#programming
1 messages · Page 175 of 1
what would happen if one mails konii a sample of their dna
the github will always be up you know
f2
nah array indices are based
which
starting at 0 is good
i dont work for alex
starting at 1 is good
starting at 2 is good
good because you would get fired
twitch api, works with streamers 
so many people in here fit that
1 i can understand, but if you start at 2 i will defenestrate you out of my office
including some of the lot here rn
starting at 3 is bad
starting at 4 is good
strong opinions on javascript, apis, minecraft servers
are you insane? get out
i have strong opinions on most things
sbin
too employed
no
again im on a separate thing to avoid being terrorized so ive avoided giving myself away
shaow is unemployed rn

to be fair i am running on like 3 hours of sleep in the past 24 hours so 
exactly
oh, you mean that cipher is too employed?
i will not allow godot slander. godot is the best engine for indie devs bar-none
sorry chum
IL2CPP
thats all
thats not a response...
uh oh we can’t have another chat war today

i like sdkman's splash
i know what Il2CPP is... but what is your argument...
godot is a fine engine, unity is decent. but both pale in comparison to my pure performance.
rai_byte 
stuck with JIT
i did not like godot's editor at all
the whole node tree like setup im not a fan of
and thats a problem?
means u cant squeeze performance where its sometimes needed and not all platforms support JIT
you can get more performance in godot than you can in unity. just because of how open the engine is
i can use Burst and jobs for some complicated and heavy work too
using a bloated engine in general means you cant squeeze performance
thats mostly with disk size
erf thinking too hard
there are so many graals
also godot dropped webm support like a bunch of monsters
not all platforms support JIT
it’s basically only ios because apple sucks
eww that's a steve jobs approved action
that alone made me ditch it
who uses webm lol
webm is great
just use mp4 like the rest of us
what's wrong with webm
which for a 2D animated game (this is for an old job i had) was terrible
other than ios refusing to support it because apple is dumb
webm slander 
id rather webm any day than ogv
why would use use webm or ogv...
ogv is forced on godot
what is this referring to, webkit? i have webms playing in apps
at least historically if i linked a webm in discord etc it does not play on ios
it's just a download
i have become uninterested in the identity situation, imma watch some anime
oh yeah that's true
gah
guh
i guess webm was too complicated for them to develop support for
i give up on deciphering you (pun intended)
somehow
the only c# person i have interacted with in here is scrub so i dont care who you are 
this seems like a discord issue because as i said i have webm playback in multiple other apps and safari seems to play them too
do we even know your real account? cuz if we dont then what is the point?
vedaulh
godot
i see... honestly i never had a workload to require video playing. but they literally say you can extend it to play h264 or whatever with gdextensions
not here
i considered this but i decided its probably too much information
oops wrong reply mb
meant to reply to this
i guess, but i dont think i could decipher it without having known the other account
i am unemployed enough, but i do have actual work to do
or well, self-obligated work
im not getting paid to make my vulkan angine
slavery
if you dont want to work there is an alternative 
in my opinion no, cuz a job gives you money which you need to survive. i get my money by freeloading so i dont actiually have to work
i do so anywyas
survival is self obligated 
survival is one of the base instinct in living creatures, because the ones that didn't have that died
well, i dont want to discuss this too far because automod will get involved

https://github.com/godotengine/godot/pull/108600 seems like someone is atleast working on bringing proper video support to godot
finally
tho looks like it just converts to OGV from mp4 internally using ffmpeg 
fair
so its the same shit but with performance impacts
its only in the editor. no runtime performance impact
it gets converted when you import the video
brother wtf do you have with tongues? why do all the gifs you send have them?
nah
i feel like its just that playing videos is not really a common usecase so.. noone works on it
👅 
how i feel in this channel rn

bro has a 10gb reaction image folder
u have no idea
lemme get some snacks
i have enough organization to cherry pick
there is also this: https://github.com/EIRTeam/EIRTeam.FFmpeg
the horrors
20 years is a long time to be saving things without really organizing them
most i've done is script organizing things by filetype and date once in a while
i just drop em in the right folders once in a while
for me that's hwen i reinstall system usually but that has just left me with folders of backups x9000 it's too scary
i think it’s time to create stay on topic images like the ones in nn but for programming
wasn’t a direct reply to you
if you consider bash scripting to be close enough
i just had a brainblast... since mp4 or such isn't supported in godot. why not extract every image of the video and then have a texture node for every frame and then disable them 30/s to simulate a video player
no
...because video formats are more compact
reinvented the gif
well... but if it works. who cares 
anyone remember motion jpeg
the only bideo gaem video format for me
(i might unironically implement something like that for my project since it seems to be the most performant solution tbh)
bink video is still in development 
oh right... its used in unreal afaik
gotta organize discord too
how i organize friends
i will never know if that's a real site because i will never check
epic bought bink in 2021
that's crazy
for some reason all these arbitrary proprietary formats used in gamedev have always been interesting to me
probably just for the fact that they aren't really used outside of their niche and also arent very transparent if you're not a b2b customer about what the actual capabilities of them are
bink claims decoding frametime of 4 ms on cpu only for a 4k video
i thought u said bing ngl
i would never
but 4ms is like gpu decoding speed for most formats, that's neat
probably just doesnt actually do much in the way of compressing or anything that requires work to decode
so many good formats hide behind patents and other bs
honestly i just tried godots ogv thingy. and its decently good. good enough for my usecase tbh
Bink ?
h265 cpu software decoding is like 50+ ms per frame kek
av1 over 100
vp9 manages 10-20
for a game that involves a lot of video though for 1080p anime styled animations its terrible
https://www.radgametools.com/bnkmain.htm proprietary video container/codec combo that is used in many games, especially Unreal since epic bought the company that created it in 2021
RAD Game Tools' web page. RAD makes Bink
Video, the Telemetry Performance Visualization System, and Oodle Data Compression - all popular video game middleware.
that sounds like a very very specific usecase that if you want to use godot you would probably want to write a gdextension for it anyway
seems pretty informative https://www.youtube.com/watch?v=RE04LQffZfs
this is an 8 year old video though
there were a lot of situations that godot just wasnt able to cover
especially with that whole job
which used alot of advanced 2D animation
the right tool for the right job. godot isn't mature enough yet to cover every usecase. but for a lot of people it will be good or even better than unity at stuff
idk if its at the better phase yet
besides disk size
and licence
its been catching up for a while
disk size? license?
the engine is smaller and you dont need to buy a license
tbh those are the only appealing things to me if i were to move to it
uhhh maybe the start of 2024?
possibly a bit before
ive only played with it a few times in earlier versions
sounds like a VN time for ren'py
but that was the era where i was thinking about actually using it
idk you said you had to play a lot of 1080p anime videos
sounded kinda vm-adjacent
vm
vn
even
early 2024/late 2023 seems to be 4.2 godot. so not too far from the current 4.4/4.5 release
its not like a completely different engine since then but there were a lot of nice QOL and features added since 4.2
last time i was going to use godot for a simple 2nd thing i ended up using https://solar2d.com/
which used to be called corona but unfortunately had to change name a few years ago for reasons
lol
tbh other than Unity id probably just use my own framework or engine
"bro why does this fking extension have an update EVERY DAY" - me 2 seconds before i read the name of the js/ts extension i have installed rn
tbh i dont even touch vscode anymore
i havent been able to find anything i like more still that also lets you swap languages so freely
its the nightly version. it gets updated everytime the repo gets pushed probably
vscode is average with all but the best at nothing
yeah that was the point of writing "before i read the name of it"
just like me fr
jetbrains is working on their own alternative but id rather use specialized IDEs
i did like jetbrains pycharm w/ python
webstorm for js/ts
webstorm was good too
webstorm is just no subscription right i think pycharm the issue was that i didnt want to pay for essentially the same shit i was doing in vscode
Rider for C# also shows me the IL code that is generated
CLion for C/C++ can show me the size in bytes for everything
its the little things like that for me
vscode does the same 
you can surely get an extension for that
isn't that just a language server feature?
but ootb jetbrains has a nice config so i get it
tbh i dont like vscode for anything that needs a compiler
(and the built in spyware)
i use vscode for everything. if it needs text i use vscode.
i used to do that for some stuff
i cant use visual studio thicc edition anymore being on linux unless winboat somehow works wit it
i use kate for random text crap, kate seems to be very configurable to the point it could probably function like a vscode
i havent tried to test that tho
i used np++ on wandows ofc for the same stuff as kate
why
strong dislike of docker
to use this
or try it out anyways
Before running WinBoat, ensure your system meets the following requirements:
RAM: At least 4 GB of RAM
CPU: At least 2 CPU threads
Storage: At least 32 GB free space in /var
Virtualization: KVM enabled in BIOS/UEFI
How to enable virtualization
Docker: Required for containerization
Installation Guide
Docker Compose v2: Required for compatibility with docker-compose.yml files
Installation Guide
Docker User Group: Add your user to the docker group
Setup Instructions
FreeRDP: Required for remote desktop connection (Please make sure you have Version 3.x.x with sound support included)
Installation Guide
Kernel Modules: iptables and iptable_nat modules must be loaded
Module loading instructions
is it just a facade for wine
nope
good
vscode works for rust 
that's what i thought at first too
Looks like screensharing a VM
yeah but i meant a proper IDE fully for it
i can search cargo in it along with other useful shit
im sure you mean built in without any config but le terminal
and usually for any given random thing there's an extension that'll do it for vsc
hard to escape it because of that
hopefully with less exploits than wine
im just hoping it lets me run paint.NET
that's the only application i have missed from windows
literally the only one
u can probably make a better one
people have tried
I've considered making something close to a paint program, but haven't really gotten started on it yet.

only legends know
no
mozilla ate the navigator
it isnt really obscure
if i were to make a paint-like program, id make it for use on ym tablet with the pen thingy
it is for the younger people here
GIMP can seemingly mechanically do everything PDN did and likely more, it is quite powerful
but they forgor how to make a ui that isn't trash garbage
REAL
just use krita if you dont like gimp's ui
gimp just has a high barrier of entry imo.
i use krita for drawing and gimp for editing
I think both have very obtuse UI designs.
mostly what i'm looking to do is editing stuff which krita really isn't designed for although it can do
it's slightly less obtuse than gimp though
nope i'm fine with gimp
im pretty sure i have a photoshop subscription already or soon because i need it for college
On god though (signed someone who also took the PDN -> GIMP pipeline)
I won't say I have PS
written in free pascal which is cool
i have PS just not normally
someone posted photoshop 2019 on the AUR or something

yeah nothing seems to beat pdn in terms of simplicity of use while having the capabilities it does
a lot of it is being used to it of course
but i have genuinely been trying to get used to gimp for literally years off and on

im pretty sure they will teach us photoshop durrign the lesson hours so im probably fine
fork it
I'm in the boat to where GIMP is the best I have but I'm considering building out software for my own case.
i am so far from knowing anything about lazarus/free pascal
feel like chay randomly is about to be like oh yeah my favourite
no
pascal feels like the kind of language people who dont want to learn a second language would use
i saw some haskell packages earlier and it was some wild scientific random shit, thought of you
mhm haskell is good
I've been considering some vulkan based editor because I rely heavily on filters.
is it just particularly good for stuff like this, i was searching prism and found these, they tend to pop up and it's always some shit like this
-# Ah yes I love being able to import 3d models into photoshop
always random scientific calculations or physics related things
lens is a concept you won't even find in most other languages
its not physics
prisms is not physics as well
so it's just a haskell thing
fr though it feels like something written in the 60s in FORTRAN or something every time i see a haskell package
probably is

haskell wasnt even popular before like 2010
maybe 2005
its advancements in pl theory are only being implemented now in some languages
it has that grandpa coder energy tho
spj is a grandpa coder
the crazy eyes of a fellow who has seen the truth
slenderman
you should watch some of his talks on yt
im watching this rn https://www.youtube.com/watch?v=iSmkqocn0oQ
Simon Peyton Jones talking about the future of programming languages
this guy seems like someone that would be on computerphile explaining some shit i'll never use and have never heard of
looks like one of the characters from meet the robinsons
keep moving forward
listening to people like this guy make me realize just how limited in scope i actually am in terms of overall knowledge and ability man
thats why i like learning stuff
i wish i could freeze time and take my time ot learn everything about a subject
not as much as id like
that's true, all you can do is learn more and be cognizant of your own ability
hard to learn without acknowledging what your limitations are
the other day i was helping someone in this discord with a little utility and they ended up not needing to use it afterwards and they said sorry for wasting my time
and i was like ... that was not a waste of time? because i learned some things that i didn't know
yes
lmao

whats the difference
ripbozo
imagine needing to go on linkedin 
everyone there is so productive they automated their linkedin posting with agentic workflows
lmao


neuro should post to linkedin for vedal
doxxed by neuro shitposting on ur linkedin
i doubt she has access to anything like that
loud incorrect buzzer
abitrary cli execution go brrr
she hallucinates
saw the OG before if i recall
not vedal censoring a random address that still is in england for the lulz
i mean yeah
i mean we all know vedal is located in buckingham palace
surely tutel would never do that 

if vedal put any personal information into her dataset that would just be asking for it
it wouldnt be dataset, it would be something like what's my ip on google
om
mf got that sandbox + vpn to australia
and a helping of filters
neuros backend is not local
it's at least partially local
the frontend
surely part of it is, otherwise vedal wouldnt need his specs
depends on where the tool calls are actually being ran really, less the llm's inference
you kind of do depending on the case
could easily be on the backend as well but
makes sense to run them in the same place the llm inference is
no reason not to
its less moving parts
is neuro too coherent to be run entirely locally
im not sure about modern model capabilities
Hmmm
she is 100% not locally ran
Let me cheak neuro parameters
Hardware requirements
right, good luck with that

most of the other shit if i recall is local though
and vedal just uses cloudflare zero trust for the control panel
well if they interact with the greater framework i could see them being local as well but yeah
everything seems to be tunneled in such a way that it's probably fairly muddied depending on what is being done
based on the schizo diagrams he's made
neurospaghetti
for her speed she has to be running on something beefy
H100 OR B200 ?
she's almost predicting the future with her latency now (and might genuinely be doing some sort of speculative decoding ahead of time almost like a branch prediction but for the conversation lmao)
Approx (1 to 3 second)
If she run locale then it will max 100 ms
Py
it's probably under a second from time that there's no longer any audio input from the conversation and the system starts to speak again
Nai
her current tts is azure still no?
yes
evil's might be local
But highest quality voice like more take time
I rather say it is voice cloning
evil is an rvc model though right? does azure support that?
why would neuro not be local... what.
she isnt
the There is a problem with my AI is prerecorded then
neuro's latency is insane and she's probably a bigger model than most people realize
What
no
Give me voice samples
how do you know?
when he was having tls issues in unity it would still play there is a problem with my ai
because i know she isnt
Is voice problem or model rate
for a fact
for her ability at tool calling and juggling context and everything i still expect her to be at least 12-20b in size if not larger
parts of her are local yes
cant make azure requests with unity when unity tls is acting up
and i think she is... based on everything i know
factually wrong
care to explain why though?
its not useful to just say "no"

Let ask vedal for ?
it's really really hard for a smaller model to be as consistent and proficient at the things she does without screwing up a lot more or losing the plot
running a local ai model like neuro locally seems like the only smart and good way to go about it...
vedal himself has said before that she is across several machines
damn. you cracked it
yes.. doesn't mean its not local
ofcourse he has multiple pc's
im talking about the AI backend
which is 100% not local
sure bud
unfortunately being distributed doesnt necessarily mean WAN, need something more concrete than that

Hmm
I have girl sample maybe she will train ?
still didn't explain how you are coming to that conclusion. im not seeing anything on stream that can't be ran locally
apparently its "leaking" if explained so i guess we will never know 
we will never know the secrets
i would expect you to be able to run a ~32b size model at q4 on a 4090 comfortably
as far as tokens/s i dont know but time to first token should be under a second with that
didn't he literally build a pc with h100 in them on stream
that was 4090s

but he does have some no? maybe it was just a inside joke that im misremembering from like 2 years ago
q6 should also fit i think
some have speculated having like a dgx tucked under a counter 
probably
someone a while back was very passionately arguing that the turtle has one of these hidden in a closet somewhere running neuro https://www.nvidia.com/en-eu/data-center/dgx-h100/
Hmm
At least one of those needs to be a graphics card as neuro games on her own pc IIRC
I can rent probably (few dollars)
the osu! model definitely runs local
was just thingken of fast
Game model ?
Heard from @oblique verge
thats it. i decided i'll build my own neuro and call her oruen 
What
U def osu
me when expensive cluster is right there to rent and waste money on
AGI the nth coming
Nai compute is very costly
make a ai to work for you while you develop the ai to replace humanity 
Yep take a risk
ahahahah i was searching to see how big your average neural osu model might be and found this post
Give me link
oh this is lambda cloud
i should just say fug it and train the next echo checkpoint off gpt-oss20b
being the MoE it is it actually runs way faster than expected even on this 3080
it's kinda crazy
are there tools yet to train and finetune gpt-oss20b. or do the classic ways just work on it even tho its strange architecture
ofcourse its unsloth
love my boys over at unsloth
How much credit u have ?
hmm should see what happens trying to run 120b @ 1b locally
Nai run somthing more powerful like deepseek
lambda is pay by usage so $0
80$ currently
In my temp account
you could do a lotta finetunes with that
B200 is very cheap ?
LoL
i did like 30 finetuning runs at least in march on lambda and this was the invoice when i was using a mix of A/H/B100s depending on what i was trying
they had a discount on b200 at the time though
70 are compute
i wish lambda would give me credit rather than just openAI
boy does openai love giving api credit out though
thanks sam
altman
ok now that may be the most worthless open with suggestion ive ever seen
are you trying to open a zip? because i think prism can open zip's to install a new instance 
yeah it's a zip lol, it overrode ark as the default after installing it

i dont like ark much and it just dawned on me that there must be a linux 7zip out there or something that i can use instead
not that it's even needed since you can just use terminal to extract shit
but for a right click type thing
7-zip does not store the owner/group of the file.```
that's fine 
windows moment (buying a physical passkey rn)
i entered a password for an older email that i rarely use from bitwarden and looked at it because it seemed short and usually my passwords are like 32 characters+
oh
Hmm
The update is trash
it's just nice of them to keep the things that are exploding after updates to nice, unnecessary things, such as the method people use to login, or perhaps the storage on their drive being wiped. nobody needs that stuff
Every code has own Loop holes no single one has perfect
my computer thinks typescript files should be opened with Movies & TV for some reason
transport stream
i see
"let's just reuse that extension itll be fine" - too many people
but yeah you would see .ts when ripping blu rays etc
the only one i've ever seen is a miku concert that i ripped off a blu ray a long time ago
interesting
mpeg-2
vedal's secret 2
5
7
1
im intrigued
1131629092475768884
neuroCatUuh
1 hour latency
bad bot
running locally

wrong
0 second poll
all answer just race condition fulfilment

i love docker
such a wonderful piece of software
it doesnt even have the right kernel making me panic thinking i fucked up my kernel config saying 6.10-14
You love docker but docker doesn't love you 💔
it certainly doesnt
okay kyuubey
but that's okay because i actually hate docker more than maybe any other piece of software ive ever had the misfortune of using
vedal's link on his github readme is out of date 

I was gonna try and verify that but then I got blocked 🤦♂️
(on the cisco router)
banned 
🌟
🪄🌟 ->
🪄
welp dunno what got automodded there ripperoni but apparently automod hates docker too
lol i see what got modded that's rough - always check your terminal output folks some kernel modules have names that might surprise you
amd-v is enabled
just verified the kvm is enabled in kernel
i checked dev/kvm has permissions set to crw-rw-rw-
compose file has the path /dev/kvm set properly for host and container
and yet docker is repeatedly screaming that kvm acceleration is not available

before you go can you slap my containerd around a bit
kil
need to turn my tech into dumb tech
bwa discord bad for me
i awake at 6am
linux phone helped me with most addictions because its so unusable

i dont really have phone addiction, i just read
sometimes watch a youtube video
sometimes read discord, sometimes browse code
desktop puter is the devil
i have been sitting down all day
its only a 7 months difference apparently
okay what if we cut it off at march 2025
i have 26869 you have 44239

sheesh, i only have 49.5k, even including the mod chats
vaniiii i see why i got automodded can you put a big ol whoopsie on that one because it's not my fault they name kernel modules things that automod doesn't like
oh yea we just ignore automod hits for this chat lol
lmao
like we just glaze over it
another thing that helped me was youtube refusing to work with vpn without signup
since i never watch yt while logged in
so far i'd say winboat is highly recommended, since it uses docker, one of the best frameworks ever, nothing ever goes wrong really
i have too much internet freedom
ive used dns blockers before but i just disable them
Good
get mullvad
vani you got a new chisato
yeah that checks out
nya
i have a new chisato! globally
The compiler got corrupt
Again
Where is
Mysg 32
every time i search for this error it ends up with goddamn yacht stuff
nyan nyan
im die thank you forever
me attempting to get across the strait between linux and windows


ablecorp
true
yep
it runs windows in a container but is SUPPOSED to kinda handle it to make it easier
it's not even winboat that is causing this issue
it's docker
just run windows 
i hate docker i hate docker i hate docker i hate docker i hate docker i hate docker i hate docker i hate docker i hate docker
Windows is crashing and low hardware support
Windows 7 is unsupported
windows 10 ltsc enterprise iot is truly all u would need
Really ?
What's (ignore)
Give me ure microsoft account
don't we all
mfw runnig windows 10 ltsc enterprise iot and the only other machines that exist in the world running the same OS as me are ATMs
I can't download lts version
it's a version of win10 supported thru 2032
lighter than usual
I already know that (my business account is not working on Microsoft) i usually use as Gmail. Personal and business
I aways do seprate personal and business

this son of a gun
COMMAND PID USER FD TYPE DEVICE SIZE/OFF NODE NAME
qemu-syst 24022 afunyun 11u CHR 10,232 0t0 1334 /dev/kvm
qemu ruining my entire life
Mee too
If Linux officially added this feature
( Linux subsystem for windows application )
Extended support hmm ?
this is basically like wsl but in the opposite direction kinda
also now that i've got docker not screaming about virtualization acceleration windows won't let me download the iso for a 3210948130459th time
2025-09-04 00:19:13 | ERROR | Command failed: docker compose -f /home/afunyun/.winboat/docker-compose.yml up -d
unable to get image 'ghcr.io/dockur/windows:latest': request returned 500 Internal Server Error for API route and version http://%2Fhome%2Fafunyun%2F.docker%2Fdesktop%2Fdocker.sock/v1.51/images/ghcr.io/dockur/windows:latest/json, check if the server supports the requested API version
runs this exact compose 10 times trying to get docker to work w/ virtualization
get it to work
"oh yeah we cant find that bro idk"

Root works
its called wine
im talking about https://www.getpaint.net/
Paint.NET is free image and photo editing software for computers that run Windows.
i find this statement increndebly funny since i jsut started watching an anime where a zombie became a magical girl
Z<l
it's crazy how well vocal/instrumental isolation works... I'm especially surprised by vocal
I ran it on NEVER
I don't understand how it tracks the reverb so well
What talking about
(Instrument)
vocal/instrumental stem separation
I guess I could post it here.. I mean, I just ran it through software, not like it's anything confidential 😅
it did pick up a good amount of instruments, I need to find a model that keeps the same fullness but also gets rid of the instruments
stem separation is crazy with recent models
verdict on winboat: seems like it should fkin awesome if it worked
but uh
docker
and uh
many other things
I think this one is like two years old, not even super new
tfw the included operations have illegal shit in them
you can't copyfile on a directory
Ralofriausoifruaiafdlakjfhlkjshglakj
welp winboat needs a few more versions and ill try again
on the plus side i can once again remove the taint of docker from my computer

Roformer or something? If not then you're missing out on perfection
that was specifically UVR V5
What model though? That's what's important
Use Roformer
segment size set to 4000
yeah, I'll try
I used some Roformer model and it blocked nearly 100% of instruments but also a lot of reverb/fullness on vocals
Well use the correct Roformer
BS roformer is the bullshit model, common mistake
eh, I hear less fullness on that one than mine
in one area, at least.. harder to tell overall
Well, this is the model we use to make things ready for Neuro RVC training, it''s the best quality we've ever got
yeah, around 1:27 it sounds really broken up.. mine doesn't sound broken up
but, I also hear a lot of improvements.. guess it's tradeoffs
Roformer is likely why our RVC is so insanely good and anything else you can find on the internet just sounds sloppy
idk, I hear a lot more waviness on the one you posted... it's hard to tell how much better it really is than mine
I hear improvements.. but, maybe some regressions.. I know it's not going to be all improvements
Well it certainly extracts important elements of voice much better than other models, so combined with other cleanup tools to remove reverb and noise, it makes for exceptionally good training data
https://www.youtube.com/watch?v=3jWOyIcKz98
And that is how you end up with this
!! FLASHING LIGHTS / EPILEPSY WARNING !!
Can you teach me to be real?
This collab is a tribute to Neuro-sama and her community "the swarm". Original song is "Machine Love" by Jamie Paige, which is a tribute to "song of the Eared Robot (耳のあるロボットの唄)" by nwp8861.
This video itself took 3 months to make, it also features many MA...
yeah, I do hear it kind of capturing more vocals.. even if the vocals it does capture have less fullness than mine
and it's also blocking a bit more of the instruments.. overall, maybe it is a bit better.. I would not say it's a huge difference by any means, though
If you just randomly heard this could you tell this is RVC?
The difference is big enough to be the differentiation between a barely Neuro overly high pitched mess and a basically perfect Neuro voice
guess I'd only know that if I tried training with them, cause it sounds good to me
maybe mine is overdoing the reverb? I don't know
Yours might just have more noise
Either way, NeuroSynthProject at this point is the leading player in making Neuro voice clones
We've maxed out RVC, now we're doing vocal synth
thanks for the work
I did listen to that a few times already... been months, though.. guess time to listen to it some more
Also check out modern NeuroSynth creations
https://www.youtube.com/watch?v=b1HNHqSdGW4
https://www.youtube.com/watch?v=49CrynNZAR0
https://www.youtube.com/watch?v=UvIr30bUEbw
this has gotta be one of my fav covers i've ever done... and its even got special vocals! swarm is the most dedicated fanbase. ABSOLUTE CINEMA 💜
come chat w/ me on discord if u like! discord.gg/8ry5dn3UjA
NeuroSynth-BETA-3 developed by Superbox & Wispers
(actually amazing work btw)
Motifs:
Raise Up Your Bat (Toby Fox)
SUPER COOL ART BY O...
collaboration with Superbox, this makes it the first (unofficial) Neuro-sama cover of BOOM!
come chat w/ me on discord if u like! discord.gg/8ry5dn3UjA
check out the instrumental version! - https://youtu.be/KSxZScNeJBo
NeuroSynth-BETA-3 developed by Superbox & Wispers
Motifs:
BOOM (Evil) - https://youtu.be/8Nt9YPnYyjs
Only present in instr...
i've been putting off this upload for TOO LONG hoooly
come chat w/ me on discord if u like! discord.gg/8ry5dn3UjA
again shoutout to Superbox for helping inference NeuroSynth! (funny neuro choir backing)
NeuroSynth-BETA-3 developed by Superbox & Wispers
Motifs:
Numbers III - _neurosama
Numbers - _neurosama
Candles - _neurosama
LIFE - Neuro-sa...
IM SORRY WAHT
i have seen and used rvc models outputs of which are actually indistinguishable from a real, high quality studio recording
btw where was neurosynth used in the last one I cant tell
You hear the choir type sounds in the latter half?
no I'm deaf :LULE:
Neuro RVC gets real close maybe, it's really good quality for sure
oh right I recognise it now
That's crazy
It's a kind of NeuroSynth as an instrument tech demo
a certain very huge and very controversial artist which i will not name drop has released music with his verses being written and recorded by a different person using rvc, but if you didn’t know that you can’t tell
It's just a bunch of vocal tracks going "ah" with vibrato mixed really deep
i’m not actually sure whether answering yes increases your chances or lowers them
yes? ok so you can’t do anything yourself filthy vibecoder
no? smh unc doesn’t know how to utilize these new powerful tools 💔
they both lower your chances 
I guess I'll try this one
unless you have a suggestion
BS roformer technically seems to have the highest scores.. but, that isn't everything
That just looks like Windows not working
a cursor on a black screen mhm
I was coding cpp I didn't notice it was py
Hmm error testing why cpp is not working
How weird, why doesn't C++ work in a Python file
(if this was Linux it would probably work if you compile the file with a C++ compiler without caring about the extension)
Ya, LSP is being chosen based on the file extension, mean while compilers don't care.
its pretty common for employers to require doing nowadays because otherwise what are you doing wasting company time smh (in this case they might be asking it for stats though)
i havent seen a single position where vibe coding is explicitly listed as a negative
in what world is vibecoding at a programming job a positive
in ours 
sure working with tools like cursor is a positive, but vibecoding in particular probably isn't one
fair enough
unless vibe coding is explicitly the job contents which i've seen as well
just have an agent prompt the agent instead
99% sure this is a
visual studio error. I, sadly, don't remember the shenanigans needed to configure it.
It's probably faster if you make a new project that's actually a c++ one.
awaawaawaawa
Trying to reverse engineer code from IDA is an absolute nightmare.

ida is fun
but i recommend ghidra if you're learning from scratch
i already know ida and i don't really want to spend time learning another tool
I used to use Ghidra, but it didn't know what the fuck strcpy was.
...
strcpy is like super basic
its not even a ghidra thing
I know, but it just wouldn't pick up on it.
you should probably learn C before getting into reverse engineering
ah
sorry
misread

what language is whatever you're REing written in?
because there are plugins that make everything easier
also check versions for different platforms (if there are any), because sometimes programs are compiled with extra debug info on certain platforms
oh so c++ probably
i use this plugin personally https://github.com/REhints/HexRaysCodeXplorer
also here's an ida thinking emoji
arguably more important
It says it's for IDA 7. And I'm using IDA 9.1.

it looks like it works but requires compiling from source and applying pr #122
sounds a bit annoying
More effort than it's worth at this point.
I'm doing a whole source code reconstruction.
then it should be worth it still
full reverse engineering takes a long time
Unless it can clean up these garbage variables and stupid jumps everywhere, I probably won't bother.
it helps with defining struct/class definitions
This is like a weird early version of C++.
what are other commands i need to know?
there are a couple extra flags you can use to make it fully run in the background for the most part 
what do you have in mind?
the big thing is just running at as admin to get rid of the uac prompts, i don't remember the exact flags 
just disable UAC 
-# is that even still possible
I love how many windows users will just admit that the OS is garbage but literally laugh at you if you use anything other than windows
how do i change the registered owner name?you know the green one
I don't think you can.
i blame my parents
run everything as SYSTEM 
revios can do that for you
but programs usually don't like that
they usually like uac at "never notify"
you can make it no popups, but UAC is still handling stuff
it's just windows
using thousands of meaningless api requests that does nothing
yeah, thought so 
at least the option to disable popups is still there, I haven't checked for it since like Windows 7 or 8 or something
I do have to wonder if it was trying so hard to remove instruments that it caused the vocals to sound not as full
there were very loud instruments compared to the vocals, and it removed nearly 100% of the instruments
I do think the UVR V5 has too much instruments/noise to be acceptable, I mean, it does pick up the vocals pretty well.. but
I'll mess around with a few models and see
scores don't mean much unless you look at a vast range of music.. one overall score isn't going to say much, even then, a score doesn't say a ton about how it did with the whole song 
2
NOW I get why CUDA inference is not enabled by default in NeuroSynth's inference engine
Fixing CPU INFERENCE was way too hard
Now it works beatifully with both though
Enabling CUDA inference: really stupidly easy
Having both CPU and CUDA inference with CUDA 12 + cuDNN 9: really stupidly hard
And all because CPU inference crashes on OnnxRuntime 1.20.0
(and all versions past 1.18.0)
okay so make the dot go away, TEAAAAAAAAAAAAAAAAAMMMMMMSSSSSSSSSSS
Use cuda
du *.qcow2
9999999 disk.qcow2
1 kb need
It's an app that should have both CUDA and CPU available, that's why it's bad that CPU crashes on 1.18.0 and up, since for proper CUDA I need at least 1.18.1
Application name?
NeuroSynthProject confidential information
Hmm ( fix to cuda )
It's written using C# and uses OnnxRuntime and that makes for a weird mess of stuff
We love parallel process
Not single (ai devs)
another week another google drop 🔥
https://fixupx.com/googleaidevs/status/1963634366287933495
Introducing EmbeddingGemma: our new open, state-of-the-art embedding model designed for on-device AI 📱
Cpu must be (npu include)
9 ghz with npu with 9456 tops
9 GHz holy moly
Have you tried it? What's it's use case? I just recently finished my first RAG system, super curious about going deeper
I did mine with MiniLM and it was working really well on memory recall, I don't know enough yet to understand if i need to got deeper on a bigger model like this
Quick reading implies larger contexts from Gemma and more precise semantics. So i guess for fact recall it's kinda overkill
konii is back
Instead of a full break i guess its just less usage overall?
nah that cross react is a paradox
she's working on her addiction
please understand
that's completely fine
this definitely didnt take an hour to finish 
cause i am the stupidest person on earth
silksong released 
so peak
im not gonna play it but im glad for those that will
konii if i were to play silksong id play the original first
dont frick mme
wait

hollow knight is 15 bucks,a nd silksong is 20 bucks, i might buy it sometime
ive already use half of my monthly allowance tho and im trying to be financially responsible, so maybe next month
why are you emoting me so much 
im getting clowned on for not spending everything immediately 
ill wait till it goes on sale okay?
ok now i remove all the reactions
bruh

i screenshotted it already
caught in 4k 1440p
its september, i refuse to get festive already
i aint skipping over spooktober
fuck i was just about to post that one
if i play hollowknight, should i do so on pc or console?
doesnt matter, its the same. use controller though.

well, use keyboard if u want but it sucks
pc maybe better if u have old gen console or nintendo switch
pc 
cuz i dont feel like setting up the xbox one s yet
pc actually, better since u can mod it
i just remembered
u can also pick which version of the game to play on
elvyn gets it
dualsense my beloved






