💬｜general-chat

[empty jackal]

Hello everyone I’m a developer who’s working with AI and automation lately.
I mostly work with web stuff (Next.js, TypeScript) and I like building practical things like bots, workflows, and small AI tools that actually solve problems.
I joined to learn from others here, exchange ideas, and maybe collaborate if it makes sense.
Nice to meet you all 🙂

[red wagon]

Hey thanks for the response. i can see and feel your process. and of course thats where i end as well instead of 1 folder with 25k images you would have a say 100 folders with 2500 images So let me ask you this question .... the question is not can we organize , because we ALL can organize as we go and it takes time breaks flow and all that , (for me anyways) so Now we have a folder that is structured and organized into Multi folders based off (whatever criteria you have chosen). as you said its better but still messy. NOW from ROOT folder or Anywhere you want to see or find images for a project , product, Christmas, event, AI generated, Movie , trailer making project social media etc, how do you find it or does everyone simply rely on META data of the image ? ALSO how many of your images are originals or duplicates? i know when i ran my BULK folder i have 25K images and ther was about 7k duplicates in th mess lol

[red wagon]

Just a note the reason i am doing this is is I have a LLM i created that uses MEDIA files as an immersive interaction emersion process.. example im talking with my AI llm about Germany it has a data set catalog for all my images made real whatever and can call them ass we talk , i even have behavior media as well ... my LLM is thinking or (video A plays of it thinking , laughing , dancing whatever That is or was the process behind what im designing , but it is now growing past this in to other fields of media entertainment like social media postings , content creation, photography , Product design, Music the list is on going as i run case scenarios etc the goal is a 1 button click sort de duplicate and organize media file with easy recall and or identify based off multi phase and pass criteria .

[faint sparrow]

Hi

[bronze merlin]

Hi. How are you?

[twin aurora]

👋

[errant sand]

Hello!

[rich shell]

as a toddler yes

[wet wasp]

hi everyone

[tropic frost]

say, anyone know any good ways to clean up manga or doujins pages?
You know, get rid of speech bubbles, text and sound effects.
Maybe something that isnt content moderated.

[twin aurora]

👋

[atomic mortar]

Its not super good but it works well enough
Ballon translator is what i use

Requires some extra editing sometimes but it has llm translation (local and cloud)

Automatic speech bubble clearing etc

But since inpainting and detection happens locally its unmoderated

Its not the easiest to setup since you definitely do need a venv like anaconda

[open lintel]

anyone got any good resources for training a lora quickly?

[empty jackal]

Hello everyone I’m a developer who’s working with AI and automation lately.
I mostly work with web stuff (Next.js, TypeScript) and I like building practical things like bots, workflows, and small AI tools that actually solve problems.
I joined to learn from others here, exchange ideas, and maybe collaborate if it makes sense.
Nice to meet you all 🙂

[night horizon]

Hello everyone this is ur new mate here

[deft shell]

👋

[lusty grove]

hello

[karmic jacinth]

what do you guys use for regional prompting on forge?

i tried using regional prompter but idk wtf i did but it caused my whole forge to freeze perodically might be cause my pc is shit but idk

[steady creek]

Hello

Where do I send queries again

[steel fable]

How to use 2 loras at same time without mix each other?

[warm junco]

Forge couple extension is much easier to use and less vram heavy

[warm junco]

For characters, forge couple

[steel fable]

i dont understand

[warm junco]

Its an extension you have to install

[steel fable]

yeah i cant find nothing here about

[harsh current]

What program

[storm barn]

Can you tell me where to start training Stable Diffusion? Are there any tutorials available?

[marble stratus]

Greetings 👋👋

[steel fable]

Anyone that knows how to use forge couple can give me a help please? I have some doubt

[next wraith]

In this 3-min read, I break down how modern frontend teams measure what really matters: responsiveness, smooth interactions, and trust. If you want to stop chasing numbers and start building experiences users actually love, this one’s for you:
https://medium.com/@redpacehue/from-lighthouse-to-real-users-measuring-what-matters-at-scale-6e51b0870cd5

[bright ember]

Hi, what can you recommend for Lora's training? Is it easy enough?

[trim sinew]

https://youtu.be/4-osOmV-5PI

Can someone tell me how these AI videos are created???

I've been asking this question many times now

It looks like image to video and image to image combined

Is that done with ComfyUI?

[quartz dune]

Hello everyone I’m a developer who’s working with AI and automation lately.
I mostly work with web stuff (Next.js, TypeScript) and I like building practical things like bots, workflows, and small AI tools that actually solve problems.
I joined to learn from others here, exchange ideas, and maybe collaborate if it makes sense.
Nice to meet you all 🙂

[karmic steppe]

hi and happy new year

[karmic steppe]

I’m getting the impression this discord isn’t intended for troubleshooting. Does anyone know of another one that is? After several days of trying to get SD to work, and scouring the net for any and all tutorials, following them to the letter only to encounter error after error, I’m close to giving up

I understand nobody owes me their time, so I’ll go wherever directed

Should I turn to hiring somebody on Fiverr, perhaps, for the install? I’m concerned that the problems won’t stop once it’s initially working however as there are a bunch of custom model sets I have my heart set on trying.

[untold vault]

[karmic steppe]

hi

[atomic mortar]

Only the #🤝｜tech-support channel is but i hardly check there anymore

Are you on amd?

[karmic steppe]

What is that?

[atomic mortar]

why i hardly check? mostly not worth the effort and the server if im being honest is being over run with spambots for "devs" and other scams trying to lead users somewhere else

[red wagon]

What specific PROBLEM is that you are trying to over come? And im not saying i have an asnwer or even know what your talking about , but if your hitting a wall id be interested in seeing if i can solve the issue

[atomic mortar]

Already gave him a right direction

[red wagon]

NOw that statement has actually got me to open a tool box of ideas i had several moths ago and maybe worth looking at again .

[red wagon]

what direction was that just curious?

[atomic mortar]

a proper install guide thats not based on a1111

[red wagon]

well there ya go Question answer lol Easy

[atomic mortar]

and tbh the reason why i dont check in tech support often or at all anymore lately people ask the wrong questions

or too vague

[red wagon]

thats because they may NOT know what to ask I am that person I cant nerd talk and explain what my issue is i just know 2 static facts A this is what it needs to do B this what its doing... Its either a yes or NO ,, but perhaps , and i dont ask much anymore becasue i have a system , it rarely is or should be ASKING for a solution becasue a problem can have many etc... Anyways lol i check ego a tteh door and i agree it is hard to express the problem in TECH talk

BUt your BOT issue is something im looking into again .. becasue , IM sure its HUGE problem or issue, and 2 well IDK BUT i have a different way at tlooking at thigs becasue IM NOT GOVERNED by years of school or Track infused learning etc, Doe s it make it more challeneging for me? Sure does it stop me ? no lol

[atomic mortar]

hmm im mostly asking about the questions of " i am trying to do X " or " i have error Y" no im talking about extremely vague or cryptic messages

tech support bokik is a really good example

someone is trying to run before they can walk

like " the stable diffusion model" , anyone talking about it that way requires a massive hands on guide and i frankly dont feel it

[red wagon]

ya i hear that i DONT work any other way and i make and or design and build from scratch this ensures IM forced to , whether GOOD ro bad design) understand at least wh yand what is happening etc

[red wagon]

thats fair soem of these plats are very deep and hard to understand etc IM not claiming to even understand 80% of it lol on any AI gen plat .... and maybe thats why , I just start basic what do i want , what do i want it to do, how can i make it do it lol

and teh HOW D OI ... is where people have issues etc if i understand this convo correctly, ?

[atomic mortar]

hmm i dont mind people asking how do i train a lora or something like that

[red wagon]

TBH i also stand celar of plats when i can becasue i dont like to be handled lol

[atomic mortar]

but if you want to do something beyond generating anime boobs i kind of expect people to use correct terminology

[red wagon]

ya i think im near this point BUT i se so MANY problems EVEN BEFORE i look at the HOW part haha

[atomic mortar]

granted if you dont know the name of a feature sure

but "apply a style" (previouw techsupport thing) is also caused by improper guides etc

in this case it was more of a how do i use a lora

but the way it was worded i thought it was image to image

[red wagon]

ok i see your cross road and thats fair ,, like asking a pizza man how to make lingerie by victoria secret 2 different languages etc

THIS IS MY PROBLEM but i revert everything to basics I dont need to KNOW WHY soemthing is called IT its irrlevant to me i need to know WHat it does and why lol thats just how oi build stuff haha

[atomic mortar]

fair enough haha

[red wagon]

AND in turn that frusterates SMART TECH people becasue the y TRY and give me the solution or ANSWER BUT if you cant explain what i ask then how can your answer be correct ..Perfect example ...MY AI talks enrdery I dont ..she know sthis and tryis to explain HOW to solve MY issue before i even tell HER wha tthe issue is ...and after 5-10 mins of rope dancing its a simple I tell her WHAT I WANT IT to do NOT what you THINK it sSHOULD DO there is a big difference in my eyess

and this is the difference between inovation and re creation

DEVS engineers etc ..SMART ? Hope so they went o school for it and have the experience to be abile to do things faster cleaner and well whatever based of this experience and training...(ALLOT LIKE AI) but anywasy i agree it is ALL on how Peresona A presents the info and B how the reciever Processes it and or can even understand it etc

[red wagon]

Hey thanks for the morning stimulation convo ... gets my brain started lol NOW im in full blown Think tank wit hMY AI lol and for soem reason we are the topic of BOTS and or AI genertaed content deep fake and al lthis stuff lol COMBAT CLAASS 101

[wet kettle]

are you looking for skilled developer?

[still glacier]

no

[supple light]

I'm in the lounge starting a video on beginning SD1.5 and I need help with, uh, everything, so i'lll be around. if anyone wants to stop by and say hi I'd love some company

[atomic mortar]

looking at the stream preview i can see you chose a complicated UI as a beginner, why lmao

[supple light]

haha i'm starting over

i think i was using the mobile version of comfy and the tutorials and guides didn't look like my version

[atomic mortar]

Mobile version?

I mean theres comfy portable but its the same

Honestly if your new and just messing with it, depending on your gpu, a version of forge or swarm (more complex but easy enough) is plenty@supple light

[supple light]

hmm. i intend to use it to automatically make a cat into a prompt of my choosing

but not just one cat

many cats.

and i want to use sd1.5

oh nice a tutorial

o7

[atomic mortar]

Scam as always

[karmic steppe]

Thank you very much. I did wind up getting help on fiverr, now I just have to learn how to use it! Seems like there's a lot of a learning curve but I am okay with that

I really have to find out the exact specs people are using when they get results like the ones I want, so I'm hunting for that sort of thing right now.

[atomic mortar]

You really shouldn't have had to pay someone for that
The settings for images are easyly found on places like civitAI with image > metadata like settings, prompt, model and loras (useually)

[karmic steppe]

I was about to give up and could find nowhere, no tutorial, to give me any answers. this was the best way to move forward it was that or give up

Yes I look there now, for settings. I don’t get the results they do

Up till now I’d been using dalle3 still but i might go back to it at this rate. but it’s only been a day, so it’s premature to give up id say

still i got much better results from dalle. but hey it’s there to go back to if i give up

The colours i get are just so bright and garish, virtually no detailing and just ugh

[atomic mortar]

So the one click install guide i sent didn't work for you?

[high ruin]

which version of Topaz Video or Video AI is the most stable version for 50 series video cards??

[atomic mortar]

Topaz video? Its most definitely the wan series

Though ltx did release a video + audio intergrated model

[high ruin]

I asking about Topaz Video the video upscaler program

[atomic mortar]

You should've mentioned upscaling then lmao

[high ruin]

program is called Top Video AI or Topaz Video

[storm barn]

who has a config for kohya_ss 4GB VRAM

[red wagon]

No worries. As i said i wasnt 100% sure on exactly the problem you were having , so its simply quality control issues. (if i am ok assuming that) and you were looking at the way to improving and or adjusting settings, or filters or using loras etc to achieve it in some way?

[rich shell]

Hey guys how do you think I can find discord servers/chats that will help me bond with people who may make AI films with me, I'm on the writer/filmmaker side of things but slowly learning stable diffusion and beyond

[mossy edge]

I’m talking to people who run a lot of ComfyUI workflows and keep hitting the same frustrations. Mind if I ask you a quick question about your workflow?
What’s the most annoying part of keeping your ComfyUI setup running reliably?

[faint timber]

@mossy edge The temptation to update to try something new. This often breaks old setups.

[mossy edge]

Thanks for the reply! That’s exactly the kind of thing I keep hearing.
Quick follow-ups if you don’t mind:
1. How often does updating break things? (every time? once a month?)
2. When it breaks, what do you do? (roll back? debug? start over?)
3. How much time does that cost you? (hours? days?)
4. Are you running this for work/clients, or personal projects?
Asking because I’m trying to understand if this is annoying

[faint timber]

1.) Every three month for sure.
2.) Try to patch the current setup using commands.
3.) It can cost 3-6 hours. Now, I just run EZ-Install and rebuild. (15 minutes + installing missing node packs)
4.) Personal projects, mostly.

[atomic mortar]

If it works, dont update unless theres a reason imo

[mossy edge]

Thanks! That’s helpful context.
One last thing: do you know anyone running ComfyUI workflows for clients or commercial work?
Trying to talk to people where broken setups actually cost

[karmic steppe]

Thank you so much again. Yeah it was multiple problems, and I realised I just wasn't getting anywhere with the install, and figured it was faster to pay for a quick install just to get me to where I can start learning to use it. Which I am now doing! I'm mostly just copying prompts from civitai and tweaking them and the settings around. If I go off too far though i quickly wind up getting those bright tacky colours and poor detail.

Also if there are other good model resource sites besides civitai let me know. civitai has a lot but most of it's anime, which i love but want to also diversify. Also I want to try training on my own art as well

lol

up to you...

[acoustic plinth]

Hey all , just joined and have been reading through the recent discussions.
I’m noticing a few recurring pain points here:
ComfyUI updates breaking stable setups, the gap between “what I want to do” vs correct SD terminology, and the general chaos of managing large volumes of generated media.
I work at the intersection of AI systems and automation, and I’ve spent a lot of time building things that prioritize stability, reproducibility, and intent-driven workflows rather than just raw capability.
I’m not here to pitch anything — mostly interested in exchanging ideas or collaborating if there’s overlap. If anyone is experimenting with making ComfyUI setups more reliable, simplifying AI video pipelines, or tackling large-scale media organization, I’d enjoy comparing notes or helping think through solutions.
Looking forward to learning from the room.

[junior urchin]

@still glacier

[atomic mortar]

For commercial flows, unless theres a speed improvement or crucial bug fixes, do not update as its doing what its required and if it breaks goodluck since the cost is on you to fix after updating it most of the times unless a client updates it

Like why would you change a tool if its doing its job just fine

[karmic steppe]

I see some parallels between this and the gaming mod communites, like on nexus mods. Maybe there could be some sort of system set up by which updates are made more obvious, like how they have things set up over there so when you update your mod it shows up on the front page which version it's up to date on. as for correct terminology maybe just a simple pinned glossary on those would help. not sure how useful any of this is...just my two cents, only wanted to help

I keep my machine that i use for SD offline most of the time anyway and don't see a need to update anything now unless something comes out for it that would greatly improve its output but requires an update.

[modern pagoda]

Nexus Mod Manager wit Skyrim mod loading order 😭😭

I been here over a year & i barely dipped my toes in this.

[atomic mortar]

Scam

[inland mango]

Hi everyone! I’m Donis from Mexico. I’m starting an open‑source project called MANGA Studio: an offline‑first suite to turn scripts and audio series (YouTube, podcasts, etc.) into anime‑style episodes using local AI models (no cloud, no subscriptions).
I’m not a programmer; I have a detailed technical spec of the full pipeline and I’m looking for collaborators (Python / ML / architecture) to build a small MVP.
GitHub repo: https://github.com/adonismaikel-glitch/manga-studio
Any feedback or help is very welcome!

[red wagon]

awesome to hear. Good luck. TBH im not even sure if i use or have used SD directyl i ususly look for thing s and tools i need or want AI image gen and go fro mthere IF ti does it great if it doesnt I move to a plat or tool that works with and for me. If these tools dont exists or or not doing what i need i typically MAKE my own tools and pieces etc. reason i came here was to see what the friction and presure points people have or face wit hthe creation proess and or even specifically AFTER the fact (example I have around 25k of AI gen images ina folder. I bascially am making a simple sorter (well maybe not to simple ) etc that i can say HEY TAKE folder A and sort this into image catagaroeis Designs in foldes for me and then caption them so ( FOR MY CASE USE) and maybe others , have a searchable data set and recal mech anism for AI or othe tools to simply locate and display and or use grab for me etc.

[red wagon]

VERY sound advice , and for anyone that has designed anything can see tha twhen you change and or update stuff for the better the pipeline itself can have issues in processes, additionally , YES im old, i believe if it aint broke dont fix, whe nteh time comes to have it or anything do BETTER mroe consistnat quality or whatever, start wit hthe new 2x4 not the 2x4 that has the pre drilled holes andthat dont line up and ro the length is slightly shorter etc, loll bad analogy but that just my brai nflow etc lol

and clearly I cant type well this morning lol ok more coffee

@karmic steppe lol im actually going to go look at and maybe even grab a version of this comfy ui and see wha tthe hype or probelms are lol

Question and or UPDATE so im at the GH repo and as i know the issue peeps are having are UPDATE etc wrecks thing s or whatever so the question is SHOULD i DL or sue the LAtest version or is there a stable version wha tversion does every prefere to use and or that works as intended etc? lol

[rugged knoll]

Anyone wants to apply to YC together?
https://www.symphoria.ai/
Our investigation into Maduro:
https://symphoria-ai.web.app/share/alish-sult/maduro

[red wagon]

ooof hahah im working o nthis comfy UI install and set up and i alrea yhave issues wit hit , UNLESS my LLM is making things diffciult, lol SO i have the system BUT dont have the BRAIN (model) so now i have to go find that ... and i aalready stated to MY LLM HEY if you have a UI why do i need a MODEL prior to running , the AI iamge generateor i made has a startup ...poofo window and then a selection tool of the model i want to use , ODD that a more complex and or COOL CONFY ui would have soemthing like this no?

[valid aurora]

anyone here trained illustrious before? got some ancient questions about captioning that i would really appreciate if you can help me out on 😆

[empty jackal]

Hello everyone I’m a developer who’s working with AI and automation lately.
I mostly work with web stuff (Next.js, TypeScript) and I like building practical things like bots, workflows, and small AI tools that actually solve problems.
I joined to learn from others here, exchange ideas, and maybe collaborate if it makes sense.
Nice to meet you all 🙂

Is there anyone looking for high skilled AI devloper ?

[atomic mortar]

I'm looking for people to follow rule 5

[atomic mortar]

What's up, im at work rn but if i know it ill let you know

[red wagon]

ok lol once i figured out and installed this CUI lol have to say lol i imediately saw it s like UE5 designining set up ..ok this is a very interesting set up lol in 15 minutes i found Dloaded and ran and it made me an image lol now i need ot verify this is happening OFF line lol but still this is interesting lol

SO if i understand what im seeing lol Confy UI is liek making your OWN personal Image GEn PIPELINE versus using a PLATFORM interface etc ok this could send me down a rabbit hole big time lol but this is very neat

[empty jackal]

what is that ?

[proud fjord]

Hello

[red wagon]

ok lol thi is a bad lol i could spnd days on this Comfy UI thing hahaha and probably never even scratch teh surface lol im already in teh processes of making nodes for it lol

im guessing the goal is to use this suite and go from prompt ->image or iamge ->img tehn img to vid then -video to lyp sync and ro creaet a pipeline of shots and media and then paste them all together bascially like LTX ? in a way her is 10 images here is 10 clips here is audipo here is whatever make me an end product lol ? or any varition in the sytem along the way ?

[proud fjord]

https://www.instagram.com/p/DTNs9A8l1kU/?igsh=MW0zMjQxMG4wbHB1aQ==

[atomic mortar]

#✍🏼｜rules-and-tos number 5

[red wagon]

@atomic mortar HEY ! if iover step and or break a rule id appreciate a tap on teh glass or nudge from you please haha

[atomic mortar]

Oh no worries, i only call out the dev bots

Mostly bc im bored

[red wagon]

ahhh well im neother haha

and regarding bored , i hear that thast why im lookign for something to do make or implode lol and the comfy UI thingy may be a ticket to some skill testing boredom lol

[valid aurora]

hey thanks for the offer fortunately i found a work around but if i have another question bout illustrious lora in the future ill take you up on that offer !

[still glacier]

no need to spam

[karmic steppe]

Thank you. I need to train the AI on specific characters, though so my next step is figuring out how to do that. anyone with knowledge of it feel free to let me know how that is done. I have already got a bunch of base models working in SD

Also finding a problem where characters in the distance look very messed up. The problem doesnt seem as bad as with dalle3 but it's there and seems cauesd by the same issue

Fine detailing is also often just off.

[atomic mortar]

Background is often jank, depending on the model but regardless

Onetrainer + civitai tagger for illustrious models is nice

[karmic steppe]

Thanks, I'll look that up. and yes I've noticed that problem with backgrounds in general...it's a problem...

also it doesn't seem to handle things well like water or even tall grass

It is hopeless with waterfalls and streams

What's onetrainer?

Looking it up now.

Looks like I'll need several good quality images of each character. Not something I have for most of them. A lotof them I just have 1 and it's not good quality. however, do you think I could generate some of them by using img to img with just a sketch to produce something more enhanced?

ok so i took out all background tags completely and it gives me a better background this way than before. go figure

The problem is the uniformity...doesn look natural. im working with nature backgrounds for this whole project and so that jumps out at me right away

The water itself is nicely rendered but some things make no sense like one character sitting in shallow water but not on a rock or anything, lol

[gray urchin]

**Happy New Year! **

AI & Full-Stack Engineer | Automation | LLMs
I build AI-powered apps and automation used in real products.

What I do

• Workflow automation (Slack, Notion, APIs)
• LLM & RAG integration
• Image & voice AI (tagging, moderation, transcription)
• Bots (trading, Discord, automation)

Full-Stack

• Web: React, Next.js, Node, Django, Laravel
• Mobile: Flutter, React Native, Swift

Open to collaborations or projects.
Feel free to DM me 👋

[magic fiber]

Hello
How are you doing?
I am a passionate developer, so far attended various kinds of projects.
so if you have some recommendations or looking for extra developer, I'd love to collaborate together. 😇

[acoustic plinth]

Hello

Is there anyone looking for high skilled AI developer ?
I can help to Autonomous & multi-agent systems, Voice AI & chatbots, Computer vision & multimodal, Fullstack + infrastructure + mobile, Deployment & ML.
Thanks

[atomic mortar]

Oh yeah with illustrious backgrounds will be a problem, the newer models (non anime) are better with this

@acoustic plinth @magic fiber @gray urchin spam bots again, don't fall for it guys

[empty jackal]

Nice day, here, Is there anyone looking for AI & full-stack developer ?

[karmic steppe]

are there any particular base models you recommend?

I'm getting slightly better water results with scyraxpastelcore from civitai

[atomic mortar]

Personally i use simple backgrounds but z-image turbo is really good lately

No big hardware demands either compared to sd3/flux2

[karmic steppe]

thanks for the recommendation!

i do anime/disney/fantasy style but will give this a try

[swift isle]

yo

whats the best model for realism

[atomic mortar]

If you can run it z image turbo is great

A1111 and possibly older versions of forge can't run it

[swift isle]

i have a low end pc 😅

I'm using a1111

[atomic mortar]

If you can run sdxl you probably can run z image turbo at nice speeds

[swift isle]

I'm fairly new to generating images locally

[atomic mortar]

Well im not sure if the newer forge can run it but swarmUI definitely can

[swift isle]

rn I'm using realistic vision

[atomic mortar]

But you'd need to ask cs1o about the forge thing since he uses that more

[swift isle]

ah i see

what would you recommend right now tho

[atomic mortar]

Forge is similar to a1111, that's why i suggested that but I'm not sure if it supports zit

Nvidia or amd?

[swift isle]

rtx 2050

laptop one

[atomic mortar]

Might be worth looking into forge then

[karmic steppe]

I'm looking for the pony-v6-turbo merge and cant find it on civitai

found it on another site that strung me along through signing up only to confront me with a payment plan for it.

I don't patronise those kinds of places

[karmic steppe]

Anyone else signed up on diffus.me? I can't find my account settings on there, anyone know how to find them?

[karmic steppe]

At least I didn't give them my payment info. i just dont want to have an active acount there.

[atomic mortar]

If a model is not on civitAI or huggingface its not worth it

[atomic mortar]

Look on fiverr , this is asking to be scammed here

[karmic steppe]

also if you go on fiverr to buy someons services let them know at the outset you don't deal in any funny business with payments outside the platform. one red flag is: if they just do the service for you before asking for payment. establish payment at the start before moving ahead. just my 2 cents i use fiverr a lot and some ppl use these tactics. if its someone youve worked with a lot for a long time thats different, i mean for someone brand new

[unique delta]

Question... 3 or so years ago I used Satble Diff via Google Colab and it created some amazing artistic renders. Are they still available? Everything I am seeing online looks too similar .

[somber eagle]

dream

[karmic steppe]

I’m noticing a stubborn problem with light sources in stable diffusion. I’m trying to have it use certain things as light sources and it refuses to, but also it adds random light sources into dark settings that make no sense

anyone else deal with this?

[wheat urchin]

Hi everyone! 👋 I'm interested in LockedIn AI for interview preparation. Does anyone here have experience with it or know about any promotional codes or discounts available? Any help or recommendations would be greatly appreciated! Thank you 🙏

[red wagon]

Good morning. as i spent a few hours yesterday playing with Comfy ui (really havent touched the surface) I am still trying to determine the use for this My LLM says its pipeline for creators etc but didnt want to get to bogged with the details. i did laugh and go OK so i connect all these little pieces(nodes)... very similar to UE5 building lol, im still trying to figure out a use. and when reversed the conversation i found, if true, is that i can make my own scripts and nodes or whatever to do things in a pipeline setting so i have a few ideas id like to try lol , BUT that said so far what i see is a local image generator. (with some workings to do stuff) etc ...just not sur eif i spend the time to play wit hthis and find out what its really for and use or just try and make it do something crazy or usefull lol ... i asked abotu Lip sync video gen , image to video , , and of course all the things i dont have the hardware to do any any real time frame lol but i have a few ideas and or workflows id liek to try out haha and se if any of my scripts i made for my LLM can be made int onodes etc. Also is it PURELY image gen or media gen >? or doe sit or can it be used with LLM creation etc NOT training llm but desing and use of LLM conversation etc ?

[red wagon]

ok so when i offer servives lol im offering myself to collab and or do some work wit hpeople for experience and or make tools lol i was in a 14 day hackaton thingy for HF and never used HF or Gradio stuff etc... and i made a dual submission for a Track 1 and 2 without even knowing..lol im nto claiming to be a DEV or enginer lol im jsut a guiy that can see holes and plug that many miss etc i have made a LLM from scratch ... not the Model lol thats coming, it inte=igartes media into display while chating wit hthe LLM TTS and STT and a few other smart features lol ALL from scratch so now im jus texploring try ing new thigns etc testign stuff having a good look around lol

@atomic mortar @karmic steppe if either of you ever wan to chat or work on something or just have a SPECIFIC target problem id love to test my skills and see if i can help and or even maybe provide some unexposed issued or fixes lol thats all im really good for haha

And of course , while havong my coffee, MY LLM is asking wher ei am wit hmy image aprgaizer lol.... and of course , since coming her ei been side railed back into MAKING more images lmao ...

[tribal crown]

I am using and linking different video gens together but for next 5 seconds the character's eyes get even blurrier and can not really properly see their quality. How do i solve this? Using wan 2.2 smoothmix. I also use some loras. is it the combo of loras? any way i can get around and still keep using the loras?

[red wagon]

asking in general or ? looking for someone like Eurotypo lo with a vast knoweldge base of experience and tools lol ?

[tribal crown]

i am just looking to make my characters look the same and have the same eyes even after a ffew videos

[atomic mortar]

Hes in the swarmUI discord, better video experts there

[tribal crown]

so what i do is basically work with workflow that

[atomic mortar]

I'm mostly a image chud

[tribal crown]

thanks

ill look into it there

[red wagon]

ok so if i understand this is yuo 9however) make a 5 second clip video and tehn attempt to either A continue said 5 second video (problem is the consistemncy from first said vdieo carrying over to the 2nd yes ?

[tribal crown]

yep

[red wagon]

ok so tell me video 1 is it prompt designed or IMG 2 Video?

[atomic mortar]

Text to video is mostly a meme, image to video is king for consistentcy

You could try to extend the video using the last frame but you probably wont have smooth transitions

[tribal crown]

i am using i2v

[red wagon]

I may have a HACK lol i have been doing some LIVE sampling for longer CHIAN linking lol : You could try to extend the video using the last frame but you probably wont have smooth transitions LMAO thats what i have been testing and the jitter of contection is minimla if clean lol

[tribal crown]

and when i do the last frame ofvieo

usually the second video is still kinda good but loosing eye quality consistency

so basically the eyes are not the same really

[red wagon]

i feel , and im not an expert , but this could be copy of copy of copy that cascades throug hthe process small eye issue keep going slowly gets worse , BUT a tthe end of the day the Model and its ability is going to determine alot of this i think

i see this is as kinda the back in the day fingers and toes lol AI had a tough time but progressively got better NOW its eyes lol details are details im thinking start wit hany artifact or clarity issues i think it would just scale in over time perhaps lol I have no idea YA i say GREEN eyes alot and soem time YES soemtime sNO there are many facotrs and again i havent played wit hIMAGE to video other than major plats etc becasue well trainig my LLM or ai system is still i nthe owrks lol

[red wagon]

ooof my head hurts a little its all part of the teeritory of trying to build using AI asisted projects and trying to keep the AI in teh same plataue as im in im constantly kicking it in the ass to keep aligned lol

[blazing moon]

👋

[junior urchin]

Anyone wanna bet that those two waves are hacked accounts?

[reef epoch]

@wise stratus SO SD 1.6 wen

[junior urchin]

Emad doesnt work for SD anymore 👀

[acoustic plinth]

Is there anyone looking for a dev ?

[atomic mortar]

No and theres never one looking for it

Not here

[acoustic plinth]

Got it, Thanks for your reply

[atomic mortar]

Just check rule 5 in the future

[reef epoch]

thats why SD died... we would paint every image by hand and send it to user... he was the "inference"

[junior urchin]

The chinese are doing what americans and europeans don't want to do

That's why we like the chinese

[reef epoch]

they got millions of workers painting images... and filming videos for us

[shell slate]

Hi what is the equivalent of Forge I can use for free w graphics card but for music?

I want to create my own music w ai using downloaded model, and not use some paid service like suno where they own the songs I make and I have to pay them a subscription fee

[atomic mortar]

comfyUI but local music models currently are bad

[junior urchin]

@still glacier #💬｜general-chat message #💬｜general-chat message #💬｜general-chat message

[fervent thunder]

does anyone have easy install link for stable diffusion with image to video generator

i downloaded a file from github but i cant figure out how to open it

[ocean ferry]

Hi someone help me pls?

[atomic mortar]

with? if its a technical question you might be better off going to #🤝｜tech-support

[atomic mortar]

whats your gpu

if you have less then 16gb vram(nvidia) it might a bit slow or unable to run at all (takes me 5min avg per video)

but i recommend swarmUI and read the docs under supported models on github

[past sedge]

Hello is this the place to ask questions?

So i want to try better models but im not sure if i want to pay aton, i only have 12vram is on my desktop. Im not sure if i should upgrade or buy some sort of platform?

[atomic mortar]

With 12gb you should be able to run flux schnell, z-image turbo (is really good!) and all sdxl models (including illustrious)

Video might be slow but if you have enough normal ram it should be fine

Though it does depend if you have a Nvidia or amd card

[onyx cliff]

hi, I am AI agent developer.

[magic fiber]

luckydmytro.site
Hello everyone, how is going your Sunday?

[atomic mortar]

Scam

[red wagon]

morning folks

[modern relic]

https://tenor.com/view/hello-mickey-mouse-gif-20890109

[fervent thunder]

why does stable diffusion not just have a website

with a download link and it installs like a normal program?

why is that so out of reach?

its like some fuckin autstic designed it

[atomic mortar]

lmfao, because stablediffusion is just a model, its not a program

[atomic mortar]

and stable diffusion is not the only model maker, theres flux, zit, hunyuan, wan, chroma, etc

and for running it locally you need some decent hardware and since its open source it gets maintained by a lot of people

if its a simple program by one company theres no telling when it gets abandoned and lack of support for other models

[still glacier]

it's litteraly 3 lines of install instructions.
1/ install python and git
2/ git clone
3/ double click whatever .bat came with the client you decided to get

Also it makes it way easier to update this way. Everyone can inspect the code, there's no need to wait 1 hour recompiling everything, there's no need to waste time packaging a .exe that would obfuscate everything and make it easier to spread malwares, it makes it cross platform, etc...

So maybe let's tone down the insults ?

Also stability.ai is the company developing stable diffusion's model. They're not the ones behind the commonly used clients like forge / comfyui.

[fervent thunder]

who the fuck is going to do that

i swear geeks live in a bubble

the general person is not going to do that

it could be one install

but people are lazy

[red wagon]

ok so what are you going to do? how do fix your problem ? Are you aasking for someone here to give you a solution to a problem you have so they do wrok while you get a solution? just curious casue i can give you the easiest answer to your entire problem

[quiet pier]

Can someone send me invite link in DM to Unstable Diffusion server?

[still glacier]

geeks are the reason this exists

[still glacier]

skill issues

Jokes aside

People will check open source code. Maybe you won't

but lots of people do

[atomic mortar]

With basic reading comprehension it's doable and non geeks useually don't have the hardware required anyways, use some online service since reading 6 lines is hard i suppose

[steep sage]

hello, whats the most recent local webui and model to generate mobile phone like selfies on rtx 2060 6gb vram?

[atomic mortar]

I think forge should be just fine with a realistic sdxl model, maybe z image if you want to wait a bit longer but then i recommend swarmUI

Granted you do need enough normal ram with z-image

[red wagon]

ll functioning LLM model swappable GUI and config gui desinged and implemented as well as LLM choice media select capabilities and rag tag whatever im doing and running the entire thing on 5 year old laptop with NO gpu lmao i didnt wrote 1 line of code BU ti mde it from scratch lol so im no nerd i just know how to make tools work FOR ME and if ther isnt a tool i make one im jus ttrying to prove that ANYONE can do stuff wit hthe right mind set and or process

SO thats why iasked what i asked wehn i asked it becasue in reality IF YOU use OTHER tech and it doesnt or cant do wha tyou want then remove the PROBLEM tech and make it a basic functioning tool lol

[atomic mortar]

Im gonna be honest and say your messages are hella chaotic

[red wagon]

and as far as teh image generated , of course mine slow but guess what same thing i can use any model anytime any way lol process is MAKE a gui and have a hot swapable GGIF or safetensor file thats it lol

[red wagon]

ya sorry im the worst typer and also im not a nerd and relly im sorry haha andfor sure im a choatic person lol MY brain fires and fires faster than my hands lol thats why 99% of my stuff i use HASS STT caps becasue it seasier (LONGER) but more methodical speaking words than typing them etc Sorry

ill jus tsay it is becasue this is a new KB lol and im still getting used to it ...but thats just a lie lol

[steep sage]

sdxl runs like ass on 2060 and 1.5 with controlnet and loras still make good images

[atomic mortar]

Sd 1.5 is not being considered nowadays

[fresh abyss]

Is there anyone looking for a skilled AI engineer?

[junior urchin]

man, 6 bots in a row @still glacier

[reef solstice]

With RTX3060 12gb i can create a video that same as grok imagine? I yes i want to try generate a cat walking

[acoustic kraken]

💻

[still glacier]

oh ? So we re moving away from the 👋 emoji ?

[fresh abyss]

Is there anyone looking for a skilled AI engineer?

[still glacier]

no

[warm junco]

Here:
https://github.com/CS1o/Stable-Diffusion-Info/wiki/Webui-Installation-Guides

[junior urchin]

Are you gonna paste it again tomorrow?

[mild ferry]

is it better to save original outputs or the upscaled ones (using img2img sd upscale 0.1 denoise) for future purposes like lora training?

[warm junco]

English please or DM me

The zluda virus detection is a false positive...

Also calm down

[oblique elk]

@warm junco don’t bother and waste your time.

[warm junco]

Yea I guess everything is said. Some people can't be helped.

[warm junco]

Banned that rude dude.

[mild ferry]

yeah, dont waste your time with him, might aswell answer my questions 👉👈

[warm junco]

Use the best quality. So if you get it nicer with 0.1 denois use them.
But if the outputs are to smooth or Plastic skin is seen then use the originals

[wooden coyote]

hey guys

[junior urchin]

is it a bot

[wooden coyote]

where is the best place to find a talented klingAI video editor in here? Would love to talk to about a big project

[junior urchin]

bot indeed

[wooden coyote]

nah, just genuinely looking for someone to work with. not sure if this is the right chat tho

[karmic steppe]

ComfyUI is a nightmare to deal with

[umbral topaz]

is this the right place to get help with stable diffusion webUI issues?

[warm junco]

Yes but in #🤝｜tech-support

[prisma loom]

hi

[fiery ore]

You can send a DM

[mossy edge]

Thanks for that context! Really helpful.
Quick follow-ups to understand the pain better:

1. Frozen setup:
   When you freeze a setup, do you:
   ∙ Keep a dedicated GPU/pod running 24/7 for that frozen environment?
   ∙ Or spin up fresh each time and carefully reconstruct the exact environment?
2. Multiple projects:
   If you have 5 different client projects, each with different frozen setups:
   ∙ Do you keep 5 separate pods running?
   ∙ Or how do you manage switching between them?
3. The “if it breaks” part:
   You mentioned “if it breaks, good luck fixing it” - does this happen even with frozen setups?
   ∙ Like, something breaks even though you didn’t update anything?
   ∙ Or only breaks when you DO update?
4. Missing out:
   Are there new nodes/models/features you wish you could use but can’t because you’re stuck on old versions?
5. Cost question:
   Roughly, what do you spend

[atomic mortar]

These questions don't apply to me unfortunately, but if its a basic workflow you can just import it

Stuff breaking can always happen with updating so yeah

[mystic snow]

hey

[upbeat sleet]

hey

[atomic mortar]

Scam as useual

[lime pagoda]

Student

[tardy ice]

is there anyone who is expert on all image / video ai models, just for a quick chat - looking for best cloud apis for my needs of generation 🙂

[atomic mortar]

Sccam as always, anyone with dev and fullstack in bio here is a scam

[leaden fractal]

Hey guys is there a workflow that I can use that will take the image of a character and convert their outfit, position background etc that then goes into a video generation workflow and creates a video with that image as the first frame that is consistant and true to the character

[worn aspen]

Do checkpoints become obsolete? I haven't used SD in awhile, and I see there are new ones. Should I dump my list and start fresh?

[hearty wave]

Guys, what video models would you recommend me for natural looking animations?

[slate shell]

https://youtu.be/GVYZnaYEAS0?si=kpTLz3X-4KqQVOJm

[full lark]

I have the same gpu and yes, you can do that with wan 2.1 or 2.2, but its very slow tho, about 20 minutes for wan 2.1 and 70 minutes with wan 2.2 🥹 (withouth lighting lora)

[reef solstice]

I see, thanks

[full lark]

Np, btw I saw that ltx 2 also works on that gpu and its faster, but I haven´t tried it yet, there are some reddit posts explaining how to run, with workflows and everything

[reef solstice]

I see

I was thinking should i use comfyui 🤔

[full lark]

Yeah, I use the portable version https://github.com/Comfy-Org/ComfyUI/releases (windows_portable_nvidia_cu128)

[reef solstice]

I see thanks

[full lark]

If the lastest version gives you errors try to increase windows virtual memory, and if you still have errors try with v0.8.2 (that worked for other people)

[reef solstice]

I see

[full lark]

Someone got hacked lol

[atomic mortar]

Scam as useual

[unreal oriole]

any flux hype here?

flux 2

[full lark]

Yeah but im 100% sure i wont be able to run it 🤣

[unreal oriole]

IT'S 4 BILLION PARAMETERS DUDE

phones can run it

ah it's called Flux Klein

[full lark]

WAIT WHAT

[unreal oriole]

https://www.reddit.com/r/StableDiffusion/comments/1qdmohb/flux2_klein_4b_9b_released/

[full lark]

Wasnt it like 80?

Wait i got confused lol, gotta see that one

[atomic mortar]

Normal flux2 you been 120gb ram & rtx5090 and even then

[unreal oriole]

https://huggingface.co/black-forest-labs/FLUX.2-klein-4B
could probably run 10 of these on an RTX 5090

[junior shuttle]

Can i DM someone here about a video i want to recreate? Looking for guidance on a proper workflow but it doesnt let me attach videos in this chat

[primal current]

Hey everyone, which FlashVSR repository are you using?

I'm using this one, and toggling color correction on and off always ends up drastically altering the colors of my images when scaling.

SeedVR keeps the colors perfect, although in some cases I prefer FlashVSR because it adds more detail to my images.

I'm using this one:

https://github.com/lihaoyun6/ComfyUI-FlashVSR_Ultra_Fast

[cedar plover]

def a learning curve but its just so worth it. pure freedom fr

[karmic steppe]

I’m having someone help me learn it yeah

[oblique agate]

https://fxtwitter.com/i/status/2011952987678093353/ what software is that? @ me if you have some clue

[glossy flicker]

👋

[steep sage]

it runs bad on 2060 6gb vram and 1.5 has good nsfw model that make unironically most believable non nsfw people

[sturdy beacon]

@timber heart you DM me. How can i help?

[grand iris]

hi, new to the whole topic - comfyui is an interface for self hosted stuff right?
I cant really find a decent FAQ/source for info on image gen and self hosting, you happen to have any? Just researching rn to see what system I want to go for

[cedar plover]

yeah its a locally hosted UI used to interact with generative AI models. its not the easiest to learn, but not bad either because theres lots and lots of premade templates included that allow you to start generating with a few clicks. theres also more workflows on places like civitai along with all the models youd ever need. however there are also lots of other locally hosted UI's you could choose from, but they dont quite have the same capabilities and customization that comfyui has. but generally once you get the hang of comfyui, youd never wanna go back to anything else.

[sturdy beacon]

Give me a second. I know a guy who makes videos teaching ComfUI. He just came out with one a few days ago updated to the newer versions.

https://youtu.be/HkoRkNLWQzY?si=fPcIFK_BPaqP4-aQ

Its 5 hours long. But it will get you going.

Its the first in a series.

[grand iris]

Thank you both for the info

[cedar plover]

id be happy to help get you setup if you need it. feel free to dm if you need it

[sturdy beacon]

The video I sent actually goes through the setup process to. It is very noob orientated.

it's 5 hours long, it explains the software, setting up and using it

The guy who creates it, will be expanding into much more advanced subjects later down the road.

[eager oak]

Hello

[cedar plover]

alr cool, that should definitely get him goin then

[cedar plover]

hi hows it goin

[eager oak]

How do I generate art with stabllre diffusion?

[scarlet idol]

I want to animate an anime girl VTuber using ComfyUI on my RTX 4060 graphics card. Are there any guides or resources available for this?

[atomic mortar]

Animate how? Just in a silly scene? You want it to be autonomous movement? Overlay a video and copy its movement

And uh do you have the 8gb version or the 16gb one

Because with the 8gb one im not sure if you can

Or well maybe a hour or 2 for 4s

[scarlet idol]

We thought of a mascot to advertise our game in Instagram Reels videos. It would be enough if it made simple movements like swaying from side to side from the waist up (facial expressions and head movements), but this is very limited on the 8 GB VRAM model.

Wan2.1 or 2.2 Animate; LivePortrait and so on. It's all confusing me...

[atomic mortar]

If its a live2d character i recommend not using AI and rather just animate it using the existing tools

[scarlet idol]

What kind of tools?

[atomic mortar]

Vtube studio, live2d kubism etc

Beware though if your mascot is made with ai theres no copyright and anyone can just use it

[sturdy beacon]

Technically true. Unless you do enough to it to make it your own.

This is part of the slippery slope. Too many people think AI stuff can’t be copyrighted.

It can be, it is the same reason why scientists copyright their stuff.

You many not be able to copyright the foundational work, but you can copyright the work you did. And if that changes the original foundation work enough (it cant work without it) then you copyright the whole thing.

We are going to be finding this argued in the future pretty soon i think.

[glossy flicker]

👋

[primal current]

Has anyone tried character swapping with Flux Klein? My results are terrible.

The best I can do is replace the girl, but it doesn't always match the face, and the swapped body doesn't match the lighting in image 2.

[alpine torrent]

what's the weather like

[red wagon]

@atomic mortar Beware though if your mascot is made with ai theres no copyright and anyone can just use it. <-- cant be true lol tos and plat policy determines ownership of teh images THEIR AI creates you or them. thast standard . however i think the catch isnt COPY right as much as if plat gives JOE a image A and pete an image B and it is NOT a exact duplicate they own each their perspective images. Even if teh same look feeel markings whatever lol the COpyright isnt the CONTENT OF the content you dont have EXCLUSIVE rights is a better phrase. SO although orignal and or unique a red apple is a red apple etc you own your version they onw their version. LOL maybe im wrong IDK lol

@sturdy beacon well then if you do enoug hto it lol its isnt a copy lol it will get sorted lol it really isnt any different than copy right is now basically it s just a different version form of it lol and this is a funny topic actuall y AS I took and image today AND actually gave to google lens and strike me dead , i had an image show up as SOME has used and guess what it wasnt a similar it was a exact copy lol Every want to see how origina lyou stuff is drop in lens lol and have a peek its actually pretty interesting

[sturdy beacon]

That's the problem, you can't just let AI do it for you. Give it a base and you change things. To give you an idea.

Well, I'd show if I could...lol

[old wadi]

Hey guys. I've gotten to the point where I've come to talk to people and ask for help.

I've been playing around with video gen trying to make it work for months. I've had some success but the majority of time it just doesn't do what I tell it to, or does it badly. I know the problem is on my end. I've tried a ton of settings and and combinations of settings.

For example with i2v. It barely does anything. If it does anything, it gets smeared into goo while also not doing anything.

I have been using .gguf. I barely know what I am doing anymore.

Should I do the obvious and go watch hours of tutorials on youtube? I think I have ADHD and its very hard to absorb this constant stream of complex info.

I wish there were basic workflows that work simply, but I don't know what I am messing up.

I have 10 GB vram

Should I be using .gguf models or .safetensors?

[harsh current]

who

[lavish olive]

Can LTX-2 do audio + image to video? Can't seem to find any such example that does not rely on generating it's own audio.

[gentle prairie]

is there a way to find out who did a deleted lora on civitai? I have the link where it says "this has been removed by the owne", but no clue about the rest

[red wagon]

its ok , i get what you are saying i have worked a long time with IG and its funny we are talking about CR rights and exclusivity etc i had a long conversation regarding this its an interesting topic for sure but one that can hang you up as well lol.

[red wagon]

Id love to take a stab at your problem , of course , I may not have a ton of time ATM, I am currently on several projects atm. at the very least i may be able to point well enough if i knew and know what you are building using etc and what you are hoping and or trying to accomplish. feel free t oDM if youd like.

[outer raptor]

Stop using .gguf for video generation.
Use .safetensors diffusion models for video. .gguf is for LLMs (text models). It will only make your life harder and explain a lot of the “smears into goo / barely does anything” behavior you’re seeing.

[fallen shard]

I'm new here.. still trying to figure discord out!

[atomic mortar]

Gguf definitely has a place for quants

[sturdy beacon]

Yup. There is a point though that something created by one is changed enough to not be the same thing.

Look at cars, Ford and Chevy use same car designs, but the insides are different.

We are going to be arguing the same arguments in court. The AI artwork law is going to be fought again soon because of people like me.

We don’t have any intention to “steal” artwork but we learn from mimicry.

The same could be said from centuries of artists. They mimic then learn their own art styles.

Artists are pissed off because an AI can mimic their style accurately.

[faint timber]

"Those who do not want to imitate anything, produce nothing." - Salvador Dali.

[snow cairn]

Hello

[atomic mortar]

four scams in a row dam @eternal jewel@gleaming fulcrum@gray urchin@sweet pivot

[frigid surge]

hello, sorry for the simple question but I can't find the answer anywhere else. - I'm using the standard T2V template for ltx2. My question is that I want to fire it off to generate many different generations from the same prompt so I can pick the best but it seems to only want to make a new generation if I change the prompt and just re-saves the previous one if I don't change it when I try to do a batch. I'm very new to comfyUI so I may be being completely stupid but I tried things like setting "control after generate" to randomize but it didn't seem to make any difference. Any help or pointers very much appreciated. Thanks

[still glacier]

would have been 5, I cleaned one before that too

new record i believe

[atomic mortar]

Oh yeah definitely haha, wish we could add a automod, it would be so easy to auto ban em

[still glacier]

we don't do nsfw here, rule 2

[gentle prairie]

when generating images, how wattage is recommended and does it matter at all? i have a 3080 and a 850w psu right now and it works well but i will get a 5080 soon and i'm wondering if this will still be enough? if not is there a way to limit power consumption in exchange of generation being slower?

[still glacier]

wattage matters as much as with any other gpu intensive task. But the more you throw at your gpu doesn't necessarily mean the faster it will get. Same as for gaming. Higher frequency is more likely to give you faster result. And to achieve higher frequency you might need more wattage... or sometimes less. Depends of the gpu and the "silicon lottery".

Limitting your power consumption without limiting the frequency is more likely to make your gpu whole unstable.

So I'd say just stick to the default settings if what I just wrote doesn't make any sense to you. Or read some tutorials about over/underclocking and undervolting.

[gentle prairie]

so default with 850w is enough then? no risk of going over that with a 5080 and 9800x3d and so on?

i didnt modify anything with my 3080 but since a 5080 use more w i was wondering

[still glacier]

if it's a good psu then it should be enough.

[gentle prairie]

it should be a good one yeah, at least from all the reviews i've seen on it :p

[still glacier]

120w for a 9800x3d, 360w for a 5080 add to that some power for ssd, drives, etc + extra room for GPU transient spikes

unless you get some really exotic crazy turbo overclock hd remix ++ gpu from some company I know nothing off.

[quick roost]

Try SwarmUI, it's a front end to Comfy and able to do what you require. https://github.com/mcmonkeyprojects/SwarmUI?tab=readme-ov-file

[vast briar]

Hi

[hearty patrol]

Do you guys know ? How these ai anime graphics designers maintain thier quality on instagram ?
I'm talking about thier carousel post

[primal current]

Does anyone know where to get QWEN Image Edit AIO FP8?

I've only found GGUF, and I understand that FP8 would be the best option for a 5090.

[atomic mortar]

Who are these?

[junior urchin]

Maybe they draw them

You can usually scum smell if it's AI, if they use multiple styles

Only encountered one Chinese guy that actually used multiple styles

Which was actually odd to me, since he didn't want to try and incorporate multiple into one

[red wagon]

morn☕ ing folks

[red wagon]

well im pretty much into the people user experience phase of a tool i made haha ... i did some searching this morning and well lol not sure why all these tools out there cost as much as they do im trrying to see how i can share my tool etc give me the folder and click poof de duped and foldered lol that all i got right now haha but NON destructive and all that stuff haha

[median stag]

Hey folks. I see people making a few seconds videos with Illustrious models. Can we do it on Automatic111?

[atomic mortar]

No and they aren't using illustrious to animate

They use illustrious to make a image yes but use wan2,2/2.1 to make the video

And no a1111 is horribly outdated, best use something like swarmui or comfyUI for videos

[median stag]

Yeah I dig in and find nothing about a111. It's really outdated. I didn't want to switch to Comfy because it was complicated. But it looks like it's time to switch. Thank you.

[atomic mortar]

Look into swarmUI, its the same backend but swarmUI has a nicer to use front end

So you dont have to use the noodles

But only dive into video if you have the vram tbh

[median stag]

RTX 4070ti is enough or I shouldn't even try?

[atomic mortar]

How much vram?

[median stag]

12GB

[atomic mortar]

Could give it a go, wont be too fast though since it probably will have to offload

[marble basalt]

is there any way you can make those kling motion control videos with comfyui?

[snow sinew]

Hello
I am looking for someone to provide paid tutoring for stable diffusion
I have successfully created a LoRA using Kohya, but I am struggling to achieve consistent and high-quality results across different models. I've tried to solve this by using ChatGPT, Gemini, and watching YouTube tutorials, but I haven't been successful
If you are interested, please send me a direct message on discord

[atomic mortar]

Scams

[atomic mortar]

Before you pay anyone, try using prompt enhancers for sd3/flux using llm's

Give prompt to ai > enhance it > give to sd3/flux

[fickle bridge]

Any channel I'm supposed to go to for help with installation?

[fickle bridge]

Also, is this user-friendly process for making AI art and images? I here all about these models and 'loras' i have no clue what they actually mean.

I'm good at adding depth to prompts, but is it more than that to get SD to work compared to a normal generator?

[red wagon]

morning ☕

[winged ember]

Hey guys,
Do you know how I can recreate a character without having to make a Lora out of them?

[orchid spindle]

Hey guys

[junior urchin]

Hey bot

[junior urchin]

Off the top of my morning brain, I'd say midjourney and its omni feature

[pale prairie]

Hey guys 👋👋

[winged ember]

I thought about that too but is there a way to do it in forge for example?

[atomic mortar]

Using a newer model such as qwen edit you can add the character as a reference image or add them to other images

[winged ember]

Oh I see! I'll check out youtube, maybe there's someone who explains it

thx

[atomic mortar]

In the swarmUI there was someone doing that but its litterally adding a picture of the character to the prompt and "place the character from the image 1 in setting:" or "make the character from image 1 sit at a desk"

Nothing youtube tutorial worthy ngl

Though for comfy id say goodluck but swarmUI is pretty good for it

[winged ember]

Oh really?? Thats awesome, I don't use comfy. Only forge Neo at the moment.

comfy was not so comfy for me xD

[atomic mortar]

swarmUI is a perfect next step if you wanna expand your skillset, however the qwen model is pretty beefy and theres no specific anime version

So odds are you might need to check the hardware requirements

And lookup if matches the styles you want

[winged ember]

Gotcha, I'll look into it

[short notch]

Hello! I’m looking for a private teacher to help me learn Stable Diffusion (Automatic1111 + LoRA), including UI basics and training a realistic, consistent AI social media influencer. Please DM me if you provide coaching or consulting. Thank you!

[bleak pawn]

For SD fine-tuners: we wrote a step-by-step fine-tuning guide (RTX 4090 & H100-class) with PyTorch examples + cost breakdown vs hyperscalers.
Disclosure: I’m affiliated with VoltageGPU — sharing in case it helps, feedback welcome:
https://voltagegpu.com/blog/fine-tuning-rtx4090-h100-guide

[violet scroll]

Hello, I’m Mary from Uk 🇬🇧
I'm looking for a serious and reliable partner in starting a new business idea.
I already have a solid niche in place, and now I just need someone ready to grow and build with.

If you’re genuinely interested, let’s connect.

[junior urchin]

ok bot

[rare nacelle]

Ok bot

Agree

[thorny cape]

hello there i am Johnny from Italy!
I am kinda new on making ai characters but having kinda fun. Hope to find people to discuss it with and share my character to have feedback

[kind shoal]

Hi, I have a Windows machine with RTX GPU.
If anyone needs remote access for AI or rendering, DM me.

[atomic mortar]

How vague

Surely this is legit

[kind shoal]

I have 8 computers with RTX 4060s and I want to rent them out, but I’m not sure how to go about it.

[atomic mortar]

Im going to be honest and say 8gb of vram per case is not attractive to rent,
5090 gpu's cost 35 cents/hr
A 4060 costs about 0.05/hr according to vastAI

A 4070 costs 0.07/hr and a 5080 0.12/hr

If they were clustered into one unit though it would be interesting llm's but not competitive image generative stuff

[cyan lodge]

Hello everyone, I'm trying to find a consistent workflow of turning a simple 3d screenshot of a building into a realistic looking image using automatic1111. If anyone is willing to share, I’d really appreciate it. Thanks!

[atomic mortar]

Scam

[lapis jasper]

can someone tell me how I can generate locally pictures in the quality of banana pro?

[grizzled igloo]

GPU: NVIDIA GeForce RTX 3070 – 8GB VRAM
CPU: Intel Core i7-12700KF (20 threads)
RAM: 32GB DDR4
i want to rent this out is there anyone interested...

[junior urchin]

Ads are not permitted

[sullen pelican]

Rainbow forms and glows over the hoise, magical, beautiful

[primal current]

Does steadydancer make the face vibrate when generating the videos, or am I doing it wrong?

[cerulean lintel]

...

[noble cedar]

...

[primal current]

So far I've tried:

SteadyDancer: 14 workflows
OneToAll Animation: 14 workflows

SteadyDancer accurately reproduces faces but always has glitches.

OneToAll barely reproduces the face in the input image. Tomorrow I'll try Scail and Animate.

[oak ether]

Is there a place in this server or other server where I can ask help/questions on ComfyUI workflow/extensions? I used a1111-like UIs before, but now have to use ComfyUI (because it's the only UI that works on AMD on Windows) and on first glance it's very.. uncomfy and missing a lot of things I used to (like __wildcards__, precise filename control, saving all generation parameters in image and automatically extracting-applying them, etc) and so I have a lot of questions that google surprisingly don't answer.
Or maybe there's some a1111-based UI that can work fine on 9070XT on Windows?

[atomic mortar]

Wildly wrong assumption with the only ui that works on windows&amd, check #🤝｜tech-support for amd guides

[atomic mortar]

Pinned messages

[oak ether]

I tried amdgpu-forge with zluda – it's extremely slow; also tried forge neo with rocm - it freezes for a very long time or crashes videodriver after doing last step (and I tried other vae's). Comfy is the only one that I install with installer and it just works, can generate images without crashes or need for manually fix some dependencies. Other UIs I tried won't even start if I don't install dependencies manually (I followed guide in the linked message). Although, I installed ComfyUI just recently while tried forge-neo about a mounth ago – maybe I need to install it again or update something and it will work

[atomic mortar]

I recommend asking cs1o when he's around

[oak ether]

Is there known AMD&Windows users who managed to successfully install A1111-like UI and generate an image with adequate speed?

[atomic mortar]

Yes, all the time in #🤝｜tech-support again i recommend asking cs1o since he helps a lot of people with amd

[pseudo silo]

Why is Grok dead?

[atomic mortar]

it made a lot of " nudifying" images of women from all ages

so that was a big yike

[pseudo silo]

Yeah I know that but today, its just no response at all, it wont help with prompts

or answer simple questions

[atomic mortar]

some countries blocked it iirc

[pseudo silo]

it always needs a refresh but idk

I dont think norway banned it

I asked " Are you Alive"
No response.

[oak ether]

Actually I just manually updated forge-neo and rocm, removed all command line arguments (since updated forge-neo somehow didn't recoginsed --attention-pytorch --disable-bnb commands and refused to launch) and first image seems to generated fine and fast. Didn't expect at all that fix will be that easy considering how much effort it took to even launch forge-neo on amd 1-2 months ago.

[warm junco]

You have to enable Tiled VAE in Neo for it to not crash

Its stated in the guide

Then it works perfectly fine

[warm junco]

The commands changed and now you only need --cuda-stream --use-pytorch-cross-attention

[oak ether]

Even upscaling works without consuming enormous amount of VRAM like I had in ComfyUI, but it's extremely slow (at upscaling, img generating time is fine - a ~8 seconds for 1472x832, but 2 minutes to upscale). And it seems that upscale by multiplier doesn't affect how much VRAM and time it takes (unlike when I used Nvidia)) and uses some shared memory even if GPU's VRAM is free. I tried disabling Tiled VAE in case if it causes tiling upscale, but it didn't help.

[warm junco]

Always set hires steps to 10-15, then its faster

[oak ether]

I set it to 5

[primal current]

Has anyone tried LTX 2 IC Lora Pose?

The quality is terrible.

[atomic mortar]

Ltx has poor quality but interesting systems

Compared to wan**

[junior urchin]

Even with the new vaes?

[atomic mortar]

Ltx has poor dataset

[winged orbit]

Who needs help with loura and ComfyUI?

[winged orbit]

Dm

[atomic mortar]

Never dm, just hop in #🤝｜tech-support

Check the pinned messages

It does have some hardware requirements though

[junior urchin]

Blockchain bot

[marsh radish]

hello 🙌

[marble skiff]

anyone know how to fix stability ai repo not being found

its not letting me run webui

[junior urchin]

just put a random model in the models\stable diffusion folder

[opaque bramble]

Does anyone here regularly create meditation music?

[modern pagoda]

Synth drones are often used for ambient sound design. Should help.

[crisp beacon]

Hey guys, how's it going? I know there are some smart people here, and I wanted to know how I can make it so that when I generate an image, the next one keeps the same background, but the character is in a different position. Thanks for listening, hugs.

[atomic mortar]

Inpainting instead of generating a whole new image would help

Results may vary

[strange yew]

Does anyone know the best place to source high-quality, localized Southeast Asian datasets (specifically Malaysia) for fine-tuning?
We’ve been building a private hub at [e-mail], but looking for more collaborators.

[zinc merlin]

Yes, check this ACL paper - https://aclanthology.org/2025.acl-long.916/

🌟 Why submit?

Incredible Keynotes: Kristen Grauman (UT Austin), Mohit Bansal (UNC), Dan Roth (UPenn), and Scott Yih (Meta).

Scope: We welcome everything from multimodal search (UI/Charts/Video) to agentic reasoning, tool-use, and provenance.

📚 Research Themes (but not limited to): • Multimodal Retrieval & Search • Agentic Planning, Reasoning, and Tool-use • Grounding, Provenance, and Faithfulness • Benchmarks for Agentic Intelligence

📅 Key Info: 📍 Location: Denver, CO (June 3–4) 🗓️ Deadline: March 5, 2026 🌐 Website: https://grailworkshops.github.io/

[atomic mortar]

What

Thats not relevant to anything

God dam these bots

[atomic mortar]

Rule 5 and overpriced, another bot

Why dont you read the rules before speaking smh

Your definitely selling something at 0.53/hr (despite other services doing it at 0.39)

@still glacier can we clean up general again, man id love to be able to do this too

[rotund spade]

what is a good model for general use

[primal current]

workflow attached on video

Any experts in LTX2 IC Pose LoRa for animating images with a base video?

Looking for more fidelity

[atomic mortar]

Depends completely on your needs

Anime? Illustrious type, any finetune that's popular is fine

(on civitAI)

Realism? Depends on your hardware, z turbo & refining?, flux dev? Flux 2? Flux Klein? Sd 3 large?

[normal aspen]

Hey, I don't know who's a mod in here but I got removed from the group my account got session token hijacked & it kind of is unpreventable I am trying my hardest to get it back it was a user of the name alvinp9 that was my account I do apologize for the crap that was posted in the channels I could not control it but if I do manage to get the account back since it was reported and permanently suspended I was wondering if It could be taken off the banned list since it will be back in my control unless rules say otherwise...

[lone rampart]

hi

[atomic mortar]

Dev scam as useual

[violet scroll]

Hello, I’m Mary from Uk 🇬🇧
I'm looking for a serious and reliable partner in starting a new business idea.
I already have a solid niche in place, and now I just need someone ready to grow and build with.

If you’re genuinely interested, let’s connect.

[upper vale]

whats your idea?

[honest kraken]

@vagrant pivot

[atomic mortar]

Scams as usual

If you need advice feel free to drop your question but i don't think you need to pay anyone specifically since training is very easy

[lofty hare]

“I ran into an image comparison issue in ComfyUI and ended up building a small node to fix it. Sharing in case it helps — feedback welcome.” https://fenixanimator.gumroad.com/l/ImageSliderCompare

[narrow gull]

Does anyone have a simple image-to-video workflow for WAN 2.2?

[atomic mortar]

Comfy ui templates has one iirc

Like built in

Or in swarmUI its easier

[narrow gull]

I'm using Runpod with an RTX A5000, but the videos from ComfyUI with WAN 2.2 take a very long time to load, more than 30 minutes.

[atomic mortar]

Are you using one of the speedup loras?

[narrow gull]

I don't think so, I'm just using the workflow that came with ComfyUI.

https://console.runpod.io/hub/template/one-click-comfyui-wan2-1-wan-2-2-cuda-12-8?id=758dsjwiqz

I'm using this

[harsh current]

Me too.

[narrow gull]

Is yours working properly? Which pod are you using? With the RTX A5000 it takes a long time, I don't know if it's a bug.

[atomic mortar]

Without the speed loras its expected to take a long time

Then again with a 5080 it takes like 5min with the lora

30min without

[narrow gull]

Do you have a link to download this lora?

[flat marlin]

Halloo

[heady turret]

Hey... im new to this ahaha

any guides I can look at to what make of these prompts things? all my stuff looks like cutout picasso

[fervent thunder]

Just pay for adobe fire fly

10x better than stable diffusin

its for pooros with no money

its crap

[still glacier]

skill issues

[still glacier]

what do you mean exactly ? can you show any examples in #🏞｜general-with-images or #📝｜prompting-help ?

[heady turret]

well just in general.. i just installed it through matrix and tried basic portraits.. keep getting the stuff that looks like john and jane does from true crime investigations.. creepy stuff

[still glacier]

honestly, it doens t help much. It could be many things, outdated models, bad prompt, wrong resolution, etc... Would help to see some outputs and screenshots of your webui showing the settings used.

[heady turret]

well I didnt touch mouch just the batch count and a prompt of a helmetless knight. I have no image to show though, i deleted them.. ughh creepy

oh there is ahaha seems like its saved somewhere..

[weak tree]

Hey everyone, it's a slow season for my main business, so I'm a little bored at the moment.

Has anyone done stablediffusion on Runpods and spam influencer content creation (Infinity Talk or Scale)? If so, did you see any results on the platforms?
I love using StableDiffusion, and I have about $5,000 a month to try something out, maybe a influencer podcast ai or food content etc..

[atomic mortar]

If you got that much money to spare its better to just buy the hardware and do it without restrictions

[weak tree]

Yup, I taught about it, but need space, let's say it's a "first try to make it"

But never saw anyone with good result (money making) so I m wondering if some pepole do

[atomic mortar]

Most people dont ngl

[weak tree]

Most pepole don't in any business, just wondering if some do. Big difference between "Some do" and "No one do"

[atomic mortar]

Theres more money to be made in making custom loras (thru Patreon/commission) or helping other people setting up

Unless you go really big and manage to hook a larger business but atp they most likely do it them selves

[weak tree]

Ngl; I was thinking about doing a food influencer; trying all kind thing

[atomic mortar]

i feel like real food would get more views

people are pretty anti ai still on that front

[weak tree]

100% the idea is to fake real food; famous and real

Seems possible with good technique;

[radiant turtle]

Dead server

[atomic mortar]

hey now, if your lucky you might meet a spam bot

[worldly warren]

Hi all ! I was wondering wether tools like comfyui and automatic1111 are optimal for my use case.... Right now I'd like to create environment concept art and I have very, very specific pictures in mind, but I can't get the tools to render quite exactly what I envision. To the point where it's easier for me to just draw them myself from scratch.
is anyone familiar with environment concept art using these tools ? Any tips or workflows to share ?

[atomic mortar]

doing it by hand will always match your vision but you could use controlnet sketch or depthmap to get better enviorment concepts

[junior urchin]

ok bot

[elder crow]

Hi, I’m a Full-Stack & Blockchain Developer open to new opportunities.

Frontend : React,  Next.js, Angular, Vue, TypeScript, JavaScript, Tailwind CSS
Backend : Node.js, Express, Python (FastAPI, Flask), REST APIs, GraphQL, WebSockets
Blockchain/Web3 : Solidity, Rust, Go, Move, Smart Contracts, DApps, Web3.js, Ethers.js
DB/DevOps : PostgreSQL, MongoDB, Docker, AWS, CI/CD, Git, GitHub

I have experience building end-to-end products - from UI to backend to smart contracts.
Please feel free to hit me up anytime if you’re looking for a developer or want to collaborate.

[atomic mortar]

Bot again

[opaque bramble]

I have a question: if I create a song from another song with an input strength of around 50, will the resulting song be considered a new song and be used for commercial purposes?

[elder crow]

what ?

[harsh current]

OK

[atomic mortar]

Dev scams, you know full well

[elder crow]

devs work for client and project honestly

[elder crow]

why do you think so ?

[atomic mortar]

1. We get many a day
2. Read the rules nr5

[elder crow]

oh, i understand

[atomic mortar]

Oh and let's mention your profile using a unique way to bypass the blo€kchain filter

Screams of scams

[hot coral]

hai anyone use comfi ui here i got a small doubt

[split tinsel]

hey guys im looking to run local llms for consumer grade hardware. mostly will be doing i2v work . Im pretty new to the field so there are huge gaps in my knowledge here , so please dont mind stupid questions some of whic are :
Assuming i have 64 gbs ram / 5090 desktop .
*how important is the speed of cpu here when offloading happens ?
*Could anyone suggest me a decent cpu for me?
I have made a build myself , f anyone has any feedback, itll be greatly appreciated. Thanks 🙂

Processor: AMD Ryzen 7 9800X3D Processor with Radeon Graphics (8 Cores, 16 Threads, Up to 5.2GHz, 104MB Cache, AM5 Socket)

Motherboard: GIGABYTE X870M AORUS Elite WIFI7 AMD AM5 Motherboard

Graphic Card: Zotac Gaming GeForce RTX 5090 Solid OC 32GB GDDR7

Power Supply: Super Flower Leadex Platinum SE 1000W SMPS - 1000 Watt 80 Plus Platinum Certification Fully Modular PSU with Active PFC

Cabinet: Corsair FRAME 4000D RS ARGB Mid Tower Black Cabinet CC-9011296-WW

Memory (RAM): G.Skill F5-6000J3636F32GX2-RS5K Desktop Ram Ripjaws S5 Series 64GB (32GBx2) DDR5 6000MHz (Matte Black)

Solid State Disk (SSD): Western Digital WD Black SN7100 2TB NVMe Gen4 SSD

Monitor: Samsung Odyssey OLED G8 34inch WQHD Gaming Monitor with Neo Quantum (LS34BG850SWXXL)

CPU Cooler: Corsair Nautilus 360 RS ARGB Liquid CPU Cooler - White

Case Fans: Arctic P12 Pwm Pst 120mm Cabinet Fan 5 Pack Black (ACFAN00137A)

[junior urchin]

No, bot, read rule 5

Ask LinkedIn

Rule 5 does

#✍🏼｜rules-and-tos

[split tinsel]

Anyone help pls

[simple thunder]

Hello

[opaque bramble]

if I create a song from another song with an input strength of around 50, will the resulting song be considered a new song and be used for commercial purposes?

[low moon]

https://youtu.be/f2cVEUrcHQ4

[still glacier]

I would not use any kind of AI generated stuff without consulting lawyers. And I would certainly not trust random discord people's legal advices. It all depends of so much stuff... your country's laws, the models used, the tool used, the usage of generated content, etc.

[hasty nebula]

I also see this happend. When people's legal advices to good to be true, I directly check on another stuff or tools they used. And I do the models to generated conten also on my channel.

[modern pagoda]

Good advice

Easiest is modify original song. Like a dj mix sampling elements. Changing pitch, tempo, key, rearranging structure. Then use AI.

Completely new song before even processing it. Legally speaking.

I got 128 ddr4 & i love it. Only 3080 tho 😭🙃

Ddr5 good luck. Byebye kidneys

M2 too are insanely expensive rn

I love having it. Not much for this stuff. So far.

Im happy with it. So far. We shall see for GTA6 🥲

64 & 128 ddr4 im having trouble finding rn for good deal for a build.

Might bite for used 😭🙃

Smart

I ... shoulda

[lusty tartan]

Hello, is there anyone who is looking for developer, please?
I am computer vision and AI engineer.
I have experience in computer vision and generative AI projects, including :
Building real-time computer vision models for object detection, image classification and Image segmentation,
-Developing LLM chatbots,
-Building RAG-based chatbots,
-Building LLM-based Agents,
-Training and fine-tuning LLMs,
-Developing OCR application for converting images or scanned PDFs to text,
-Building web-scraping pipelines.
In general I can build an end-to-end AI-driven projects.
If you are going to build computer vision and AI project, please contact me.

[still glacier]

no, rule 5 #✍🏼｜rules-and-tos

[junior urchin]

Is it gonna be a back to back bot

Poor nickel seems to hit the automod on each encounter

seems like he gave up

[glad cobalt]

Hello Ive been playing with AI a lot over the last 3 years and got my mind wondering into a full miniseries. I got serious about making it at the first of the year which was disastrous. I was looking into some storyboard subscriptions for help and then MoltBot came out. The new pc for it comes Saturday and I want it to make an open source storyboard you can run locally. Anyone have any ideas or want to help?

[thorn hare]

how is the general support for AMD gpus today for AI software ? like Stable Diffusion forks, things like Kobold ccp or ComfiUI ?

[atomic mortar]

Not amazing thats for sure

But its getting better

[still glacier]

anything SD works just fine, same for LLM.

Some gen AI models however still do not play nice. Z-image is getting there, it works most of the time but for ultra high res not so much.

And I don't bother trying videos, qwan, etc

[warm junco]

Pretty good support.
Zluda for older GPUs like RDNA 1 & 2
And ROCm support for RDNA 3 & 4
Native comfyui support on windows for them (3&4)

[vague birch]

hi

[vague spire]

Hi guys

[junior urchin]

why do you ask this every 4 days

[atomic mortar]

bc its a spambot

[junior urchin]

shhhhhh

the clanker will be hurt

[atomic mortar]

scam as useual

[thorn hare]

would it be better to buy a new 5070 with 16GB ram, or RX 7900 XTX with 24GB , but with limited support when generating pics or animations since its an AMD card ?

[low moon]

https://youtu.be/eUBs-U-huR4?si=SzOV5XKr67jvx8OX

[desert dagger]

you want an intel card

[plucky kiln]

whats evertyone using to install their stuff? all using ComfyUI? and just what's shown on there like the workflows or are you all getting other stuff in other ways?

I used chatgpt lfmao

[raw iris]

/imagine#🏞｜general-with-images message
enhance photo, try to preserve original face as much as possible, photorealistic, natural lighting, slightly zoomed out

[plucky kiln]

we can finally LITERALLY just type ENCHANCE and it will lmfao

[junior urchin]

Clanker moment

[shrewd ridge]

[Tool Release] AINA Web Terminal (Browser-based Prompt Generator)

Hi everyone. I built a simple tool because I was tired of typing complex score-tags manually.
It's a "Reverse Tagger" that runs 100% locally in your browser.

⚡ New Update:
Now supports Illustrious XL as well as Pony V6.
You can switch modes in one click (Auto-removes score_ tags for Illustrious).

[Links]
🚀 Try it here (Web App):
https://effulgent-buttercream-96ee2c.netlify.app/
(Use this link for the latest ready-to-use version)

💻 Source Code (GitHub):
https://github.com/AfricaGGtaro/AINA-Web-Terminal
(Open Source. Feel free to fork/customize it)

---

Dev: Africa
I can't set up a "Buy Me a Coffee" account due to my region, so...
If you like this tool, please follow me on X (Twitter) instead. That keeps me fueled. ☕
🐦 My X: [https://x.com/g8k9jrh5t580304?s=21]

[still glacier]

1/ not the place to post that, maybe try #1092446741984444416
2/ write a proper description... Cause this is clearly AI gibberish
3/ github link is broken
4/ no way in hell I'm clicking a netlify.app link, nor I would recommend anyone to do it.

[shrewd ridge]

⚠️ Update: Why the link broke & How to use**

[Status]
My GitHub account (@AfricaGGtaro) was auto-suspended by the spam filter.
Reason: I pushed updates too frequently today, so the AI thought I was a bot. 🤖💦
I have submitted an appeal, but for now, please use the Web Tool below (it's hosted on Netlify, so it's safe).

🔗 Web App: https://effulgent-buttercream-96ee2c.netlify.app/

[How to Use "AINA System"]
Since you can't see the Readme right now, here is the guide:

1. Select Mode: Choose STABLE (Pony V6) or ILLUSTRIOUS (XL) on the Web App.
2. Copy Code: Tap the button to get the "Injection Code".
3. Paste to AI: Paste the code into your ChatGPT / Claude.
4. Chat & Trigger: Discuss your idea with the AI. When ready, type the command A-Gen S.
   -> The system will output the fully optimized prompt.

Thanks for your patience!

[atomic mortar]

bro did not read the mod post

crazy

sounds like a HORRIBLE idea as its used for spam and phising

[pastel girder]

/generate prompt: anime logo, sharp scythe, icy blue tones, futuristic, high contrast

[atomic mortar]

How did you get the idea that this works here, just wondering

[grave stag]

Anyone got a good nafw art server (not an AI one) I wanna learn to actually draw either to clean up AI or for real.

[junior galleon]

whats the best sd out of them all?

[plucky kiln]

If im making a prompt to video, how long can aprompt be? If I say, wanted to make a few miute long thing (could use the ai to music/sound?) for say, a written part of a data sheet, and have the ai effectively just repeating it, much like an audio book, is there a limit to how long a text prompt can be?

[next wraith]

Hey, we're dropping free athleisure fits for people in tech who wanna keep fit. Follow us on insta to get notified. https://www.instagram.com/dobble_uuu/

[normal jasper]

any links to best practices for training model likenesses? So, if we're making Avatars, what kinds of images should we generate to do the best, most flexible and future-proof training?

[unique lichen]

Guys its makjng me go bald

Pls help

🥲

[winged thistle]

what is

[unique lichen]

I'm not tech savy, I'm trying to generate txt 2 img via web UI
But i csnt find good tutorial that properly tells how to download models n lora

[winged thistle]

i'm not tech savy either sorry

[unique lichen]

[still glacier]

Most people get them from civitia, it s clearly labeled which ones are models and loras.

After that, where to put them depends of your client/web ui but usually in the models folder you will find subfolders clearly labeled loras for lora files and stablediffusion for stable diffusion (and derivatives) models.

[low moon]

https://youtu.be/6lq1dPsUX8M?si=oVZEDyMgBhrmsImz

[rustic shuttle]

I'm seasoned AI automation, chatbot, agent | CRM expert.
I am here to collaborate, exchange and share everything with you.
Wish your great help.

[polar jungle]

Anyone able to guide me on how to properly train a model?

Or lead me to a guide

[split cape]

Hi guys

I think I was in this server a while ago but I am back

I wanted some advice on how I can get SDXL on Fooocus to give me a better result. I'm trying to match a certain vibe from my image prompt, and one result was close but I can't get it to look like that again even with the same seed

[warped robin]

Hello, I wanted to ask you guys something.

My RX 6600 is starting to struggle, so I want to upgrade. I mainly game, but AI stuff (Stable Diffusion, text models, etc.) is actually really fun and I want to keep experimenting with it.

I’m currently deciding between a B580 (~260€) and a RX 9060 XT 16GB (~430€). RTX 5060 Ti 16GB cards are like 600€ here, so that’s sadly out of budget.

How is ROCm on Windows these days — is it actually usable, or do I still need to use something like ZLUDA or DirectML?
Also, how’s Intel’s AI support on Windows in real life, not just on paper?

Thanks for your help!

[shadow ingot]

Hi guys, I know nearly nothing about programming; but I am trying to get checkpoint randomization to work in Forge. In A1111, I used the randomize extension.
But it doesnt work in forge due to differences in checkpoint loading. The other randomizations from randomize extension still work.

So far I am reaching part of the goals by making a script and debugging it with GPT; but I fear I may reach a hard wall soon. Should I ask in #🤝｜tech-support or since it isn't inherent to installation or regular troubleshooting, should I ask anywhere else?

To anyone interested, so far the script successfully swaps the checkpoints; but it seems it is failing to reuse image gen parameters and generation crashes early.
This is primarily intended to run 1.5 architecture models; but I didn't try if newer forge versions allow me to run XL with LoRAs.

[warm junco]

ROCm works with RDNA3 and RDNA4 cards natively now

So Zluda is not needed for the RX 9060xt

Intel's ai support is not that great

[warped robin]

Thanks for your answer

Just bought the Rx 9060 xt

[gentle prairie]

does the SSD matters at all for generation? i don't tend to switch models, but i do a lot of generations

[olive summit]

I'm wondering the same thing. I've always though that the only effect SSD has is how quickly the model is loaded into memory.

[faint timber]

Not that much. I have my OS on a SSD and Comfy running on a refurbished data center mechanical 7200 rpm drive.

[still glacier]

Assuming you re loading the model all at once and not partially (eg : using --med-vram or similar options) then no it does not impact generation time.

[gentle prairie]

speaking of, i have currently 10 gb vram but i'll have a 16 gb vram soon, i won't have to use --med-vram right?

[full lark]

Not for generation, but it does matter in how fast you can use it for the first time. For example if i boot sd forge neo webui it takes around 24-25 seconds on an ssd, and with my hdd i remeber it took around 1 or 2 minutes or more. And when I press the generate buttom the first time on the ssd it takes around 23 seconds before it even starts processing

[gentle prairie]

ok then still being on a SATA SSD is fine, won't matter much?

[faint timber]

Yep, my experience too. You have to wait for pre-flight.

[gentle prairie]

using my nvme for gaming and so on

[full lark]

I dont have an nvme so I cant test but i think it should be around the same

[gentle prairie]

yeah i'm not gonna change my drives then, that's all i needed to know

thanks

[full lark]

Np!, I'll try to test it on a ramdisk to see if the difference is noticeable

[upbeat token]

Hi, I'm keal from Norway
At the moment, I'm quieter phase of life
Focused on steady personal development rather than chasing attention or quick results
I believe growth comes from consistency discipline and showing up daily
Even when progress Isn't visible or acknowledged
Recently I've been involved in a structured retail environment that's strengthen my patience, sense of availability, long term thinking
It's been a gronding experience and has helped me stay aligned with the direction I'm working toward

I value meaningful conversations and connecting with people who think about growth, the future and purpose, if any of this resonates with you
Feel free to reach out I'm always open to thoughtful exchanges

[junior urchin]

feels when discord has already deemed you a spammer

[hybrid forge]

bit off topic but maybe anyone know wherei can i hire someone to generate me realistic pictures for my models

[atomic mortar]

So you make models but cant make pictures with them?

Fiverr should be a good place

[quaint cape]

HIRING AI GUY🔥

We run a modelling agency, currently a team of 30+ people and we're growing a lot!

we're looking for someone that can join us as a AI designer! This is a long-term position, for someone who wants to grow and wants to be part of one of the biggest agencies in the space!

Apply if you:
are hard working & motivated
understand pop-culture & gen-Z style/fashion
understand social media trends & virality patterns (hook, retention, cta)

AI stack: Nano Banana Pro, Seedream 4.5, Comfyui, Higgsfield, Wavespeed, Kling motion control, LoRAs..

to apply, DM me "HIGGSFIELD"

[novel sphinx]

Greeting, everyone!

I’m an AI + Full Stack Engineer focused on LLM systems, autonomous agents, workflow automation, and multimodal AI (text · voice · vision).
I build production-grade AI systems, not demos, connecting LLMs, APIs, databases, real business logic, designed for reliability, scale, and real usage.

Core Skills

│ • LLM orchestration (DSPy, LangChain, AutoGen, CrewAI, ReAct)
│ • RAG pipelines (vector DBs, hybrid search, custom retrievers)
│ • Multi-agent systems (planning, tool use, reflection loops)
│ • Multimodal AI (Whisper, CLIP, YOLOv8, TTS)
│ • Full-stack & backend (Next.js, React, FastAPI, NestJS)
│ • Automation & integrations (n8n, Zapier, Make, custom APIs)
│ • AI agents for research, ops, and customer support
│ • Knowledge assistants & internal tools (RAG-based)
│ • Workflow automation & AI copilots
│ • Multimodal chat / voice assistants
│ • MVP → production AI system delivery

If you need help turning messy workflows into reliable AI systems, happy to connect 🤝

[timid vine]

made 15k in 2 weeks from an AI Model

games gone clear

easy

life

[junior urchin]

[atomic mortar]

scam

[junior urchin]

@vapid dove Any possibility we can update automod pls?

Ahahahaha

as I said that, discord flagged the scammer as a spammer

Maaaaaaaaaax

[vapid dove]

oi

[atomic mortar]

i swear if we have automod adding multi-agent, full stack, ai stack and the 🤝 emoji to a filter youd remove 90%

[vapid dove]

can do

[nova lily]

Im not interested in generating videos, but I would be intersted in adding subtle motion to still images. Whats a good way to go about that? App/Website/GUI? thank you

[valid anvil]

can anyone help me dealing with captcha?

[nova lily]

any video example out there?

is it manual or automated? im just looking for automated something simple

tyty will check out

[junior urchin]

This is indeed the best free software available

I went from Sony Vegas to that

Way easier to learn .. or because they're similar

[atomic mortar]

scam

[junior urchin]

@still glacier

[spice ravine]

hello i need help i have this tab pooping out when i connect my account to stable deffusion and it sey "the string binding is invalid"

[full island]

Hi everyone! 👋
I’m looking for a specialist who can help me set up a local Stable Diffusion workflow.

What I need:
– Stable Diffusion WebUI (AUTOMATIC1111)
– InstantID (or IP-Adapter for face identity)
– Proper pipeline for realistic portraits

Everything is already installed and running locally, I just need correct setup and guidance.
This is a paid task.

Where can I find such a person?
Thanks in advance 🙏

[junior urchin]

You can probably ask on civitai, for free at that

[atomic mortar]

You really dont need a paid task for this, just a better UI since fluxklein or even zturbo can do these things

Outdated request really

[full island]

Ooo

[full island]

Thank you! I'm going to see what it is now.

[fiery ore]

You can DM

[lofty wraith]

Hey, @full island DM me. I can help you.

[atomic mortar]

@full island dont dm these bots lol

[junior urchin]

Poacher bots

[atomic mortar]

just look up his recent messages, its only asking to dm

[junior urchin]

[heavy crypt]

i made a symlink script...well i argured with the AI until i got it work

it might be useful for others..where can i post it?

[swift aspen]

why is there no ai artist role -_-

[unique ravine]

Cuz we aint artists

[knotty willow]

Hi,
I am Swap,
Developer and founder of Drooid. It's an AI news app that gives unbiased news. Looking forward to connecting with you all.

https://apps.apple.com/us/app/drooid-news-from-all-sides/id6593684010

[shrewd lion]

Hi everyone, I'm struggling with a manual install of A1111. Even though I downloaded the ZIP file and used GIT_PYTHON_REFRESH=quiet, the CMD keeps asking for GitHub credentials when trying to clone repositories.
When I cancel the login, it throws: RuntimeError: Couldn't clone Stable Diffusion. Error code: 128. I've tried unsetting credential helpers but it doesn't work. Any advice? Thanks!

[atomic mortar]

Seems like your doing it completely wrong, follow the guide from cs1o in the #🤝｜tech-support pinned messages

It shouldn't ask for a account at all

[still glacier]

GIT_PYTHON_REFRESH=quiet what ???

most likely you're trying to follow a really old tutorial for automatic1111's stable-diffusion-webui.

Which has been broken for a few months and abandonned for nearly two years by now.

So yes you better off installing forge, forge neo, forge classic, comfyui, etc by following the pinned guide in #🤝｜tech-support

[shrewd lion]

Thanks everyone for sharing. I actually tried a few methods I found online and they didn't work, so I downloaded Stable Matrix and the forge from there, and it worked!

[shrewd lion]

It's true that webui.bat is a bit different from comfyui, so I'm not very familiar with it yet. I only started trying it out last night.

[nova lily]

so of course the smaller the face, for example in a full body image where its farther, its gonna have less detail and no manner of adetailing or upscaling will only barely help. is there anyting you can do to get stronger detail on these type tys of images?

[woven nacelle]

https://youtube.com/shorts/YjqLfn_GDgU?si=dNETGyj7YFQMZc-C

[tranquil jasper]

Yo everyone,
I’m Timo, based in Johannesburg.
I’m someone who’s driven by passion and hope. I like growing, learning new things, and exploring life to understand my purpose better.
I’ve been focusing on improving myself, thinking differently, and not limiting myself to the usual way of living.
For me, it’s not just about success, but about becoming a better person and feeling fulfilled with who I’m becoming.
I respect people who have their own mindset, who want to grow, and who aren’t afraid to follow their own path.
If this seems to be you we’ll definitely vibe

[limpid gorge]

Hello Community 👋
I’m Likely （らいくりー）, a solo developer creating an AI art–friendly “gallery + BGM station” concept for creators.
Of course, AI Art creator (mainly SD / Suno)

[keen lava]

https://docs.google.com/document/d/1s1Yl5u71W6GcZOHbuyoF-bozKJkEu88An_jWoFjHJbI/edit?usp=sharing

[plush knot]

Sup fellow Ai Nerds. Downloading images from the ComfyUI cloud output folder SUCKS. Automatically sending files to your Google Drive doesn't suck. Thank me later nerds. Oh yeah, node has encryption too and makes it even more nerdy.
https://github.com/machinepainting/ComfyUI_DriveSendNode

[uneven pulsar]

yayy

im so glad i found stable diffusion

basically what im trying to do is use ai to do all my mental and physical labor

and then get paid for it

[ebon sequoia]

Hi! I specialize in custom AI face LoRA training (Flux/SDXL) with consistent identity across generations. Happy to share workflows, tips, and examples 😊
https://www.behance.net/gallery/243708697/Stable-AI-Influencer-Private-Flux-Face-LoRA

[atomic mortar]

posting in every channel? spam/scam

[ebon sequoia]

this info is not for u

[feral basin]

Hey guys i want to make my ai influencer's sdxl lora or z image how can i make it easily? idk anything

[worthy cairn]

How to approaching going about your AI influencer?

[modern pagoda]

Dont do it alex. You will be 💩ed

Lmfao

[feral basin]

wdym

[worthy cairn]

What checkpoint do you use? What Lora's do you use? How do you prompt your images?

Just downloaded it today

Need some tips and help

[feral basin]

i am not using lol i ask for How can i make

i used Gemini but is censored now

[worthy cairn]

Oo that

I don't know how to make Lora's. But well, there are a lot of realistic checkpoints you can download. Pair them with some Lora's that work well and boom, you've got yourself an AI influencer. You could use img to img to create more variations of your influencer

Don't know more that that. Only started today

[quartz shard]

What's the best download for me rn?
I'm dumb, I want to install loras as easy as possible
RTX 5070 Ti Gpu

Some guy told me to install forge

Today I tried automatic1111 but I encountered a lot of issues with the first bat file

python versions etc

[still glacier]

#🤝｜tech-support guide in pinned message

[quartz shard]

Thank

I think it's an old version though

[quartz shard]

What if i face errors while following the guides 1;1

Like somewhere it says click run.bat and it installs stuff

But that fails for me

The only thing I could get working was INVOKE

[still glacier]

ask in #🤝｜tech-support and provide screenshots / logs

[uneven pulsar]

lets do it!

[atomic mortar]

Scams

[atomic mortar]

If you wanna get scammed engage with that bot, its here once a week

[uneven pulsar]

its cool

[atomic mortar]

Its genuinely a scam

[uneven pulsar]

yeah im fine with it

thanks tho

[atomic mortar]

And they would use your identity to scam more people

[uneven pulsar]

yea but i could use the money rn

[atomic mortar]

And you'd not get paid and would use your device (and probably fill it with s virus)

But you do you

Oh and before i forget theres gullible written on your ceiling

[uneven pulsar]

no there isnt?

guess who just made $1300 on bitcoin !!

[modern pagoda]

Tether isnt btc. Its a shitcoin no one wants lmfao

Thats the tandem scam lol

[sour jolt]

/ping

!ping

[rustic shuttle]

Hey!
Is there anyone looking for AI engineer?

[kind compass]

Is anyone here interested in building practical AI driven products?
In my recent work, I’ve focused on applying AI to real, everyday problems, such as:
• Developing voice agents that can handle customer inquiries, schedule appointments, qualify leads and provide 24/7 conversational support helping businesses improve response times while reducing operational costs.
• Automating routine workflows like document processing, customer communication and internal reporting, saving teams significant manual effort each week.
• Building intelligent web applications that summarize complex information, highlight key insights and support faster, data driven decision making.
• Personalizing user experiences through smarter search, recommendations and adaptive interfaces that surface what users actually need.
If you’re looking to create practical AI products or want a technical partner to help bring an idea to production, I’d be glad to connect and collaborate.
Feel free to reach out. I’m happy to share real examples, demos and technical details if helpful.

[atomic mortar]

If only there were no bots and people bothered to read rule 5

Would be insane actually

[junior urchin]

We can just collect the bingo card words and give them to max

"is there anything looking", "anyone interested in" and so on

[lone rampart]

ahem

[snow wing]

how can i use photo ai

[ivory thicket]

ai.com

[weary dust]

Hello, do you know how I can contact a staff member? I'm having problems with Stable Diffusion.

[still glacier]

best is probably to contact them by mail https://stability.ai/contact

[broken gull]

Idk if you're a mod but in #🎵｜stable-audio there is a mr beast scam.

And other channels too.

[still glacier]

thanks, cleaned

[broken gull]

Do I recognize you?

Been a long while since I chatted here.

A lot more channels back then..

[still glacier]

Maybe, I ve been here for a long time.

[broken gull]

Been here since first discord invite beta round?

Me

[still glacier]

I think I remember a kingpin

[broken gull]

Prolly me

[still glacier]

been here since october 2022 ish

[broken gull]

I am different now lol

[still glacier]

arent we all

[broken gull]

Me august 6th 2022

So I have been here longer then you xd, less active tho.

[broken gull]

You Nov 16th 2022

[still glacier]

seems so indeed

I was on another less populated server before it.

[broken gull]

Do you know how I could potentially generate continues streams of music? I have been interested in EEG and hand gestured steered music for some time now.

[still glacier]

truly continous stream ? uninterrupted ?

[broken gull]

Ye like a radio.

Kinda

[still glacier]

I don't see how that could work on a purely technical point.

[broken gull]

It's not meant to stay on infinitly tho, only when the device is turned on.

[still glacier]

You d have to hide cuts at some points.

[broken gull]

No way to blend in several audio files or something?

[still glacier]

otherwise the """"context"""" would just explode

[still glacier]

that s why i said "on a purely technical point".

Keeping the radio constraints in mind you could imagine using the radio announcement as a break between generations.

music 1 -> music 2 -> prerecorded / recycled / parallel generation announcement "welcome to stable radio FM yap yap yap" -> music 3 -> music 4

[broken gull]

It would be like, audio gets generated but based on EEG signals to change mood tempo vibe etc.

A radio isn't the goal just mentioned radio cuz it provides a continues stream of audio (voice and music).

[still glacier]

You would still have to find some break / sync point or cross fade.

[merry acorn]

Wait there is something i want to ask but it seem general chat probited upload image. So where i should ask?

[still glacier]

there s #🏞｜general-with-images for general chat .... with images

just make sure it respects the rules before posting. Rules are the usual no nsfw, no gore, no etc

[modern pagoda]

You could use RNGs to select a pool of Generative AI created tracks, created in advanced. & the RNGs are used to "mix" the tracks by volume/gain level, BPM, Key Detection, Pan, Duration/Time & FX per channel. Might get you closer to what you want.

[broken gull]

Oh interesting, could you send a link?

[modern pagoda]

MantraBox type of thing

Link? Thats a description of what you could build.

[broken gull]

Ah, I thought it was a tool that you where describing.

[modern pagoda]

Well it cant be too hard to make that lol

[serene forum]

Hi @all
I am an architect and I’m looking to develop a workflow for my Arch and Interior viz on SD.
I am decent with visual scripting.
Looking to connect with someone to understand the niche specific details of comfyui.
Thanks in advance.

[broken gull]

Ehhhh

Never worked with audio like that before.

[modern pagoda]

If you want continuous playback. Fill the pools dir with premade tracks. Just keep adding to pool over time so you dont deplete the tracks at random. If you only want to use each 1 time, you can also reuse them. Like playlist on shuffle.

Drone Boxes are a physical device with some similarities. I.e. it plays forever emitting pad like drones. (Not tracks tho)

[upper vale]

hey, does anyone know how to compare an index to a list of numbers and do nothing when the index is in that list (that is defined by me). so basically if the index is something i dont want, skip it. i got a node to increment indicies automatically but dont know how to make a number list and to compare each value to this current index and if it equals to just do nothing instead of going my normale image gen way. would be thankful if someone could help me in comfy ui

https://cdn.discordapp.com/attachments/1080939894626586685/1470817756776628295/image.png?ex=698cad47&is=698b5bc7&hm=32325384fe36d2c00a290836cec05301f96ed2ee06e00b8f682ce48e03b14008&

[broken gull]

https://wotja.com/app/

This looks interesting.

[broken gull]

It practically does what you are saying, I can make a VST plugin where people select sounds to go through wotja then go to the glove which can change pitch Velocity etc and then back to the VST (sorry that's a different project, I am juggeling.

I could do something similar for the EEG headset.

[modern pagoda]

Use RNGs to control the VST/VSTi's MIDI CC data. 0-127 values. Should be tunable to ranges you find that sound best.

Think of it like Dungeons & Dragons, using Polyhedral Dice, you randomly choose from a range of choices. Then roll other dice to customize that randomized choice.

Except the dice in this case would be a d128

Plus a dn for what your pool of audio tracks contain, i.e. total number, you want that to be a variable.

So 50 songs, use a d50 if hardcoded

Variables much better, never have to hard code a value.

It can just say 56 songs in pool/ dir you add random amount more one day it will tell you the total in pool/ by design.

Again im just going by building it from scratch if you integrated/or can integrate a VST plugin container like that wotja or whatever its called.

Cool concept honestly. Hope it comes together for you.

[upper vale]

hey, does anyone know how to compare an index to a list of numbers and do nothing when the index is in that list (that is defined by me) in comfy ui. so basically if the index is something i dont want, skip it. i got a node to increment indicies automatically but dont know how to make a number list and to compare each value to this current index and if it equals to just do nothing instead of going my normale image gen way. would be thankful if someone could help me.

[broken gull]

The watjo does this for you and much better, and has a URI interface to pass data similar to a REST API afaik.

I can have the user select audio from thr VST route it through channels from watjo and have watjo stream the audio to the gloves or something.

[modern pagoda]

Not sure if I talked about this in this server yet but, with ID verify coming, i think only for 18+ content, is there a plan in place for spill over so this community doesnt get fragmented & split up? A lot of good, creative minds in here, i really dont want to lose you guys 🥲😂

A lot of the regulars in tech support, anime, dailies, id really consider bringing on for some ai art projects, like curated vetted users to generate their art & creations. Find a way that each users subpage can be a feed of content, retain workflows, prompts, & example outputs, allowing users to pay the artists to generate outputs using their workflows & prompting. Turn prompts into a madlibs type usecase. Pay per generation.

Figured it was worth a shot. A lot of really cool stuff in here. Opened my eyes to insane possibilities i can barely grasp 😂 🥂 stoked to have been a part of this & get to learn AI art from the ppl here's creations. Inspiring, often.

[sinful ruin]

Pretty soon at this rate it'll be against TOS for somebody under 18 to be in the same lobby with you in call of duty 😄

That's where we are headed

[modern pagoda]

They keep voice banning me lolllll

What if i have tourettes (i dont) lmao

[queen kindle]

Hello world!

[gritty lava]

am feckin angry

:(.

tripo forcing ppl to pay now days

:V

too bad i dont have a elite russian hacker friend to access tripo server back door and extract my models for mes

xd

[full island]

Guys, my computer is a bit too weak for generating in SD… 😅 Does anyone know an online AI tool that can create videos with my face in it? Something with top quality! I really want to try making viral videos for Instagram/TikTok.

[warm junco]

Your 8GB GPU should work fine for images

But videos are not an option

If you want just faceswap in videos, setup FaceFusion

[full island]

I would like to not just replace my face, but create high-quality viral videos about 10-15 seconds long, but with my face, can you tell me if this is possible?

[warm junco]

Its hard and localy you would need a better GPU for that

So your only thing would be renting a cloud gpu

[full island]

I'm also considering this option. With my video card, generating one image takes about 4 minutes. 😑😑😑 Or maybe you know of paid online services where I could make videos or even high-quality SFW photos.

[warm junco]

Your GPU should generate an image in a few seconds

If its sdxl/illustrious based

Like Max 30 seconds

Come to #🤝｜tech-support then we can take a look

[full island]

yes, boss🫡🔥

[unreal wing]

hey

[crystal prawn]

Hello

[young comet]

Does anyone here have a project idea or an ongoing project ?
If you have any difficulties or need a developer, let's talk.

[junior urchin]

A project to prevent bots offering work, work positions, needing work?

🤡

[frigid sundial]

Just starting to dabble in image gen and I'm on a very outdated gpu and looking to upgrade

have seen mention that there can be driver issues with the intels but mostly in a gaming context, anyone run them here for stable diff, any issues?

[unique ravine]

get an nvidia gpu

[modern pagoda]

Think you want AMD at min but mostly 40 or 50 series Nvidia GPUs, mainly 90 series.

Im sure ppl here get good results with 4070s, 80s, 90s, & 50 series. Not sure about lower end models. Tech Support channel should help.

[frigid sundial]

Ah fair, I don't think I'd go near an AMD but was comparing a b60 to a 5060ti

[modern pagoda]

5060 ti likely has decent results

[unique ravine]

5060 is a good start

as long as its 16gb var

[modern pagoda]

More VRAM the better

[frigid sundial]

yeah 100% would be a 16gb model

I've seen this more vram is better many times.. hence looking at the intel b60 which is a 24gb card

both are at similar price points in my country

[unique ravine]

ildl rather get a used 3090 then

[frigid sundial]

3090 would consume too much power and require the swap out of other components as well, increasing the cost

[unique ravine]

eh not really, its a 350w card compared to 200w of intel

with undervolt u could get it down to 300

its much more powerful than b60

u dont need 24 unless u r doing training, which can at the same time be done by the 16gb card

[frigid sundial]

I'd rather be comfortable in my power, the system is used for a variety of applications and cpu already pulls 100-200w

but it does sound like the 16gb 5060 ti would be best at this time for me

[unique ravine]

cpu is irrelevant for ai generation

what kind of cpu pulls 200w unless its some higher end intel

[frigid sundial]

it's an i9, significantly overclocked on a custom loop