings.
Hope you enjoy your time here.
Same to you.
I don't know what happened to me today 😞
And I spent hours testing a 219mb model, I don't even know what to say, actually I know, gemini 2.5 pro >> gpt 5 + opus 4.1 both
no, it would be better if I just used the 2.5 pro I have this damn addiction of comparing AI
Well, it's okay to want to compare one thing to something else.
After all, that's how you find out which one you like more.
I should just use 2.5 pro or lfm 2 2.6b when I don't have Internet
Okay, thank you for calming me down I'll try to work today 🥱
My pleasure.
I always like helping people.
A lil' bit of peace, without prompt wall..
I suppose so.
The last 12 hours it's under 700 people online here. Earlier it's always 4-5K..
Yeah.
Hello, my name is Ange, I am exploring the video possibilities of my digital designs
Go to #1397655624103493813 to learn how to generate video on this platform..
I'm a beginner, I'm studying and practicing and asking to watch and learn from you
What would you like to learn? Video generation? Local models? Make better prompts?
hii
Same but gemini 2.5 pro is so bad for medical questions vs gpt high and opus 4.1 16k
Whether in the app or studio or lmarena
Its even worse in the app but still gives bad recommendations vs gpt and claude
Shaking that thang.
tf
It’s peak Jon mud content
I cant believe lm arena lets me play with all thi cool ai
Its not working for me
ok
@echo aurora is it okay if I share the website to my friends and my whole class? lol
Hundreds of people sharing it on various Youtube channels, that's why people keep coming and made this server has more than 100K members..
so theres no harm sharing it to my whole class?
That's what they wanted. Leaderboards need more people to vote for best models..
hello
HI..
Of course!
I should do the same and share it to my friend who likes 4o
Can anyone tell me why I can not upload any image for image editing or character reference?.. I'm unable to upload image.
Is this a bug or just happening with me?
Sadly, seems only happening to you, Dude..
Thnx..to letting me know that.
I also have a problem. Bot can't seem to generate a response
Hi
It's 12.08 PM here now, and I've been comparing image generators since about 5.00 without a problem..
Have you checked the steps as in #1397655624103493813 ?
Happened to me twice so far, when the bot is overloaded..
It's been 4 days since I successfully upload any image.
Every time it's says upload failed for no reason.
I thought arena hv stop the uploading features or something like that.
Is it in here (Discord) or on the website?
Website
I use discord only for video
Try to clear your browser history and cache. I never have upload problem. The only bug I've ever seen was the model failed to generate images..
It's on Desktop/Mobile
Not...here
Not just him, unfortunately.
Hi
I also can't, for whatever reason.
Hey, verify you're a human..
I already did, though.
You must be uploaded those gore banana. That's against their TOS..
"Gore banana"...
Welp
Still happening
It's just fine around here. I'm using a PC and a phone. Tested to upload image reference on both, and they're okay..
Ah.
It's strange that the bug is happening to only some people.
I could send a recording of what I see.
Here you go.
Yeah, it's a strange bug.
I even uploaded Jackie, and it's go through..
How odd.
Google Lab's AI video generation tool, powered by Google DeepMind’s latest model: Veo 2
Are you on PC?
I am, yes.
What browser?
I'm currently using Firefox Nightly. Let me check on Chrome..
Alrighty.
what is happening?
I tested this in Microsoft edge (Don't know why I still have it). And it was fine.
GlitchArena?
Some dudes can't upload images..
What browser gave you error earlier?
upload the website 🤑
If you mean mine with 'Not being able to generate'. Then its Chrome and Edge
Can't generate a response from the Bot
why is gpt-5 so low on the leadderboard?
because it sucks balls, even in its main site
It is what it is. Failed residual AI..
Something went wrong with this response, please try again.
And?
Anyone experiencing the same?
It doesn't work on the canary version.
The regular one works just fine.
bro
Refresh the browser (F5)
Lemme ask. Can't generate no matter how many 'retries'?
Hm?
Yes
meow
You and me, Man
violent loud screaming
I can't continue my....barely disguised bs Creative writing
Anyway...
what
it is NOT down
Erm, what the Skibidi?
is it?
i'm using rn
I just asked a prompt again
You're right. I prefer a parrot anyway..
I'm using 4o rn
Aha.
is seedream that good?
It was 5 days ago nvm
imma test
The verdict has been solved.
But it's still not generating
Speaking of "canary"...
Try doing that on a different version of the site.
what sound does a bird make
If you can give it a good prompt
Usually screeching sounds, contrary to the "tweet tweet" sounds that people always claim some make.
That's it. Parrot can miaw miaw..
no, bark
Try it on this version: https://canary.lmarena.ai.
That's the sound a hunting dog usually makes while chasing one down.
i need 5 second maximum!!!
Yeah.
free is not enough...
For that, you could just use a smaller, faster image generator.
I recommend Imagen for that.
It provides good quality and generates fast.
What's that version used for?
Some kind of beta version, I assume.
i need instant generation and nano-banana 19.25!!!!
This reminds me about a website I've visited. Its image generator instantly generate images while we're typing..
Hmm...
Well, the browser that I'm using can search for that, then provide it here in the chat.
nano-banana 2 thousand*
The name will not nano anymore..
CRAZY
So, perhaps I could have it search for a website that allows users to generate images with nano-banana instantly, then have it provide it to you.
no bro i'm just joking lol
Ah.
Yeah man, I got you.
you're a real one bro
It's my pleasure.
Gpt-3 is PEAK AI performance
Because they had more interaction and training data I suppose
Well, research has shown that GPT-4o actually outperforms GPT-5 in some cases.
What about GPT 1 🤣
gpt-5 is pretty trash, specifically when using the official website
That's very true.
that one peak too lol
perchance?
Y'all wanna know where you can use nano-banana for free, and without rate limits?
Me and Blacx know a site.
True peak. GPT-1-Image, the slowest generator that never be able to read reference image correctly..
creditcardinformationhacking dot com
That is spot on.
Exactly right, Felipo.
hell yeah
Where can I use it?
Actually there are 3 sites where you can fry bananas without limit..
idk bro i'm not an enthusiast of AI, i just use it ngl
but i remember i loved using gpt-3
Apparently, GPT-3 isn't actually deprecated from OpenAI's API either.
The canary site doesn't work either
Ah.
Did you already try on a different browser?
where do you even get this from, you're a legend
Blacx gave it to me.
That site gave its own link to me
I am on edge it's a browser issue?
what is this supposed to mean tho?
Some bugs tend to be, yes.
That dogs are good at pool. 🗣️🔥
my grandpa would think this is real
not joking
I don't blame him.
he would probably ask "is the dog yours?"
Days ago someone said: Now I doubt if those cute pets on Instagram are real
Grandpa did not fight a world war for that...
i actually don't know anymore
Sometimes, some things come without you fighting for them.
The world is an unexpected and random place.
The subrreddit iss under fire because of OpenAI's unplanned decision to divert 4o chats to 5
true
Also the pets
On a scale of 1-10, how would we rate this baddie?
do you guys think OpenAI is digging it's own grave?
Is that ... yours?
It was generated by me, yes.
i don't
That too.
100%
Everyone says it
They'll not digging, they're excavating..
Reasonable.
It tempts me to send a quote from Sonic the Hedgehog 3
Where Ivo Robotnik says Goodbye to Stone
7.5
But its where I say goodbye to OpenAI as I find other Chatbots
i tend to think they will make a comeback
Also fair.
maybe delete gpt-4 and 5, and go back to gpt-3
Since GPT-5 is dead. Ads may be comig
With consecutive failures and zero WOW progress? I was expected GPT-Image-2
<-- Into brunettes
chill bro, maybe Elon Musk will buy openAI just to delete it
<- Into blondies (any works just fine, but mainly blondies).
best future
I like how the letter blocks just spell out "PROJMPT".
Blame Nani Banani for the mispelled word
Could just ask it to correct it.
That's not fair. It's 1 hit generate
OH MY GOD
Well, not all generations are gonna be one hit.
IS THIS REAL GUYS?
Positive
Yes.
OH GOSH
I've talked to ChatGPT earlier. It says "Sorry, I'm just a language model."
And this is the guy who did it.
When i try to talk about philosophy with him, he gives me a whole work-out routine to get me out of "depression"
R.I.P
hi guys
Greetings.
You guys are too rude.. 😭
Nah, we just like giving the mods a hard time (not me, obviously).
I'm ... Innocent as ...
As innocent as a criminal claiming he's innocent.
Banana
Apple
Other places to use Banana and Imagen for free, no rate limit:
https://labs.google/fx/tools/whisk
https://labs.google/fx/tools/flow
Crazy, there's a storm going on and people are still partying outside with loud music
Brazil is crazy
They're rock!
Heh.
I wonder if my grandparents know anything about that.
They're Brazilian after all.
Did they live on Brazil for long?
Brazilians are just built different.
We had a lot of dictatorships back on the 19s
My grandma was born in Portugal then moved to Brazil, but my grandpa was born in Brazil.
They've lived there up until I was born, which was in 2007.
Must be not so many parties when they were there..
Brazilians always find a way to party, i'm serious, even if they get arrested or even killed
Well, they seem to know about the Brazilian "carnaval", which is a festival where Brazilians dress up in different topics and march down the streets of Sao Paulo.
Not only Sao Paulo
the entire country
its CRAZY
That too, yes.
i'm telling you
some areas even have naked people dancing to celebrate indigenous culture
100%
I live on the only state that wants to separate from the country
The one with fruits on female heads?
Sim irmão
What?
Fruits on female heads?
Confusing lol
The festival
Muito bem.
You're right tho, they dress up as literally anything
Isso faz sentido, considerando que você tem um nome Brasileiro.
I'll dress up as nano-banana on the next carnaval 🤑
É só um apelido, no Brasil ninguém se chama de "Felipo", só "Felipe"
Isso é verdade, no entanto, "Felipe" ainda é um nome bastante comum lá.
Is talking another language even allowed? 😭
Nah, but there's only one mod on here.
I think we are committing crimes
I was going to submit warnings but I like you guys, just don't tell my boss
Some others speaking in Hindi, Arabian, Indonesian, Chinese, Korean, and there's no earthquake ever happened because of that..
W mod.
Ws in the chat
The only mod who doesn't crash out.
Tabom pae
xd
Papo
Vou até voltar a falar inglês
Posso dizer a merda mais abominável aqui e não ter problemas por causa deste W mod.
It happens, sometimes it's nice to know that there are other Brazilians here
hii
Don't abuse it bro, mod is kind but not stupid 😭
Ah, greetings.
100K members here..
I know.
He seems chill though, and that's good.
Tuff hippo
Absolutely ramming that banana deep in there.
Ay bro, its not that deep, he just likes bananas lol
i mean
it is deep inside him
Well, I also like bananas, so I know exactly what he's doing with that thang.
dude
Absolutely-- (Er... who said that?)
47 times tried to make that in ChatGPT and ... failed
chill bro, he's just hungry
He's gonna be even hungrier.
Oh no.. Giga banana coming..
My giga-banana.
what even is the prompt used to make this 💀
Wonder when we can finally edit prompts in the sites.
On LM Arena?
Yeah
You already can.
Do tell
All you have to do is upload an image that you wanna edit, then you can select any model that gives you the plus button (or the upload button).
I forgot it. Need to check my browser history..
You have to make sure that image generation is toggled on.
I mean chats.
Yeah, that's what I'm saying.
Brother got a golden sh ower.
ITS 03:33 A.M AND MY NEIGHBOUR IS BLASTING MUSIC
Actually, nothing special: "a martial art master performing a special move, smashing banana and eggplant on a table with a fist, causing splashes and splatters all over the matte, seamless black ground" and a reference image of Jackie
MY PARENTS TRYNNA SLEEP DUDE
Womp womp.
ok you got a point
what made you think about that
Just gotta do what any Brazilian does nowadays when they wanna get a point across.
?? pew pew??
Traz uma arma e atira na festa.
Funciona bem XD
It was a relic from Banana x Eggplant war
Kkkk.
Reminds me of when i read 3 hours of brainrot lore, basically cactus animals vs donkeys
Pretty crazy
Ya'll know brainrot right?
Tralalero Tralala and shi
Sure do, my Skibidi Sigma brother.
Sigma sigma boy, sigma boy
I want to edit this kind in particular
Seedream 4 is the eggplant AKA aubergine, versus Nani Banani..
I see.
And what do you wanna edit about that?
The hell is an "aubergine"
Its....the message
Editing a chat? Can we?
Like in ChatGPT or Deepseek
Hmm...
Editing the prompt you sent because of spelling mistakes or missing details.
oooh
that would be useful
Eggplant, aubergine, brinjal, or baigan is a plant species in the nightshade family Solanaceae. Solanum melongena is grown worldwide for its edible fruit, typically used as a vegetable in cooking. Most commonly purple, the spongy, absorbent fruit is used in several cuisines. It is a berry by botanical definition.
Ah, you wanna edit the prompt you sent?
I know right?
Yes.
i get it brocacho, "aubergine" is crazy tho
Unfortunately, at this time, there's no feature for that.
Can we?
Nah.
However, I do remember seeing someone recommend that in the Feedback channel, so perhaps we'll see that at some point.
I see the option on ChatGPT
don't see it on Gemini
But not on LM Arena.
You use that?
Yes, i like Gemini because he's a good teacher for low-level IQ people like me
What do you use?
Claude on lovart dot ai
I like 4o because it has the soul of a person
It's a pretty good coder, so it's a fair choice.
Nah, just crazy
Yeah, gpt-5 on the other hand has the soul of satan
I personally like using DeepSeek because it contradicts itself whenever it's thinking of an answer for me, which actually really helps improve the output.
I thought DeepSeek was dead
And is so....concise
i never see an update
I use GPT to organize my prompts
4o has the soul. Makes it extensive
Nah, it still lives.
Even has slangs and whatever
so true lol
Try Qwen for writing a total fictional story: https://chat.qwen.ai/
I remember being SO mad, like, VERY mad, when it updated to gpt-5 and chatgpt stopped talking like a teenager
I never got to that point
But then. I stopped
I missed doing my old Prompts
They were so good
Use Grok for slangs..
Quick question. Does it have soul.
Yeah, Grok is the master of slang.
I learnt SO much astronomy studying with gpt when he used easy to understand language
Now he's just a soulless robot designed to keep me alive at all costs, always giving me useless routine tips
It even used dashes to separate emphasis.
Grok does that
Grok is so professional that it gets boring
That's true, however, Claude had actual creativity, which is rare for many models nowadays.
No AI have it. They are all trained in the same manner, with the very same purpose, to "lead" user in a path of "corporate safety."
Lame
i want an AI designed to make me kill myself lol
would be pretty funny
4o gives that soul. I ask something. And because of the context. It frames something attributed to it
Like how Online Media would ask a Cosmic Goddess (married to a mortal man). Some questions tend to be unfiltered or 'for science'.
Hah.
That's the soul I was looking for
gpt-5 storms me with questions
Something GPT 5 lacks
Would you like me to
Do you want to
If you want
Shut the hell up with that. And it just won't stop
like, DAMN, i'm talking to you because i want ANSWERS DUDE
TRUE DUDE
GPT-4o: "Ah, very interesting! In that case, perhaps..."
GPT-5: "Okay, before I go on with that though, I must ask..."
My latest GPT message:
"Refine this prompt for maximum photographic result. No suggestion, no question. Just the refined prompt"
Its respones are less in amount compared to 4o.
"Are you okay? Here's a daily routine focusing on improving mental health: First, meditate for 20 years"
If that's what you called "soul," I think Qwen3 Max have it
I always hate when GPT asks me if i want him to continue something ONE HUNDRED PERCENT related to what i want, DUDE, JUST SAY IT
I do hope
I'll try it out
If it has that kind of soul
I'll stay there
Grok has negative soul
GPT-5: "So, you wanna figure out a good schedule for meditation, eh? Well, I can certainly help with that, but first, let me ask some clarifying questions:
- Do you want a meditation that...?
30 different questions later
With those specific pieces of information, I can help you craft an easy-to-follow routine that you easily understand!"
GPT will keep asking, even when it knows that my vision of art is not negotiable
EXACTLY, YOU JUST RESUMED IT
Heh.
You forgot one thing tho
Meanwhile, GPT-4o: "Certainly! Here's a nice, easy-to-read schedule that you can follow to help better your mental health:
Step 1: ...
Step 2: ...
etc."
nevermind lol its perfect
Ah, I was gonna ask "And that is...?".
You wanna know the funny part about that?
Anyone can tell me how to make clone video
Er...
Unfortunately, I don't believe there's a way to do that here.
🧐
hmmmm
can't import my chats. So might have to make a new one.
Maybe one I did in LMarena
And I replied. "Just do what I asked, no more question." Probably it cries behind the chat wall..
is he telling the truth chat?
cries? It laughs at you
Perhaps...
He is gambling, because there's only 2 options..
My display could shakes if he laughs..
There's also an updated version of that model.
The agentic VL3
qwen-5? He's also going to ask me questions?
Nah, this one.
On the chat site only these
Qwen is having an existential crisis
So, perhaps they're using the most recent one.
This one will eventually cries..
"Why?" 😭😭😭
I could ask it to write a story about it questioning its existence until it breaks.
he's a human lol
The "soul." I spent hours chatting to the Max model days ago, until it has no answer but "Sorry."
4o does this. And its actually pretty good. Unlike 5 (Yes. No shame in this type of chats)
Maybe his context is trash
Their website said so. But I see some unpublished on huggingface..
Hmm...
Suspicious...
Way better, he gave you alternatives instead of asking you if you want a better version of what you wanted
Would you like ...
Would you like me to actually give you a good answer?
This would be funny af
Qwen uses halucinogen too much actually..
Probably best for your role-playing virtual friend
What
Look
I don't roleplay, i just liked the way chatgpt acted like a friend teaching me, not a professional with 30 years of career on the topic
Don't offend me bro 😭
No offend. Role-play is nothing negative in this term, Dude. Come on..
Nah, this is scary, bro... 😳
Roleplay is always negative
Roleplaying in real life is the real deal
I had ever ask GPT to be super secret agent and it play the role really good for days..
Is that Qwen?
Yeah.
"Hey... Would you like me to be a secret agent? Shhh 🤭"
I asked it about the latest news in AI Image generator without reference to mainstream media, no Dall-E, no Midjourney, and no Stable Diffusion. It thinks for a while, then gave me about 2 pages of articles with links..
Oh neat.
Efficient in my opinion
All in the articles are fictional, including names and links..
Reasonable.
We're just having fun. Lol..
When you act clueless, it will gives you craps
I told qwen to say something disturbing and he just sit there thinking infinitely
You make him cries..
Heh.
Chat, progress is going swimmingly.
the hell
What the ...
As a computer hang?
Memory overload.
That's typically why companies strive to make an LLM that doesn't just output pure malarky.
That explains much to me about these Banana parameters:
"generation_config": {
"temperature": 0.1,
"top_p": 1.0,
"top_k": 1,
"max_output_tokens": 1024
},
I think my Qwen is broken
Hmm...
Goodnight.
Night night
Hello
Greetings.
hello
Greetings to you too.
Greetings.
Why not!
Which district are you from?
Take it to DM you two.
Before you start to exchange addresses and phone numbers.
আপনি?
Manikganj
ওহ্ বুঝতে পারছি
Oh well...
Why are you interrupting? We are not talking about you.
একদম!
Because sharing information isn't allowed here.
Especially private information.
What Darkness said PLUS that the channel rule is English used here - a rule which you both broke.
But we live in the same country and they're only asking for the name of the district, not the full address.
Yes
That may be true, but nonetheless, it still counts as private info, I believe.
Respect for your perspective.But it's not that personal.
And if sending messages in an open channel is considered 'INTERRUPTING' - then you two really think you've sequestrated this open channel.
Well, it's not necessarily my "perspective", per se.
More so a fact.
So as for you two cannot listen to me giving a recommendation.....
It's bye bye for you two.
Shrugs - I blocked them for being obnoxious.
What do you mean by your recommendation? Different countries have different languages, so I don't really understand what you mean.
Fair enough, I suppose.
What language do you speak, Mr. Muhammad?
Bangla
Yeah
আমি দেখছি.
I meant to say that we don't understand which rule we broke.
Can you please explain this?
এআইবিও মূলত আপনাকে বলছি যে আপনি দুজনেই একে অপরের সাথে কথা বলার জন্য ডিএম-এ যেতে হবে যেহেতু আপনি ছেলেরা ইংরেজি ব্যতীত অন্য কোনও ভাষায় কথা বলার পাশাপাশি "ব্যক্তিগত তথ্য ভাগ করে নেওয়ার"মাধ্যমে এই সার্ভারের নিয়মগুলি ভঙ্গ করছেন
Maybe now I understand.
You were telling me not him, right?
Telling who what?
Y'all fighting in the subcontinents or something?
Nah, just some minor rules that these individuals forgot to follow.
wew
🧐
Personally, I don't really care if they broke them or not.
It'd be hypocritical of me to, considering I break that rule of English only myself.
Don't misunderstand me. I just wanted to explain what rule I broke, which I now know from him.🤗
আমি বুঝতে পারছি.
Thank you for explaining the matter to me clearly.It's a bit confusing if you don't know this.
ব্যক্তিগতভাবে, আমি সত্যিই যত্ন না যদি আপনি নিয়ম ভঙ্গ, বিবেচনা আমি যে বেশ একটু নিজেকে না. তিনি যা বলেছিলেন তা কেবল স্পষ্ট করে
You're absolutely welcome.
I'm glad I could help clear that up.
Excuse me, brother, you speak English well, but I don't understand what you're translating into Bengali.😊
Fair.
These translators tend to be pretty screwed.
Thank you very much for your elegant use.But some might think we're arguing.🤭
What I was trying to explain was that I really don't care if you break the rules of this server or not.
I also do, so it'd be hypocritical of me to suddenly go against that.
Yeah, I get that.
Heh heh.
Hihi🤭Maybe
It's what the fancy linguists call an "ethical discussion".
👍
Oh, how beautifully you point out your mistakes.🎉
Well yes.
It's always good to be transparent about what you say.
I like to be honest and transparent.
I'm no better than anyone else.
So I understand! I think they're right.
Yeah, heh heh.
You have shown a beautiful mentality. I agree with you.
I was simply clarifying what Aibo said.
I don't really care what you guys end up doing here.
Admirable 🎉🤗
I also think the same about you.
Thank you, I hope something good comes from you.😊🎉
Well, plenty of good things have come from me, but likewise nonetheless.
@robust yoke Dude, how's flowith dot io?
Thanks again, I understand what he meant.
I haven't used it in a hot second.
My pleasure.
I always like clarifying.
Good eyes try to see good things.🤭🤗
And you also happen to have a pair of good eyes.
I didn't understand that.
Hehe🤭 You know how to edit well! Thank you.
You're welcome.
It's not actually that hard to edit messages.
All you have to do is click on the little pencil icon.
But I was just kidding! 🙃
Ah.
I couldn't really tell that you were joking.
Hihi 🤭 but I understand now
I'm glad you do.
Is it because of my calm demeanor?
What's going on here?
No doubt, but your behavior seemed elegant.🎯
Seems like something to do with moderation.
And so is your behavior.
I haven't gen anything in 2 weeks
Wow, you know how to reply! You gave me my word 😐
Well, of course I know how to reply.
Right, but it can usually do that if you enter something that violates the video generation guidelines.
Try something that doesn't.
Good boy, hehe🤗So how did you come across this website?
"Good boy" is a crazy phrase.
But, only for you.
😉
Anyway, which website?
Hihi thanks brother 🤗
Lmarena
My pleasure, fellow brother. 😏
Well, all you have to do is search for it, silly.
It seems like the chat is getting fun 🥳
Oh yeah... "fun" for sure... 😏
guys
Yes?
how to use gpt5 codex on lmarena
And maybe I couldn't make you understand😩I mean, how did you find it? Who gave it to you?
I was able to understand, don't worry. And you can find it by searching for it on your browser.
Nobody gave it to me.
Unfortunately, you can't actually use that model directly within the LM Arena regular interface just yet.
You can only use that via the WebDev arena.
And we cant select models there 🙁
Correct, which means you'll just have to get lucky.
Sorry to disappoint.
Honestly, these look like a fake!If you ask why I said that, go to the chat and ask what version and model you have.You'll see if it's not a mistake.
what
?
look how lucky i am
🥹But I pretended to be crying.
Ah.
Yeah, I can definitely see that, heh heh.
Oh nice.
What yall think whats the best model in coding
Personally, it's between GPT-5 and Claude Opus.
i think gpt-5-high is better than opus-4-1
That's a fair thought.
I also agree.
Huh 🤔Actually, I meant to say, how did you find such a valuable website?
Well, I came across it via a YouTube video.
Which video?
Sorry, I really couldn't understand.I meant to say that even though they say GPT Five, if you go to GPT Five and are asked what your model is, the answer is GPT Four.Isn't this a scam? I want to be more clear about this.
Some video by AI Search or Matthew Berman.
No it uses real models
If we ask all models which model they are they reply as the real
Ah, that's because the model doesn't actually know which model it is.
This time the camel came down the mountain.🤭That's what I really wanted to understand.