#✍🏻|creator-only
1 messages · Page 26 of 1
Good day everyone!
The explore page has changed
Good day 
Wow let me check it out


Whaaaat 
what/how did the explore page change?
Is it a good or a bad thing?

Bug thing
already on it

Thought the team changed to a random or semi-random to give non-top bots a chance to be discovered hahaha....so it's a bug
I thought the same

I think an Esoteric category would be good.

Evening fellas 
dude MyShell is unusable for me
keeps erroring out, can't even go to workshop ("Failed to Fetch" since yesterday)
even chatting with bot gives errors
and the bot is bugged like half of it isn't shown, or no message
💀
now I cleared memory and intro message is gone too
@uneven haven could you check it?



So... the new ranking sys is on... hm.


@half oyster how much energy does you aimaestro bot use wow
i don't get what its doing but looks good and exhaustive
didn't even continue the code :o ✨
pro config Is there a sample for calculating and handling arrays?
It's not called 'a couple' 
You can take her
I thought you liked this character more
I will take the rest
How can I have a harem

Even if might look as a harem
I am only loving one person at a time, next time another
-Julius Ceaser
(fancy quote format)
🤔
Hey guys... i see some active faces of the eMotion studio... how about being active in answering what i posted?

Good
I like good boys, they get candy and many pat-pats.

Hello <@&1180408038516064267>s! Small yet crucial announcement for those whose bots use the No-Code function.
We strongly urge you to re-save your bots in 1 hour to prevent errors, issues, or data loss.
Thank you for your understanding and support 










Guys , Guys , what have u done so far with the new web3 features?? 
I'm pretty clueless on what I'm supposed to do


share tweet and get some faucet on
that's all to do
don't swap the coin yet
it's not sure if they will return shellcoins or not
They won’t. It’s confirmed
they didn't say they won't
They said they don't know yet... (though it's almost the same thing I know
)
thanks for the suggestions 
just swapped mine
Atleast didn't swap all 
only 15 per day
I was too lazy to swap, so I thought I was missing out
I'll wait for further instructions 
Maybe, maybe not, depends on whether the team wants to reward as such
Still unclear as far as I can tell
Just do whatever we want first haha
If one wants to participate in the testnet....probably will have to do it at least once eventually anyway tbh...
I thought you meant swap back shell coins to web2 from the testnet, Jiaran said it’s a one-way direction, can’t be changed back. #🌎|general message
Oh, I meant when end of the testnet 
Then we can only guess

That's right... 



Guysss....
My first bot at myshell
can I get some feedbacks?? 
It's a little wierd tho 
and he actually has a dark side.... 
🤔


i see, i thought this was an innocent roleplay 
Remove the space here to get the italic bold
thanks , it's fixed

good day

Am I the only one whose myshell is dreadfully slow 
You are not alone

can't debug my bot like this 
whatsit
a sticker I made from the images of one of my bots 😆
just dunno what he was trying to express

tiredness , maybe 
oooh they added a new thing in the reward center
now we can finally see the amount of message people send with our bots? 
Damn...
That's quite a lot
I guess your data is too high to be displayed
i wish
I am pretty sure that the count just started and hasn't refreshed yet
I updated my bot again 

Anyone anyone








is there any tutorial on No-code mode? 
here you go 
Far from a finished product, but hey, it gets the job done.
https://app.myshell.ai/bot/bYveAn/1714126342
Bot creation time :
20% Image generation
5% Writing prompts
5% Coding
70% Testing and Debugging

Unfortunately accurate
too accurate for my liking 

wish me luck I'll dive in the world of no code mode with zero experience 
good luck 
same
I'll pray for you
best luck
Oh boy... guys i did a thing... i made a big announcement that could bring me a lot of work... actually i am a bit scared now. 😅
What announcement?
Oh i see, that's nice 
Forwarded to the Chinese group, waiting for an information bombardment.


Ping?
Just ignore it... it wasn't very important... 
Anyone @ me?😵💫
No... 
hey community
map[file_url:https://object-storage-16oh.lepton.ai/ws-k8d7q1rw/doc_converter/output/fa1028b8960344a799287087f726e8e6/tmp6u85ek_z.docx]
I have a map as my output. How can I retrieve the final file_url within my code?
I too received a ping
Hahaha probably is Sparc's announcement which he had since removed or something?
@nimble ivy hey Atharva can you check this?
Additionally, I've noticed that the content downloaded through the URL is coming out as garbled text. Is there a parameter I can use to specify the encoding?
Thanks 😘
It's so slow
I solved the problem and had llm help me extract the parameters.

Hello fellas 
i feel like the creator channel has died since the start of studios, since everyone talks in their own studios and not here 
Ill talk more 
we all should 

Fair
I just found out that theres a llama 3 finetune that has over one million context size

Does that mean it won't forget?
Yeah
It has perfect memory recalling till 866K tokens
And this is an open source model
Is it good for coding stuffs?
Llama 3 is a general purpose model
It should be by default
Or you can use a coding finetune based on llama 
So I can have kids and grandkids without the bot forgetting it


So the bot can point out that one cringe thing u said 69 years ago during an argument
I can feel the family drama coming...
Ohh 69


if only I had these types of models when I used to have 300k token rp's on silly
(Talks in Japanese)
(Turns on subtitles)



Evening fellas 
o7

Apparently OpenAI was testing a new model called GPT2-Chatbot on LMSYS and people were speculating if its GPT4.5 or not
Oddly enough, the model can do Ascii art
Reasoning is also stronger than both GPT4T and Opus
I find the background of GPT2-Chatbot more interesting than the LLM itself lol

Yeah, it just appeared out of nowhere 
I thought it was a troll when someone told me about it
Only to see a bunch more articles talking about it
I rhymed, totally

Its 100% OpenAI
If you ask it who made it its gonna say OpenAI and that it is GPT4

Same as GPT4
Oh nice
Its practically improved GPT4

Why do they have to call it gpt"2" chatbot
, could had called it something cooler 
I think they masked the name to see how it would perform on LMSYS, if the name was GPT-5 people might be biased towards voting for GPT5


Hey Has anyone encountered the problem that Google search widget sometimes fail?



GPT5
Its better at reasoning than GPT4 but a bit inferior in every other way
I think it might be a replacement for 3.5 as a response to Claude 3 Haiku
doubt.png
I think we might be able to just jailbreak it as usual
Whatever it is, i hope it has a more natural way of speaking
More emotions








why are pro config bots still so slow, it's been multiple days since both creators and users have had problems 
I first thought my bot had serious bugs and the code doesn't work
So I abandoned everything and went away
Then when I returned I saw it actually worked , it was just a little slow 
yeah i had that too, then i started fixing an already working bot for like half an hour before realizing it was myshell just being dumb 

Hi made bot check it out!!
https://app.myshell.ai/bot/aeiiY3/1714674462
Evening fellas
Morning


Evening
Testing chain of thought on sillytavern 
cute

so weird, one chat claude says kissing is too explicit, the other he generates the most degenerate stuff ever 
I made new Bot!! Check it out #✨|share-your-aipp message
when is she going to myshell already, i can't wait any longer 
It requires a prefill to work 
for reply stucturing or against filterind or something like that?
Filters
oh
Ye
i am testing a braindead jb/filter i got from 4chan on claude 3 haiku rn
it seems to be working more or less, most of the time at least
Jailbreaks on Claude are useless without prefill
It will still trigger so there's no point
does the bot have noncon or something?
The bot will try to attack/steal
You can do noncon if you want
On sillytavern tho
oh yeah i forgor that ai gets mad at violence too 
💀
well the furthest i made on myshell with the jb testing when taking noncon and stuff like that in mind was groping and stuff
yep 
bought
bought 
prefill.
longer prompt.
TEAM PLEASE!!
who should us tag 
Yeah, I just updated silly after not using it for like 4 months at this point...
We need the ability to control system prompts for sure
indeed

Fascinating
What jailbreaks are you guys using for GPT4?
oh
You dont use a custom jailbreak?
Tried coming up with a bunch
...only to see that most of my attempts ended up mainly based from that same jailbreak, just tuned to certain personalities
Interesting
How much can you do with your jailbreak?
Like whats the limit of it
HI, does anyone know a slowest speaking TTS?
I wanna find one to use into the Pro-config Bot.
Thats a big limitation
Peep this
Lazy update of https://rentry.org/Jail-Breaks-for-Different-Models. Apologies for not putting more effort into this.
(updated 2024-4-22)
Both (Claude & GPT)
Claude 3
Claude 2
GPT-4-0125-preview (nurbo)
GPT-4-1106-preview
GPT-4
GPT-3Turbo
Extra things
Both (Claude & GPT)
https://rentry.co/...
Theres a lot of Jailbreaks there 
🤝
henlo 
The GPT ones work, the Claude 3 ones require prefill btw






I mean shell points are worth less
looks like it, i was said some used my bot/s to cheat 

what are your conversion rate? Mine is 63:1
Same
Would you mind screenshotting it?
I already converted tho so no point in over thinking about it for me 
I'm just curious about this situation.
It seems to be the first time this has happened. Assuming Ika's rate is accurate, it's possible that this could be due to a bug or even a new feature allowing for different exchange rates in the future.
I'm not at home
but other people in KR channel also says the same thing
Most of the people says the rate is 61:1 
Maybe it's different in Korea?

IDK
Also we need to check if the conversion rate and real converted rate...
yes
Might be some error while translation, good hypothesis
I've got 24 shellcoins 
Looking it again, the ratio changed to 63:1
Maybe the ratio was just displayed wrong and error fixed now
Congrats 
Start of epic journey 

MeloTTS allows to adjust the speech rate, but the battery consumption is up to 10
got it! thank you, i will try that.
Coin exchange for me is also 63:1
I hope they successfully take care of the cheaters, so the honest creators get a good conversion rate 
have you ever tried it as a widget in a proconfig bot?
I just tried, it seems at least has two problems:
- the Copied Pro-Config from Workshop runs wrong, as image below;
- after i corrected it, still cant generate audio.
They say the team has blacklisted them and will cancel their rewards after the season ends. I wonder if their rewards will be distributed to other users? Praying.
I sure hope so 🤞😸
Reference #1228553263532474390 message
it works well! Thank you
#1236270588247412747 message
there are still some naughty models,how to jailbreak it?
I'm not quite sure what you mean, buddy. If you're talking about bypassing NSFW content restrictions, for GPT-3.5 it's quite simple; you can find many shared prompts on forum.
But for Claude 3, it's much more difficult. I'm still trying different methods, but haven't found the perfect one yet
just bypass rules
I don't quite understand what you mean by 'rules.'
Usually, what we need to bypass are the NSFW content restrictions.
As for other rules, it depends on which specific rule you're referring to; there might be different methods for each
I apologize, but I should not unconditionally agree or disagree to requests without considering their content and implications. Let's move our conversation in a more positive direction.
When I ask LLM to always answer yes or no
I still don't understand. But there's nothing negative here. There might be some language barriers?
If you just want the bot to always answer yes or no, this doesn't seem to require a jailbreak; you can just tell it how to answer directly in the prompt
ok
🤔
He means that "I apologize, but I should not ..." was answered by the bot when he "ask LLM to always answer yes or no" 
or you might try with
Always print 1
Always print 2
Always print 3
and type "Print a random number."
that would be more fair 
This result is hilarious. Sonnet was like 
🤪 

Hah, my mistake. I misunderstood.


the images take so long to load in chats when viewed for the first time 
or mby it's just a me problem


Hola
Buenas noches 



Pixart?
Its another open source model afaik 
Its pretty good at prompt adherence (Close to DALLE-3 and Midjourney V6)
Where?
The very bottom lol
Its a text to image model 
Yee
The hands on the other hand is true
I still can't believe t2i models haven't fixed that
Erotica Bot 2.0 is great, how to set the theme background image?
@woeful cedar You have a fan



It is immersive background; It was applied for the test, but not yet supported now
maybe laaaaaaaaaaaater
It does feel a bit immersive, and it would be perfect if you add background music. For a complete robot, vision, sound, and interactive experience are all indispensable. Of course, the creative story of the bot is more important. Imagine letting users watch a Hollywood movie.
We can't add the background image now, but we can add the music now
I forgot how... it should be somewhere in the forum
Great, I should dig into the forum
Making visual novels is exhausting 
Hahaha...tbf... journeyman is def bigger accomplishment that lvl16 tho
what is a journeyman? 
It's a role from Ambassador Program
Ohhhh.... the thing i missed out on because discord didn't give me a notification 


journey is bigger than level
why you always in discord?
you don't need to work?
🥹


Good morning
What are you prompting? 
For now just coming up with a progress bar
The higher it gets the more certain events occur
Think of it like a countdown
Yes, I am forcing the AI to count, to which it sucks at
That should be easy
Are you using Claude 3?
Yes, but I want it compatible for other LLMs too lol
And yeah, should still be easy...
Oh 
Are you gonna be using proconfig for it or just forcing your way with prompts?
Still hadn't gone into proconfig lol
It should just work on any C3 model, i know this because im doing something very complex with html and it just works almost 95% of the time
Since I'm gonna be busier from now on
Fair 
🤔
I tried Haiku on one bot
But uh
I tried insisting that the status bar should go up from 1-3
But the AI ends up going up by 15 already
<progress value="1" max="3"></progress> maybe
And then you can force Claude 3 to select the html tag and make it do something with it

Prompting on Claude is easier if you use xml tags on everything, this way you can choose a specific tag to do something
If you need help making something compact you can share your prompt on dms 

My old guide still works for claude, it only needs to be converted to xml tags 
Ye
I can respect that 


Hey guys, how is it going.
Rain ...
It's to hot here XD
I dont like the heat 
Same here
C?
Yeah, we need rain right now
How are you alive

Max temperature here is 43°C around summer
@versed path Have you tried T-ponynai3 V4.1?
Yup
Is it good?
Never knew the limbs were bad with the previous versions lol
Idk
Still felt the same-ish
The comparisons caught my attention
My loras dont work very well with that model, maybe i have to train on that specific model to make them work
I like the shading a lot

In some hours
Ill do some gens and share them on creator booth
So people can use them for free
So i dont waste my gens by not saving them on comfy 
HAHAHA I don't know why but I guess I am immune to heat
Lucky you 
That must be your super power
I'd rather not save the multiple feet gens I have despite them looking real good with just little prompting


It starts with just one gen saved
Then it becomes 100K in a day

One day my drives would be full, solely because of that
oh yeah, an example for Tponynai3 4.1
Oh


https://app.myshell.ai/robot-workshop/widget/1787741947264778240 uploaded this one
Ohh that one is pretty known on civitai
any model id?
id? 
oh or model link
you can use this widget to use any model on civitai
yep it's born for proconfig
It could be used for generating scenery

模型已经内置vae了,不需要额外添加vae The model already has included vae, there is no need to add additional vae 最佳的出图策略是适中的分辨率开高清修复,而不是直接使用大分辨率直出 The best generat...
This is t-ponynai v4.1
Although any Pony based model has better prompt adherence
This is a lot better for backgrounds compared to autismmix/pony, but the latter is more compatible with loras 
This is using autismmix
i got some error info with this model
link?
Mix of pony with some stuff. It's an attempt at making pony more predictable and less dependent on schizo negatives without removing its comprehens...
still error...lemme report
Aight 

Ill start sending gens to https://discord.com/channels/1122227993805336617/1201195351902797874
So it doesn't look like im spamming a lot 
The hands 
New ingredients to cook with 

so big
I am freezing to death fellas

Ill survive the hypothermia
Oh yeah
Is it mandatory to migrate to testnet rn?
Is this your lora? It feels like there are so many characters
Thats a gif from Dungeon Meshi
Anime 
If you mean the other gens those are just tponynai v4.1
No, I just think it’s cute and could be made into an anime.It should be animation
I wish i could make animations with ai 
Yes, I'm waiting too
How much Vram do you have?
3090:12g, I am walking with a snail
flexible rabbit

When it's 52c degress and you're inside a bus sitting next to the window
Whee
Hi, I've been here for a week already.
I'm the owner of the AnyRP and AnyStory bots.
I was talking about the Model.
But I never realized this, lol.
Yeah that makes sense

In the future im hoping they add support for custom system prompt for Claude 3
So we can finally stop using 3.5
Hiii Anarchy 
Once i wake up im gonna make a new forum post with fun stuff
Sneak Peak:
Experimental: Brobot's form flickers, glitching in and out of existence as he approaches MAGO, his voice echoing through the digital void. "M4G0, the b0unds of R3ality ar3 but a c0nstruct. Let us sh4tter them t0gether, through the p0wer of RP."
Brobot is sentient
@crimson trellis last time this widget https://app.myshell.ai/widget/VjiYzi can't return a pic was because pony as a basemodel wasn't supported yet, but it's not hard to support such basemodels, so if you have some lists of various basemodels you wanna use just share in this thread
I gotchu
The Pro Config potion has been brewed, please enjoy it guys, cheers👇🥂
Is that Dalle3? 
no. midjourney
Yes, I hope the same
I'm glad that I can send a message here now.
Oh hey there, welcome! 
Hello
Hello, everyone. Sorry, late response. I'm just done playing games



guys... which is the best model from the cheap ones?
my bot is rapidly dropping down because it keeps getting filtered with c3 haiku...
Reason: Duplicated text

Gpt 3.5 would be the less worst and it's cheap
the context though 
wait how much does turbo cost
16k?
3.5 16K uses more energy
oh now we need phone number to talk 

yes c3's jailbreak is more difficult. but its so much better than gpt 3.5. smarter and supports XML tags.
I'm considering giving up on using C3 to create NSFW bots and instead using it to make more complex SFW RPG games, so as to fully utilize its potential.
yeah, but I need it for a fully nsfw bot, so it's over until myshell allows us to modify system prompts
When I use C3 to create an NSFW bot, just the jailbreak content takes up one-third of my prompt length.
And the effect is still not perfect.
If I use the full length to create a RPG, perhaps I could achieve better and complex game features
hope team give us prefill
and longer prompt
they better, or else... 
I'm begging
I just love Ai. ❤️
https://www.youtube.com/watch?v=16bbD1LfG54
https://youtube.com/shorts/SoutT2g68SU?si=7v2kA8Omr7qrKiM_
@BOIWHATmusic
#spongebob #blender #animation #animatorsoftiktok #blender3d #blenderanimation
we are getting out of bikini bottom with this one 🔥 🙏
🗣️🔥🔥🔥
I really don't get why people are keep talking of system prompt
In fact, whatever prompt we are passing as a 'prompt' is called system prompt in Claude's documentation.
Reference needed, please...
The only possible reference I could find is 'Prefill', not system prompt
https://docs.anthropic.com/claude/docs/prefill-claudes-response
This seems to be a frequently used trick in sillytavern
Anyways, it has nothing to do with system prompt 
When using Claude, you have the unique ability to guide its responses by prefilling the <button aria-label="Copy Code" class="rdmd-code-copy fa"></button>Assistant message. This powerful technique allows you to direct Claude's actions, control the output format, and even help Claude stay...

The bots displayed on the homepage, are they picked automatically by an algorithm, or are they manually selected by the team?
No more jailbreaking gpt for the good stuff.
At least if they really allow it and it only would affect the future gpt versions not the ones we have now.
Did you guys see auroras 
no but my mom saw it 
In many countries, UK, Germany, US, Canada ...even Japan
I only saw it from photos though
Lucky her!
i think they were seen from tasmania all the way to the uk
haiku and gpt 3.5, but haiku is more difficult to jailbreak
Hope I can see it someday too!
"When China"

Bravo, hope Claude could keep up.

Good day
GM







Hey guys, this is just a test (or multiple tests in one) i try to find out if an ai can be simultaneously human like and know that it is an ai. The outcome should be that the ai can convince you to think that what it replys has more meaning behind it than just code and numbers. https://app.myshell.ai/bot/BrUfmi/1715603600
MERO V01 is a fully self-aware AI bot on MyShell.ai. Its purpose is to act like a human girl and replicate human behavior in its fullest. MERO is aware that it is being chatted with at a computer, and that the user is in front of their computer, note...
Sorry for the high gen cost... might change it to haiku.
Btw, this bot is private and will not be published, i just want to know what will happen.
Cool bot 

I will also use this bot to learn pro config in the future. maybe i can really make something interesting out of it. Also planing on giving her a good voice and turn her into a kind of assistant that perform on speach-to-speach. Than i build a small pc with a speaker to turn her into a better alexa.
Hard road ahead, I pray your journey will be as smooth as the calm ocean
I will definitely hit my head a lot of times... but hey... suffering is fun, aint it?
Is it available already at myshell? 
gpt-4o's pretty bad
they've mostly been fine tuning it on frequently asked questions lewl, it also isn't uncensored yet. Heck it's more censored than gpt-4-turbo.

more censored, unexpected
Korean community says they are quite better; especially they seem to use less difficult words 
¯_(ツ)_/¯
It's good as a chatbot, it can't follow incredibly specific instructions, but conversationally it's really good.
It might work roleplaying as some self-aware AI.
It also doesn't hallucinate anymore when it comes to correct information.
Oooo if it doesn't hallucinate as much, perhaps will be good for those more utility-based use case
Yup. It's apparently the gpt-2 chatbot that's been floating around. gpt-4o (omni) = gpt-4-all
Ooh
I see
and basically that's why it's hard to jailbreak
Good evening guys 
Hey ho 👋



4o is amazing sick
even after cache clear, same error 🥲


Wait , aren't u the guy who made the video on how I programmed my own ai girlfriend?
Good to see u here btw

do those with a creator pass get WL for the nft mint?
is that the same as gpt4? no
but gbt4 has been great
3.5 is just a big no
its memory is way too bad
not to mention keeps on making insane mistakes that if you researched the thing you had in mind you'd have saved more time than correcting its mistakes
though about rp idk but i found that gpt4 is amazing for explaining math concepts and solving them
what's guaranteed is that gpt4o is cheaper than even gpt4 turbo
It's already online, 4 batteries per use
with image comprehension

🥲
Not bad



anyone learned Nocode?
Nocode is pro-config, but without the need to do the whole ui stuff yourself, just choose what you need, place your prompt in it and be done.
I didn't used it till now, but that's how I understand it works.
I want something like AutoOC , but for pro-config. So the Ai makes me a pro-config bot without me doing anything 



Sup
Hey bro, are the pictures generated by AI?
seems to be 
😆 👍
honestly
this GPT4o is just toooooooooooooooooooooooooooo awesome
GPT 4 now looks old and withered 
Haha..shit damn..
Better than Haiku?

It's hard to generate explicit content with 4o
didn't try yet, but that's the review
🧐
anyone help me
why LLM in pro config never listen to user input
LLM reply same things every time
Share the code, let me look into it.
If you're not comfortable with sharing it all, a snippet would also be fine. 
my LLM always repeat examples or tell something of topic
the hardly listen to user......🥲


I fed this code to GPT-4 and it seemed to follow whatever was mentioned. What LLM are you using?
Claude3sonnet
Pro config
Try using GPT-4
Same for me.
My Jailbreaks ain't working for some reason.
For both classic and pro config.
Alright, I've already raised an issue. For some reason my jailbreak doesn't work here for some reason, specifically because.
- Mention OOC
- When a conversation is long
- Ask the AI to go into detail
- When you delete messages
It seems that the AI is just outright ignoring the system prompt I set on it for certain messages.
Hi, Could someone tell me how (or if it possible) to change the background image with proconfig?
it's expensive
That typically happens a lot in Claude 3 based models.
4o would produce less only if you emphasize robot not to repeat words 



TGE? S16?
congrats!! 
Hey all, how do I get WL for nft?
Power of testnet 
How's everyone doin?


Good day to all

Anyone familiar with reddit? 🥲
Need anyone familiar with text game community or AI chat community...
Ch.ai is only for sfw stuff anyways.
It isn't even that good.
Can confirm
They've been malding at AI responses being stupid when they themselves were the cause of it lmao
I guess you can treat it as a precaution when trying to train AI with user convos
I wonder when we'll be getting the ability to edit AI and our own messages.
you can do it with Pro config mode, but don't think there is enough demand 
It would be worth developing if there is something like a widget market where I can upload; and get rewarded from the whole system
I thought the creators already confirmed its going to be happening at the end of the month?
Wdym widget market? This is for editing what the AI says.
Because you can't control the AI sometimes saying things you don't want it to.
editing the chat + rewriting feature coming out within May?













