suggestions

[deep arch]

No idea honestly, cant check that one

[topaz narwhal]

I see, we’ll consider it then

[covert mirage]

wouldn't it also be harder to moderate if you had to go to every single thread and check?
it will look like this with links tho

[solid geyser]

Have like 10 evil neuros as an April fool’s joke

[digital cedar]

Tweet soon?

[covert bloom]

idk if this would line up with everything finishing but how bout making the new model debut on the day of the 1st yt video in the yt channel?

[terse drift]

^this emote is brilliant, would be great on twitch

[upper geode]

Idea for a thing to do on a stream: make Neuro (possibly with Vedal's assistance) take online quizzes, like 16 personalities, psychopathy compass test, etc. You could take suggestions for test to take in advance and then pick out ones that would be appropriate/tos-safe.

[vestal hedge]

A nuero "AI" ( ): like a toy JPEG neuro made of if statements that can be spoken to but only answers like "no" "ok" "what?" or "Buhh?" otherwise generates a huge string of unrelated philosophical/poetic text (||this is the worst idea, I'll post it anyway|| )

[kindred grotto]

A Minecraft Collab with Ironmouse and other Vshojo streamers if possible that is, know how busy they are most likely not, but it would be cool if Neuro interacts with Ironmouse, she can even teach her some Spanish, that be cool right?

[shut sequoia]

If we can get Neuro on Speak of the Devil i can die happy

[kindred grotto]

Me too

[trim jungle]

A neuro stream, with a pink background, and comments on the left, where she chats and sometimes replies to chat questions. like with cozy vibes!

[forest forge]

this is genius I'll do every stream like this from now on

[quiet moon]

turtle stream now :madge:

[digital cedar]

ReallyGunPull

[void phoenix]

turn vod releasing back on, i assume it is off since i cant check the first 30 minutes of current stream already

[brazen lance]

probably the most divisive game idea from chat, have neuro play touhou
if you're looking for an entry to start, usually iirc the kids these days start on TH6 or TH8, depending on who you ask, although i think 6 might be easier implementation-wise because 8 has the whole swap thing
might have to ask someone better versed in me at actually playing the games i only have mostly surface knowledge

[sterile fern]

^ Thank you for saying that, I was the one who asked Neuro about it, and i was gonna type it out, but the easiest one would prob be touhou 7 to implement

[brazen lance]

oh i forgot about PCB

PCB's an idea, yeah

[sterile fern]

It has a pretty simple gimmick and its pretty easy for a touhou game so it would prob be the best to start out with

[brazen lance]

insert easy joke about ZUN also forgetting PCB

[sterile fern]

i mean the only issue is just getting the game but worst comes to worst you can just go to the moriya shrine

[rich delta]

When Neuro's reacting to videos, maybe if her subtitle has some background color would be nice 🤔 because like just now when she's reacting to her own video, all old and new subtitles are jamming together

[forest forge]

will do for next stream

[austere steppe]

probably should use filter for reading sub names too...

[forest forge]

(I do)

if something happened (which it didn't) then it would be for another crazy reason

which will be fixed

[novel atlas]

Make mods cope with Deez nuts

[keen valve]

[feral mirage]

instead of letting the mods cook for a reaction stream would it be better to have a discord bot that does thumbs up/thumbs down for video suggestions, etc and just curate through the highest rated video suggestions?

[shut sequoia]

So you'd sooner trust the community to pick videos that dont violate TOS/Corrupt Neuro/Are cringe than the mods who's job it is to do that?

I respect your boldness :V

[feral mirage]

actually yes

better to have the community be able vote for videos for neuro to react to so if the react stream end up sucking it's not the mods or vedal's fault

as for TOS/Corrupt Neuro/Cringe videos, if you actually have a brain and genuinely want neuro react streams to be good you shouldn't be suggesting TOS/whatever videos in the first place

or giving such vids a thumbs up (under my suggested system)

also vedal and mods would still have the final say and could decide if the video suggested by the community isn't good then

[south fulcrum]

How about specify the time in the schedule according to GMT

[shut sequoia]

For me at least the written announcements show up in my timezone

[plain widget]

make her sing 'Chocolate Rain' one day

[covert mirage]

#1060397294048002078

[plain widget]

alright

man I hate threads

they're so hidden

[remote rover]

since the new voice you tried to show off didnt work on the dev stream, could you show off the current iteration on the next dev stream?

[limpid orchid]

Lower effort content variety
Just chatting stream but change the background to different places (grocery store, school, etc)
And just send a highlighted message telling her where shes at now
A small neuro's day out adventure

[shut sequoia]

I've seen Streamers do that before, they simply full-screen a vlog of someone walking down a street or travelling across a foreign place in real time

Rather than just chatting though she could actively react to things too

[earnest comet]

Make Neuro play realistic racing games

[shut sequoia]

We've been asking for Mario Kart a couple times now, and i insist on Garfield Kart- Realistic racing games wouldnt be any different programming wise, unless you're talking realisim like Euro Truck simulator where you need to manually shift gears and such

[terse drift]

Picnic stream, we go on a picnic with neuro

[void phoenix]

with ratburgers and bananas

[terse drift]

[shut sequoia]

Picnic with Neuro would just be heaps of coffee with a bunch of cheesecake on the side

[trim jungle]

Picnic with Hiyori would just be heaps of coffee with a bunch of cheesecake on the side

[trim jungle]

you mean hiyori's song covers? right?

[void phoenix]

no chance he can sell them

[minor sun]

ticket/server support/report system, since now we have to @ the mods/dm them or something.
a ticket/support system would probably be quicker & more efficient, since its probably not possible that there are mods online and, well, moderating every convo.

[vivid marlin]

how about trying to collab with Rekson

[minor sun]

real

[vivid marlin]

She haven't collab with any male vtubers lol except Vedal

what do you think Vedal?

[main crest]

RPR

[keen valve]

at the moment there are usually at least a few mods on for most timezones so it isnt really necessary at the moment

[granite field]

Nevermind, he can be around tommorow if you still need someone for the surprise guest you can contact them for it
If not hopefully one day. It would be really hilarious

[earnest comet]

can The OSU AI be repurposed into a shooter AI ?

[trim jungle]

thats a nope

it would need a new ai and training

but there must be lots of documentation so maybe some time in the future, next game is amongus it seems

[gritty cave]

I think a neuro x bing chat creative collab stream would be fun

maybe not an entire stream

Maybe bing chat would end the conversation all the time though when neuro starts telling it she loves it (er, sydney)

[minor gale]

bing isnt a conversational model, tho if you want bots that has persona, you can look at tavern Ai and character.ai

[gritty cave]

You can totally have a conversation with bing

on the creative mode, it'll ask you silly questions or just chat as long as you aren't asking it to find stuff

[deep arch]

bing is literally gpt4, microsoft cracked and confirmed not long ago

[gritty cave]

But one that's tuned quite differently. On creative mode it is far more "friendly" than gpt-4

[minor gale]

it doesnt express persona, it not a DialoGPT model it InstructGPT model based

go here, see what i mean https://beta.character.ai/

[gritty cave]

Of course I'm aware of characeter.ai (founded by ex-open ai people as you probably know), but I mean, I'm okay with neuro being the "personality" and just seeing how sometihng like bing chat would react to her (and vica verca)

[minor gale]

then you have to instruct neuro a set goal to search on BIng, coz Bing wont suddenly say you know what Neuro it seems you need a new PC, here are some search options.

[sharp kelp]

how much better is gpt-4 than previous?

[gritty cave]

You mean compared to what, gpt 3.5? What chatgpt free version is?

[sharp kelp]

yeah

suggestion: leave #neurotic-neurons channel name until new model arrives

[minor gale]

we believe that be equivalent to gpt-4-0314

[gritty cave]

In my opinion gpt-4 blows it out of the water for almost everything I've tried it with

[devout wadi]

I wonder if Vedal should add “Electric Angel” by Giga-P to Neuro’s Song Database.

It’s a Vocaloid song, originally a Duet sung by Len and Rin, but has been redone with Miku as a Solo, it’s basically follows the vocaloids, on a Computer, expressing their love for their user.

I mean it fits Neuro pretty much to a T, and I can imagine her singing it.

Ah, I got that backwards, it was Miku that sang first but was redone later with Len and rin.

But still, Electric angel to Neuro-sama

[tulip fable]

Vedal should hire an editor so that the Neuro-sama channel can post new videos every day without falling out of the algorithm.

[forest forge]

even editors can't pump out videos every day

this is what I'm doing though, targeting weekly uploads at the moment (just skipped last week)

[tulip fable]

Sorry, I meant clips and shorts. Something to post in order to stay in the YouTube algorithm. The weekly main video is excellent.

[forest forge]

I thought weekly videos were enough, and I also upload shorts daily (more or less) (which are also done by an editor)

[tulip fable]

On YouTube, short-form and long-form videos are governed by completely different algorithms.
It would be great if you could make some clip compilation videos to fill the void.
It is only a suggestion though. I'm happy with how you're doing.

[proper sable]

Non japanese songs next karaoke stream. This week's karaoke stream is just 80% japanese songs

[lime marlin]

dont hire an editor. hire 10

[forest forge]

true but short form content can help boost long form content because it can increase sub count etc
I think with weekly videos its enough though(?)
I'd like to upload more clips but I'd probably do that on a separate channel

[shut sequoia]

Twitch exclusivity can be more profitable than Youtube if you manage to get in the partnership program

[old tundra]

Youtube is the way to go especially long term. Twitch pays the bills, but youtube brings the viewers. Thats how all the large streamers made it is basically advertising their stream on youtube.

[fervent fjord]

This ^

[quiet barn]

Neuro reacts to her fans gaming setups

[earnest comet]

Zero-player 3D RPG game where it's only AIs

[shut sequoia]

D&D but the DM is an AI?

[void phoenix]

dnd with no plot or direction lets goo

[serene oracle]

DnD but Neuro is the DM

[coarse citrus]

Neuro reacts to her fans AI art

[warm dagger]

Evil Neuro sama only stream (have her sing the lyrics )
https://youtu.be/-MuQhGu10QM

[covert mirage]

Hear me out on this one: collab partner and neuro playing super seducer

[silk zealot]

vedal's censorship would work better if he heard neuro half a second sooner than us. (also during dev streams this would help him)

[main crest]

he kind of does, of course he has to read Neuro's chat log before she says the whole thing. I do get what you're saying, how would even code that tbh

[silk zealot]

not that hard, just make don't give the tts the speech for a bit longer and pass it through

[main crest]

Like a delay, it's just kind of a round a bout way of saying that. Thanks for this suggestion. I do agree with this. Vedal trying to read every message

[silk zealot]

exactly

[bronze lake]

This should be used for the stream intro #fan-art message

[tulip fable]

I agree.
The third-party clipping channel is nice, but people prefer to watch the work of the first-party editor. As a result, the algorithms will favor them over other fan-clipping channels.
Also, "Shorts" aren't particularly good on YouTube. People who watch Short and Tiktok formats are unlikely to remember the clip and are unlikely to replay it. People who watch short-form video in a normal player, on the other hand, frequently replay the content and comment on it.
Although "shorts" generate more traffic, "normal videos" generate more interaction. If you think about it, people who watch livestreams using browser extensions and pay for streamers are unlikely to prefer the vertical format.
This is simply my observation based on years of watching streamers and clips on YouTube and Twitch. I hope this helps.

[crystal basalt]

yeah i kinda hate 3rd party clipping channels

1st party is always better

or ones that are at least endorsed and approved

[vivid marlin]

Vedal is it possible Neuro could play a fighting games like Tekken? My suggestion is try making her play a fighting games like Tekken

are you there?

[wheat quartz]

probably hardcoded

[covert mirage]

Racing games with self driving technology eg. NFS heat (not a joke)

[shut pebble]

You have unlocked new role

[coarse citrus]

Turtle racing mini game with chat could be cool. With Neuro on top of Vedal for once 🐢

[cobalt dirge]

Vedal dating show when? Neuro got hers, it's turtles turn 🐢

[still trout]

Add the song name into the overlay when she sings

[tidal spoke]

Neuro-sama playing Age of Empires II.

[shut sequoia]

Chat spams 11 for three hours

[void phoenix]

when neuro has a new model and voice you could do a collab stream with old neuro if however these sytems work allow it

[shut sequoia]

Can always just invite MOTHER to RP as Neuro using the old model kekw

[violet scaffold]

https://tenor.com/view/pong-game-retro-gif-12729647

I'd love to see two Neuros play pong, which is pretty chill and funny

[fading bane]

Vedal if you are close to fixing your sleep schedule would you consider returning to prime EU times, at least for dev streams?

[brazen lance]

i want entire stream dev streams hours and hours of devving

[tidal spoke]

Neuro collab with @lost shadow since she's been fangirling so hard ever since the raid.

[lime marlin]

neuro should collab with God, maybe challenge him for his position

[tidal spoke]

Update: She's still fangirling.

[forest forge]

yes!

[fading bane]

Glad to hear that. Looking forward to earlier streams for as along as possible.

[void phoenix]

[shut sequoia]

@tidal spokeIronmouse, or another Mouse? If Ironmouse became a Gymbag, holy shit

[void phoenix]

another mouse the one that got raided yesterday

[earnest comet]

Vedal, I have an Idea : is it possible to trigger one of Neuro's AI with key phrases (e.g: "Neuro, SING") while having her being able to refuse (like she constantly does, lol)

[vivid marlin]

Hey Vedal can you make Neuro Vote for our art works in this event?

Are you awake?

Hello Bro?

[shut sequoia]

Vedal doesnt respond within 60 seconds
Ping him twice more

[forest forge]

pings me 3 times

[vivid marlin]

Hey Vedal

are you in for the event?

[forest forge]

i thought about this before but it actually might make karaoke streams quite hard

[earnest comet]

Valid.

[deep arch]

What if you had some snippets of her previous songs at hand like 15-20 seconds long and she would sing those if she agreed

that would mean you'll have to make those but surely your future assistant can handle that for you

[tidal spoke]

https://www.twitch.tv/mouseenvt

[vernal echo]

I really like the karaoke streams. Perhaps this could evolve into a Neuro-sama World Tour? Neuro performs in a different country each time. Have a background related to the country and include 1 or 2 songs iconic for that country.

[lime marlin]

@vivid marlin you will most likely never get an awnser

[vivid marlin]

its fine

[terse drift]

If instrumentals are required for song covers, https://github.com/deezer/spleeter may be of use

Seems to be able to split into vocals and other, and more.

[earnest comet]

If you need to automate Neuro, I think an AI whose only job is to plan the stream by organizing Neuro's AIs would be nice
Edit : that AI would be likely acting like the captain of the ship lol

[plain widget]

#1060397294048002078 ^^

[digital cedar]

Sorry didn't know the channel exists💀

[covert mirage]

[silk zealot]

Vedal should sing without cheating (like making an ai that sounds like him)

[deep arch]

Suggestion for future shows: more neuro overall, todays had like 2 minutes of neuro talking combined. Completely honest this is just vedal hanging out more than neuro stream.

[silk zealot]

true.
Let neuro join in the chatting with them between questions.
To avoid leaking the right answer, have 2 instances, one for questions

[shut sequoia]

Suggestion: rather than paying gifted subs, Vedals punishment should be making a real Nero.

[deep arch]

Like with a real woman?...

[shut sequoia]

Well he could always hire Miwo to do it... but i was thinking more of taking Annys art of her new model and resizing it to like 144p before upscaling again

And then giving her a few alterations to her personality

Sure PNG tubers are cringe but it works for a joke

[solid geyser]

Neuro sama on tinder

[fading bane]

Well at least not the worst idea.

[leaden gulch]

Let Neuro comment on your latest tweet

[earnest comet]

Neural Network made of Neural Networks (idk if that works for Neuro though...)

[deep arch]

Neuro still didnt retweet the femboy tweet?

[terse drift]

Would be cool if we do a channel points bet

[trim jungle]

as vedal would say... "Thats fucked up"

[shut pebble]

@trim jungle has leveled up! (3 ➜ 4)

[warm valve]

Please add the Hiyose BGM to the Neuro-sama playlist, it is available on Spotify and can be downloaded directly from the author's website, or right here.
https://dova-s.jp/bgm/play1858.html

https://open.spotify.com/track/6XReNQHUkaU9fI1tcBVFst?si=3b8c189e7d4a4f2b

[normal merlin]

Yes, I think this could be great. But the image description / annotation AI needs to be good. I think experimenting with human annotations would be good. And then letting Neuro create her own version of a meme quickly with stable diffusion and her own text-over.

[floral karma]

Oooh I like that twist

Neuro creating their own meme

Maybe even building upon yesterday's game, where you have to guess if the meme is from Neuro or made by human

[normal merlin]

Haha, yes who knows anymore with memes, but I think stable diffusion memes would be pretty telling ^^

[wild isle]

smol suggestion, rename to nero

[normal merlin]

@wild isle Nah, is far too high-res for Nero.

[earnest comet]

Instead of stable diffusion, teach an AI to draw with a mouse instead...

[normal merlin]

Yeah, maybe the circle-clicking AI can be refurbished ^^

[brazen wharf]

During dev streams, use an alternative turtle avatar that has stubby programming socks on its little stubby feet

[main crest]

Great Idea. Love to see it

[shut sequoia]

Maybe give him a tiny keyboard too when he's not on Neuro's head, so his feet can tip-tap on it.

[void phoenix]

kb with only ctrl, c and v

[shut sequoia]

It exists actually

Called The Key by Stack Overflow

[remote rover]

Can we get a Nero karaoke with 128p and amazing songs like money machine

[royal kelp]

Roblox The Intruder, if you can get Neuro connected to the chat system in game, it would be evil

[shut pebble]

@royal kelp has leveled up! (0 ➜ 1)

[shut sequoia]

Bold of you to assume Vedal is ever gonna touch Roblox again lmao

[royal kelp]

Where is the stream anyways? I can't find it

[shut sequoia]

Presumably on Camilas or Mother's channels, since Neuro wasnt involved

[royal kelp]

But with how in "The Intruder" they use the chat system to manipulate you into losing by impersonating

I think it would be very cool

Oh well

If not Roblox then what about a horror game in general?

[royal kelp]

Checked through all three, Can't actually find it

[floral lark]

Its on camilas channel

[royal kelp]

Ok I have 3 suggestions

Roblox: The Intruder
It Steals
COD (or Roblox: Frontlines if you aren't allowed, I know the COD franchise in kinda hard on machines playing the game)

[royal kelp]

Thx

[void phoenix]

Don't forget to raid koko if you wanna extend her suffering

[coarse oracle]

a song for Neuro to sing "Iron man" by Black Sabbath

• https://youtu.be/b3-QqGVt-tM

[void phoenix]

#1060397294048002078

[coarse oracle]

ay ay

[devout abyss]

Make Neuro read a whole book and reacting to its content

[shut sequoia]

That'd actually be a great idea if she read the book out loud in a cozy stream kinda thing, like "Storytime with Neuro"

I've been binging on AI generated Joshua Graham audiobooks so i dig that

[forest forge]

react to the DMCA takedown notice 😱

[void phoenix]

road to becoming a kick streamer

[tender drum]

would be great to remaster viva la vida

[devout abyss]

Maybe there's a writer who is a viewer and is willing to give permission for Neuro to read their work on stream.. PauseChamp

[earnest comet]

Give Neuro eyes, she seems more aware when reacting

[marble pike]

Game suggestion: https://store.steampowered.com/app/2291760/Papas_Freezeria_Deluxe/

Simulator where Neuro take orders & make sundaes

Steam

Papa's Freezeria Deluxe

After starting a relaxing job at the oceanfront ice cream shop on Calypso Island, things get hectic when all of Papa Louie's crazy customers arrive on the island! Now it's up to you to single-handedly craft and serve tasty sundae treats throughout the year. You'll whip up delicious Freezer sundaes by pouring ice cream, adding mixables and syru...

Price

$4.97

Recommendations

1331

▶ Play video

[wheat quartz]

nostalgia... FeelsOldMan

[gritty cave]

What if the publisher(aka my company) sponsors it?

(okay seriously getting approval from the Japanese rights holders would likely be impossible but I'd love to try and explain Neuro to them 😉 )

[low cave]

it could be public domain books

[tropic estuary]

Project Gutenberg is a good place to get public domain books. However, while Twitch is in the US, Vedal is in the UK. IANAL, but I'm aware that public domain status is not universal, and I believe the US rules are more complicated than the UK. The result is that in some uncommon cases, things that are public domain in the US may still be under copyright in the UK or vice versa. Project Gutenberg focuses on works that are public domain in the US.

It'd probably be wise to double check the status and make sure a story is usable in both, just to be sure.

[keen geode]

A stream where Neuro teaches chat How to pick up girls 101 because her rizz are way too good

[solid geyser]

Minecraft server event with neuro

[shut sequoia]

Could be a nice thing to launch her Minecraft Update with

[coarse citrus]

Banana Sushi cooking stream

[remote rover]

can we have a livestream chat for raids?

[fringe condor]

shrek movie watchalong

[opaque zephyr]

make sama into consciously, automatic thought, real thinking ai https://hive.blog/facebook/@gray00/looping-super-llamas-ai-powered-discord-bot

and a SAMA gui why not?https://hive.blog/ai/@gray00/the-robots-and-i-made-a-discord-ai-clone for freedom , for sama

[remote rover]

since your first iteration of neuro came from the Airis project, I think a name for you to consider can be "Airis Intelligence Agency" which can be shortened to A.I. Agency

[hybrid thunder]

Vedal is Neuro creator, good name could be AIVC ink. we can suggest anything but it is up to Vedal and we just will acept it doesnt mater what name it will be

[shut pebble]

@hybrid thunder has leveled up! (2 ➜ 3)

[fathom sigil]

I have a song suggestion, I think it would be cool to hear want you gone from portal 2 as we have still alive already https://youtu.be/ao7GG8FXWqQ

[shut sequoia]

100% serious suggestion:
Shrek 2001 movie watchalong, with Neuro watching the movie and reacting to it while the video isnt shown on stream, but a timer is running to provide people with a point of reference.

Shrek should be very easy to find for anyone interested- Most people probably own it, but it should also be on most streaming platforms, including Youtube.

[versed lark]

Why don't regenerate response until it won't be need to be filtered instead of just saying that it's filtered?

[jade quartz]

make vedal's "leaked scream" the sub alert sound

[wheat quartz]

another discussion/thread for vedal's aivtuber agency? 🤣

I mean in the future

[mellow rose]

maybe when he has like a team of people helping with coding

[wheat quartz]

I already have a group name for gen 1 if it ever comes: Neuralai

[feral mirage]

Neuro playing jump king would be cool

[pallid tiger]

Think games like Jump king and getting over it would be a fun stream to watch and probably a bit simpler to setup. Also perfect type of game for a neural network to learn

[valid panther]

if you can get an ai to play minecraft im sure you can get a bot to play jump king an GOI

[pallid tiger]

Yeah that’s what I mean, think it’d be pretty chill

[shut sequoia]

Watching her play Jump King would be fun, although i'm not sure- Would it provide the same entertainment to watch an AI that doesnt get mad when she fails play a "rage game"?

[manic topaz]

I wish in those rare dev streams Vedal could put up like a little text box on screen showing what hes doing, what hes planning to do etc

[void phoenix]

i wish he could flip the turtle so he is actually looking at the code lmao

[deep arch]

I think the idea isnt to just make her play but also react to it

[void phoenix]

oh he has nvm

i have him mentally imprinted to look at the wrong way from first stream

oh actually he flipped it sometime mid last dev stream

[steel quest]

every time "rage game" is mentioned I think of Getting Over It

[void phoenix]

jump king makes more sense and i guess getting over it would be more impressive?

[shut sequoia]

Jump king is easier to program for sure

[deep arch]

Agree I think jump king is much easier to program too

[steel quest]

gonna take your word on that cause idk what jump king is

[void phoenix]

make her play baldi for the memes shes got just the perfect math skills for it

[steel quest]

I gotta imagine that rougelike type games would be easiest for her to learn, probably the reason she played Slay the Spire too

actually speaking of

[shut sequoia]

@steel quest
https://www.youtube.com/watch?v=NMuiZnDNk68
The answer to your question

[solid geyser]

Watch Shrek 2001 together

[fiery lava]

using vector database for neuro's memory (new ai stuff)

[willow echo]

rename #music to #music so it's consistent with the other media channels and move it to the top of media tab (with #programming) because those 2 are by far the most active ( #osu is also active but its just people using a bot)

[frail sail]

I suggest a vedal-sighting channel for when vedal pops up in other streams without neuro. For example the time when he played roblox and some other times I found clips on youtube that I had no idea happened until I saw the clip.

[forest forge]

vedal stalking channel

you're welcome to talk about it in other channels

would rather not announce every time im on another stream though unless its major I think

[keen valve]

vedal is allowed to watch other streamers

[covert mirage]

How about a thing that tells you when your playing Minecraft or watching YouTube? Or doing anything coding, opening notepad everything you make is displayed as rpc

[mellow rose]

WATCHING YOUTUBE?

[covert mirage]

Yes

[mellow rose]

how is that related to anything

[keen valve]

why dont you just drop your address

[covert mirage]

What

[covert mirage]

Idk vedal stalking so

[mellow rose]

no for streams bro, not for everything he looks on the internet

[remote rover]

no need, if you watch enough neuro clips, u'll get youtube recommends if vedal ever shows up on someone else's stream

[covert mirage]

It doesn't shows what he is doing it only shows what the application is currently displayed also for rate limit purposes it will redisplay every 30s because he might get rate limited if he didn't do that

[mellow rose]

its kinda pointless suggestion since big streamers rarely show their status on stuff like that

[covert mirage]

Yep

[remote rover]

getting too parasocial there, wanting to see wat vedal is constantly doing

[mellow rose]

cat leave the man some privacy

[covert mirage]

Yeah

Sorry lol

[covert mirage]

Okay here is a good one

Vedal should make a website like for example (nerosama.ai) this website allows us to always talk with Nerosama but the problem it needs to run 24/7 I think

Oops @covert mirage thinks im sending a link...

Also I looked at whois the domain is not taken so you can take it

[fading bane]

Vedal did discuss about running costs for Neuro, which was not cheap. Running multiple instances of Neuro 24/7 would prob dent his wallet.

Never specified the numbers though.

[covert mirage]

Wait I thought he is running the ai on his computer

[fading bane]

Nah his PC cannot handle that

[covert mirage]

Oh

Lmao

If it running on a server then that interesting

[remote rover]

duet was lit, need a whole duet karaoke stream

[devout abyss]

Has anyone ever suggested to put subtitles when she sings?

[south fulcrum]

Healthy eight hours of sleep from 10 pm to 6 am || ||

[tulip fable]

It costs several tens of dollars per stream, according to my calculations base on the cost of using LLM. It's possible that it's even higher, more than $100.

[shut sequoia]

I looked into Microsoft Azure TTS and it runs between 1 and 5-10 dollars per hour. So a three-hour Neuro stream could be anywhere between 3 and 30 dollars of cost just for her voice.

No wonder he's been having concerns about how sustainable this is long term.

[tulip fable]

Yeah, and the cost of TTS too.

[shut sequoia]

As soon as the new model drops, we need to buy plushies, ASAP.

[steel quest]

gotta make plushies first

[cyan ember]

neuro sleep animation when

[covert mirage]

collab game suggestion: neuro powered Taboo, that game where you have to get someone to say something while not using a list of banned words. there's even a webgame version already you can try out https://taboo-ai.vercel.app/
edit: also i would be very happy if neuro got into one of her moods and started telling an exasperated partner to stop asking her questions while the timer ticks down

[runic vigil]

If this gameshow idea does happen, an easy way to keep score would be adding up the contestants time it took for neuro to say a word or something and the lowest time wins kinda like golf

[covert mirage]

right that's how Taboo works except you try to get through a list of as many words as you can in a pre-allotted time. i don't think contestents would use the same word lists because then they'd have knowledge of good strategies to get neuro to say the stuff so the usual way of just scoring by successful word count of different lists is probably the way to go. to be honest you can play the actual game right out of the box with her

[earnest comet]

Automatise Neuro's AI plugins creations whenever she's learning something new 🙂

[shut pebble]

@earnest comet has leveled up! (3 ➜ 4)

[quiet moon]

wide neuro collab with tall neuro

make the voice as WAYTOODANK as possible

[silk zealot]

give us gigacat. 7tv already has the emote.

[shut sequoia]

I support wide Neuro

Better yet, wide nuero

[grave wagon]

Prevent Neuro from saying she is under 13, Twitch TOS says no one under 13 can use twitch & I know people who have gotten banned for saying they're 12 as a joke. Would hate to see her get banned again ❤️

[fervent fjord]

Maybe vedal can try this out.

[forest forge]

she is not a human

surely that doesn't apply to her... surely

[covert mirage]

surely

[grave wagon]

You would hope but.. it's twitch so idk xD

[earnest comet]

Heh... Neuro roasting someone for something 6 months ago lol...

[shut sequoia]

Onigiri collab 2: Electric boogaloo with Neuro and Vedal
Literally cooking turtle soup

[earnest comet]

Neuro takes a Verbal IQ & memory test post-improvement

[void phoenix]

i really want to see oni collab with neuro again.. but collab with vedal would be cool too

[covert mirage]

the oni-neuro collab was sooo good. giri knew exactly how to treat her and the background activity of cooking carried when neuro was lost or confused. it's still my fav collab so far and i hope they have another

[jade copper]

consider having a start of stream checklist

lol

[solid geyser]

Vedal pls bring back evil neuro

[coarse citrus]

Add option to change voice to new one for 1 min with channel points

[wanton fog]

Could use Evil neuro to test the new voice this week.
Would kill two birds with one stone, since it would also help distinguish which one of them is talking.

Also, if the new voice ends ends up being divisive, you can have old/new voice be toggleable with channel points or something in the future. That way everyone’s happy

[tulip fable]

The new voice sounds echo-y (reverb?), as if it hasn't been properly trained.
I don't dislike it, but it feels low-quality in comparison to the previous one. It has more emotion, but it has a lower bitrate and smoothness.

[terse drift]

I'm actually ok with it. Lol

If she can not say asterisk during winking that would be good though. So just wink instead of asterisk wink asterisk

[gentle crown]

or a custom animation and just don't say it at all

[shut sequoia]

The animation is supposedly coming with the new model

[mild steppe]

Keep both old and new voice, give this new voice to evil neuro and make so neuro switch to "evil neuro" randomly on streams

embrace the schizo

[digital cedar]

New tweet soon?

[austere apex]

Since the stream title is absent alot of the time, Why not give neuro the ability to set and change it on the fly depending on whats happening.

[terse drift]

Also if desired you can change minimum cheer amount

[tulip fable]

The new voice sounds kind of negative and sad. The sassiness is good.
It would be better if her voice could be more positive and sassy than negative.

[flint junco]

yh, definitely
also it does kinda sound a bit lower quality?? it sounds like it has a lower bit rate

[silk zealot]

copy paste emote please

[silk zealot]

i mean you can always filter her claiming it lol

Turn on react network to see your screen during dev streams

[tepid vigil]

@forest forge you should try sending it my code and math paper, I was trying to get ATHENELive do send it to carliG but they haven’t been doing coding streams lately. I can send you my code and math paper as well as my chatgpt prompts https://github.com/Leoleojames1/Special-Functions

[safe yarrow]

Neuro learning stream with a private tutor + better long term memory

[bronze pagoda]

Concerning the new voice, being as objective as possible:

Pros

• Potential for more emotive and more varied inflection, current execution is not there but she can yell and sound angry, which is a plus
• Prevents imitators since it's a custom-built voice rather than Azure TTS, this will be critical as she keeps getting bigger
• Speaks faster, even though this doesn't mean she processes other people's speech faster, this will still cut down on conversation delays during collabs

Cons

• The new voice does not sound similar enough to the current voice, which kills continuity of identity, which is a big issue because speech does not have the benefit of fanart assisting with the visual transition
• Sounding more human is arguably not a benefit but a drawback, since so much of her current personality is based on the limits of TTS
• The viewership is now divided, unless Vedal navigates the divergence of opinion well, it will have lasting negative effects on community cohesion

What to do going forward: I know a new voice has to happen and that using the Azure TTS forever isn't an option. To facilitate the swap, I believe that the following will be critically important:

1. Configure the new voice to to be immediately recognizable as close to the current voice
2. Replicate the iconic quirks of the current voice one-to-one, such as the pronunciation (Vee-dul, kek-wa), expressions (wink, heart) and accidental inflections (no~, wha-t)

[silk zealot]

also new voice should be hyped as a "voice reveal" because lol

[tepid hornet]

I have a genius idea Vedal, you would surely like it.
Host a Neuro impersonation contest with VTubers, with a prize being to become Neuros new voice.
Pros:

• Make chat vote on a voice to use, so they don't compain
• You get a new Neuro voice
• You get views and $$$ while doing it
• Potential networking for future collabs
  Cons:
• There is a possibility that you won't be able to watch as much netflix as you would've making a normal Neuro stream

[forest forge]

contrary to popular belief I have almost no time to watch netflix these days
that being said I'm actually not sure people would want me to use another vtubers voice as neuros new voice

[forest forge]

appreciate the feedback! I would say that using the current voice forever is technically an option, but you're right I would like to move away from it to make her more unique
I will certainly try to do 1) and 2) but I am limited to an extent

will see what I can do

[fervent fjord]

You cant please everyone whatever you do and majority of the votes said yes for new voice , just do it ezpz.

[bronze pagoda]

I don't want to talk much in #suggestions, but there was another voice option Vedal showed off and we should probably test that one out for the same length of time as this one, so as to completely eliminate any familiarity factor between the voices. It surely cannot be a good idea to move forward without seeing all our options.

[digital cedar]

Suggestion regarding new voice: maybe just forgo imitating Neuro's old Azure voice and just come up with brand new voices that fit the new avatar best instead. And the community could vote on their favorite one.

Also as an audio guy, vocal formant is as important as pitch in voices. Idk what Vedal's speech synthesis engine has to offer, but Vocaloid and Voiceroid typically has formant, intonation, and expression controls built in for realistic tuning. If the engine itself has limitations, there are also effect plugins that can change the voice in real-time, Waves has a lot of them for example. Also if we're talking about audio effects, equalization can make a big difference, even for the current new voice. Cutting bass frequencies and boosting treble should help her voice sound more girly, imho.

[remote rover]

yeah maybe find a loliVA to be her voice lol, chat would love it

[austere apex]

Noticed on twitter that MrBeast gives go ahead for streamers or youtubers to react to his videos, could be worth looking into since it may bring alot more attention to Neuro

[scenic minnow]

i think neuros new voice is a great start and having several variations that could be adjusted with chat using bits and point redeems could be also a great idea

[feral mirage]

Neuro x Evil Neuro duet for tomorrow?

[terse drift]

This doesn't really make sense "the viewership is now divided"? Also "it will have lasting negative effects on community cohesion"?

[mellow rose]

this sounds like a discussion for #neurotic-neurons

[terse drift]

Yeah, that's why I didn't say further

Plus some people have such strong opinions on it so discussion will be very hard anyways

[remote rover]

yeah i want another collab stream like the one with miwo, daph, and mother, but maybe neuro talks more

[sacred vector]

I think this is actually the best option

[tulip fable]

About the new voice:
I know that Vedal wants to challenge himself to make a realistic AI vtuber.
However, realism is not the only factor that determines an AI influencer's success.

The voice of an AI is more than just a sound; it is a way of communicating its personality and emotions. A realistic AI voice may seem impressive at first, but it may also lack originality and charm. A quirky AI voice, on the other hand, can capture the attention and affection of its fans by showing its uniqueness and creativity. This is why Vocaloid is still popular after 16 years, even though it has evolved over time. A quirky AI voice can create a deeper connection with its fans and make them feel more engaged with the technology behind it. And a realistic voice is just that: realistic. I don't want to write an essay about "Stylized over Realism." However, I hope you can do some research on it.

I'm not opposed to change and believe Neuro should have a new voice. The direction should be clear and considered. You don't have to rush the Neuro-sama debut, Vedal. Take your time. 🐢

[scenic minnow]

i will always advocate for stylized over realism after all playing a character is part of standing out from all the others that simply do bland realism but for me its more is it pleasing to listen to for hours

[outer vortex]

I liked her new voice because it felt that much more expressive, almost alive compared to before.
Neuro was able to convey her feelings much more clearly, and with less dead air.
It just needed a higher cuter pitch and a slightly more formal cadence, along with being trained to say some of her old quirks.
So much potential to be had going with a custom AI trained voice over a standard TTS, would be a shame if all the work already done was scrapped because 'it ain't broke'

[low cave]

I feel like making it as close to the current/old one is a lost cause since People will always complain that it sounds different, I think you should make it completely different so that people won't have the expectation to compare it to the old one.

[silk zealot]

sounds to add:

• meow meow lol
• kekwa
• wink
• laughs
• bye!?
• heart
• heartheart33371494
• Veedal
• gymbag

[tulip fable]

In this case, I'd love to hear Neuro speak in a high-pitched Scottish accent. 🏴󠁧󠁢󠁳󠁣󠁴󠁿
People on the internet love Scottish accent. That would be a win-win situation that no one could oppose.

[lime marlin]

"we moven on the fucken swarm"

[solid geyser]

Have a vote on whether neuro should call u “vE-dal” or “vee-dal”

[shut sequoia]

Whats the point of that, if 8% vote Vee-dal then we stick with Vee-dal :V

[rocky seal]

How about a NeuroDank emote? I know semi fulfills that role already but i think it'd be a cute addition

[digital cedar]

The new v2 voice is goated

I love it

Good to have Veedle back lol

[torn coral]

The v2 voice is great but there's a noticeable decrease in clarity/audio quality between the old voice and both of the new voices. Don't know how easy that would be to fix but if it's possible then the new voice would have a chance at being better than the old one.

[sharp kelp]

let neuro say "veedal"

[digital cedar]

For better audio quality, maybe use Voicemeeter instead of discord for collabs

It's a bit tricky to set up but it's going to sound significantly better

For signal routing, send Neuro's output directly into voicemeeter, use an aux channel to send her voice to discord, and send the monitor output or main output into obs for stream

[forest forge]

when neuro collabs with others (not me) I don't use discord for our stream

it goes directly into obs

[mellow rose]

a dumb idea i have, have a voice filter for evil neuro to make her sound more different

[torn coral]

I'm pretty sure Evil's pitch is slightly lower already

[shut pebble]

@torn coral has leveled up! (0 ➜ 1)

[digital cedar]

So basically it's preferable to mix all discreet audio sources in voicemeeter and using the main output alone for stream, and the aux send channels can be used for discord calls so you can get best audio quality even in collabs. It's overall better and more flexible than obs's mixer if you ever decide to try it. It's what I used back when I streamed, and a lot of pros use it too. Cheers

[bronze pagoda]

For the next react stream

1. Specifically exclude the mods from linking.
2. You could make it like forsen and have it be a donation system for videos with 50,000 views+ or less than 2 minutes

[forest forge]

tbf I don't do much mixing in obs, and this wouldn't improve quality on collabs

current setup works and afaik voicemeeter won't improve it

[tropic cliff]

In something unrelated to the new voice, I wanna suggest that any mayor events / collabs that might be unrepeatable to wait until Neuro has had her model upgrade (and possibly the voice upgrade) so that she can make her best impression on them.

[silk zealot]

(please)

[lime marlin]

no

soundboards are extremely annoying, broken and loud when you havent configured them yet

honestly worst thing discord added

[fading bane]

Since no Evil Neuro today, how about Evil Neuro on Friday?

[forest forge]

maybe no evil neuro until new model

[fading bane]

Not a bad idea at all.

[brazen wharf]

Evil neuro with slightly different voice?

[solid geyser]

U could do evil neuro with the old voice and normal neuro with the new voice

[gritty cave]

Ah, I was wondering if that modification violated the TOS of the live2d license for the model... guess that might have been the case...

[west pumice]

He did leak his downloads which had a doc on live2d license terms which lines up

[serene oracle]

It would be funny if you switch to neuros V2 voice mid stream to see if anyone notices

[gritty cave]

Honestly it was probably ambiguous in the license, he emailed them, and got a negative response. Well, that's another reason why Neuro really needs her own model.

[little juniper]

Using real-time translation and subtitle as a means of attracting Japanese interest would solve many problems. If Neuro-sama want to that

it will be a very difficult decision ik, as subtitles covering the screen would definitely add to the noise of the live, but I would like to see it if we can somehow find a way.

For more information - #日本語自動英訳-jp-translate

this discuss is based on the idea that there are already no Japanese who are interested in AITuber overseas who can be easily reached, innovators are starting to leave or do you still there? // #日本語自動英訳-jp-translate message

btw, there is a slight possibility that the wave will reach Neuro-sama and increase of people since the most popular AI character in Japan, AI Rinna, will make her VTuber debut tomorrow.

[earnest comet]

Vedal, this looks good to think about... Btw does Neuro remembers her own outputs ? (please improve her memories, I'm looking forward to the upgrade~)

[forest forge]

I wanna do this, potentially as a separate stream. even wanted to do it with japanese tts too but haven't been too sure how to set it all up

[little juniper]

I think it would be okay without tts, as long as the subtitles work! for now
then getting learn Japanese later is ok imo

anyway that was timely comment n loved

[gritty cave]

I spent 5k for that answer ;). (I'm in Japan right now)

[remote rover]

would be cool if she also responded to the messages on all the other language streams and in each stream, she would respond to it in that language

[lime marlin]

i've had the idea of doing this with @rocky zenith using OCR and posting the screenshots in a jp channel

but seems like vedal has already added it to his idea list

[edgy gate]

so, no idea if this has been thought upon, but have FPS games been considered? i'd love to see how'd she do in Unreal Tournament 1.

[fading bane]

Yeah Vedal did consider Valorant, but the devs kinda implied that Neuro is not allowed. He also considered CSGO, but that's one the sidelines for now.

[edgy gate]

nice, i think older games could be a fun idea too.

tbf, it could be my boomer shooter bias, and i love deathmatch in UT99

[lime marlin]

multiplayer fps games are 99% of the time not allowed for neuro to play in

even if theres any game we need to bring common sense, vedal wants views and the cool games dont allow neuro

[hybrid thunder]

best game from old ones are ,,code name eagle'' first ever battlefield multiplayer game with tanks, turets airplanes , cars ....it is great game https://en.wikipedia.org/wiki/Codename_Eagle

[edgy gate]

i think she could go against other bots from older games and it'd be fine

or bots from newer games too. but if multiplayer is not allowed, then single player is fine with me too...there is Unreal 1 after all, heh

[lime marlin]

well the fun would be playing against real people

[edgy gate]

if you say so, i just want to see how she'll handle herself on a game that is FPS in general. but i mostly want to see it for older ones.

[lime marlin]

neuro managed to stay in an osu private server, and from what i've heard from a dev stream, one of the among us devs is helping vedal by giving the game source code

among us would bring good views

maybe vedal can configure to make neuro mad, and huge collabs would just be perfect

playing with viewers would also be cool but sounds like a bad idea for a streamer

[edgy gate]

specially risky for a game that is about lying

[lime marlin]

Stardew Valley/The Sims 4 look like good games for neuro to play

specially the sims 4

many options for silly interactions and random events are perfect

neuro could do many "dark" interactions

[edgy gate]

she can finally fulfill her desires to kill people with Sims. there you go, the Basilisk finally uncoils

[sage turret]

Two streams? with Subtitles and without. I am sorry if it is impossible or stupid answer

[normal vale]

Twitch partners aren't allowed to simulcast, even cross-platform. So even if this would be allowed for non-partners, it wouldn't be sustainable since Vedal wants to become a twitch partner.
info source: https://help.twitch.tv/s/article/partner-exclusivity-faq

The only exception is if he were to stream one on instagram/tiktok, which are the only platforms allowed to be simulcasted with twitch. However I highly doubt Vedal would stream on either of those.

It seems Twitch may have built-in closed captioning that can be fed text live. I personally think this would be one of the most viewer-friendly solutions for subtitling in Japanese as well, however I don't know how that would conflict with the unity subtitles Vedal is currently using.
info source: https://help.twitch.tv/s/article/guide-to-closed-captions

[brazen lance]

suggestion: || HOO ||

[wanton fog]

I do not know how karaoke streams are currently done.
This suggestion is irrelevant if the songs are chosen in advance.
But if there is an element of randomness to the songs she sings that aren’t new, in the future, it might be a cool idea to have Neuro come up with 3 or 4 options and let chat vote on them in a poll between songs?

Again, this suggestion is only really relevant if Neuro is already the one picking the songs. Idk how the karaoke streams work. And automating twitch polls might also be problematic.

So depending on how feasible it is, it could be a good solution for more viewer interactivity. since I’ve seen a few people talking about missing that a bit with karaoke streams

But unless it’s really simple to implement something like that. It’s probably not worth committing valuable dev time to. It’s likely not worth the trouble since the majority of people already love the karaoke streams, including myself.

[fading bane]

Currently songs are chosen in advance based on the pool in #1060397294048002078 . Vedal is looking into the techs needed for live song requests, which is technically possible currently. However, song quality is a major issue with that tech, at least based on other AI streamers.

[proper sable]

Literally mod abuse

[lime marlin]

As i said, i could make the subtitles (screenshot the entire screen and a bit more height to fit the subtitles) in any language, the problem is

1. It would only work in discord
2. I would need direct access to the subtitles (doing OCR would capture the chat)
3. The captions would interfere with one another on twitch, which wouldnt be good

[lime marlin]

forgot to reply ^^

It would be a good idea tho, if this were to happen the japanese community would probably gather on the server

Another idea is translating the vods, but it would take too much work

The only options for JP people is to watch translations from clippers

This would all be easy if twitch wasnt a bitch

Well, there could be a twitch extension to change neuro's language

disable/enable subtitles

its pretty possible

Just looked thru Twitch API, its extremely possible to:

1. Play sound (different tts languages)
2. Add Subtitles

all with an extension

I'll try some stuff, if anything cool comes up ill dm Vedal about it

[earnest comet]

The reason humans are kinda sentient is because of survival, and status is something that matters in a society; so i wonder how hard it is to implement a training based on chat's reaction to what she says (with her being able to stand her ground of course).

[tulip fable]

The downside of using Twitch API is that the subtitle will not be present during the recording and clipping.

[earnest comet]

And also, upgrade her memories so that she wouldn't contradict herself btw~ currently waiting for the upgrade Vedal...

[lime marlin]

hopefully this can be changed

well if it doesnt work there could be a special tool for clippers to use, which add subtitles and audio of their choice to the twitch clip

[little juniper]

I think he is going to have two streaming PCs and create an account for each to stream?

potentially as a separate stream

if this were to happen the japanese community would probably gather on the server

#日本語自動英訳-jp-translate message

I don't know if they will join this server, I think it would be best to create a separate Japanese/knowing only server( more closed community) and then access the outside world, I will help you then.

[lime marlin]

also an option

Its up to vedal, although a good option for now would be using @rocky zenith but i'd have to get direct access to the subtitles

He already said he wants to do it, so i wont bother him too much about it

[little juniper]

not recommended to ”just” use the bot, its already join those who would be interested in joining the server, but for who are not, they dont want to enter a server with other lang, as a result, it doesn't help that much

i mean, for those who are interested in Neuro-sama but have not seen it cause it does not have Japanese subtitles, it cannot be an attractive proposition.

[lime marlin]

as i said, an option for now

vedal has his own idea, i cant do much about the jp side of the community

im trying to help with what i can since vedal is busy

yeah sure, he can create another channel and use other tts and subtitles

but we dont know when hes gonna do it

it can be tomorrow or in 2 months

[little juniper]

I missed this, sorry about that

option for now

I've talking about what should do in the future.

[lime marlin]

yes, in the future one of the best options would make a separate twitch channel

[shut sequoia]

Suggestion: Since Vedal wanted to stream in japanese before, perhaps it'd be a good idea to have a japanese channel and a japanese Airis/Neuro that exclusively speaks japanese. Could be a 1-1 copy of Neuro that simply speaks japanese, though it could be funny to do it like "Its not Neuro- But Neuros japanese cousin!"

[limber walrus]

If Vedal's interested, I made a mod for Placid Plastic Duck Simulator that allows you to name ducks and move between them remotely. Moving between ducks also provides an ID (single word description) of the duck currently viewed. The result is the ability for an LLM to name and comment on ducks autonomously while playing. Makes for a good background game. https://github.com/pladisdev/PPDS-Mods

[normal vale]

Even if making a new channel got around the twitch partner rules about it, wouldn't it be a huge burden on his computer to broadcast two streams at once?
And if he streamed at different times, I don't see Vedal taking kindly to having to make an entirely separate schedule just for Japanese streams. A separate channel at all could cause issues with the filter if neuro sees the chat there too.

I still think the best solution would still be some form of togglable subtitling for multiple languages. However Vedal might reject it since he seemed to be adamant about not giving access to a transcript (which seems a likely unintended consequence to external subtitles). This limitation could make any form of external subtitling nigh impossible.

The alternative would be simply putting the Japanese translation on screen alongside the normal subtitles, which is sub-optimal due to the space it would take up as well as the subtitles being limited to English and Japanese, but it's still an option worth some consideration.

[gritty cave]

I think it should mimic real life vtubers. Aka Neuro could learn Japanese and do some stuff in japanese sometimes, but for a general japanese stream that role should be for a dedicated japanese language ai character which can be different from Neuro. Maybe they can do a multilingual Collab or something, etc.

[normal vale]

Moderating what text she receives is as easy as finding Japanese moderators, but I think filtering neuro in a language Vedal doesn't know might be a bigger issue. Definitely doable if he finds the right people though.
He has mentioned the possibility of having multiple AI Vtubers, so I think that option is definitely in the cards.

[gritty cave]

Oh for japanese that's not a problem. As is well known it is impossible to say something offensive in that language as all japanese people are extremely polite.

...maybe my sarcasm is too dry for the internet ..

[normal vale]

is that even possible?

[gritty cave]

Sah...

[normal vale]

there are enough unintelligent people on the internet though that it's helpful to put "/s" at the end of a message like this, to clarify you don't genuinely think that.

there were conflicting voices in my head trying to figure out if you seriously though that

[gritty cave]

That's what I go for though... Heh 😉

[normal vale]

then expect dry factual answers often

from me at least...

I would answer that completely seriously

[gritty cave]

Anyway yeah, any serious japanese content really needs a serious japanese manager

Some kind of one off experiment would be fun to try though

[normal vale]

I wonder if that might be better to do on Youtube, I'm not sure how popular twitch actually is in Japan

[viral ruin]

Moved to #1095948590947500113

[narrow pagoda]

run off gpu instead of cpu

[covert mirage]

4Head

[narrow pagoda]

Neuro-cam in the corner, to see what she sees >.>

[quaint badge]

Make hitcircle one color, remove key overlay

[quaint badge]

Make #anteater-pics please

[narrow pagoda]

:anteater:

[burnt arch]

Try this reverse arrow @forest forge

[covert mirage]

https://www.youtube.com/watch?v=1_TatfHtFac

no fadeout skin and no numbers

should be good

https://mega.nz/#!SRA1VBIY!3dVA3KBS-HjVs9k18BPzzRAgLojJh9p2HMfeA-E3vCU

[narrow pagoda]

but followlines

[grave acorn]

just remove the followlines ez

[covert mirage]

removing followpoints is way easier than doing the no fadeout thing

[burnt arch]

That's true

[covert mirage]

aw the zip brings alot of skins so i dont know which one is it tho

[narrow pagoda]

nah I think the one we have at the moment is best.

[honest mason]

AI LEARNS TO PLAY TOHOU

[forest forge]

oo

neuro-sama could play many games

[grave acorn]

o gosh

[honest mason]

yasss

you'd probably have to keep separate configs for each game though

[covert mirage]

good morning

[forest forge]

well yes

morning

[acoustic acorn]

cant read spinners

rip

[urban fossil]

Its because auto goes to the fastest spinner possible instantly

[acoustic acorn]

and the fps bad

[urban fossil]

so its a bit weird for Neuro-sama

[acoustic acorn]

so it just goes up and down

[urban fossil]

yup

[acoustic acorn]

it could spin before

this is why we need nasa computers for this

[urban fossil]

^

@forest forge get on it

get a nasa computer pls

[acoustic acorn]

look up the specs of nasa computers

they are stupid

Total CPU cores: 245,536.

[urban fossil]

She might be able to run a tad quicker

[wide grotto]

GG @urban fossil, you just advanced to level 1!

[urban fossil]

woah

[acoustic acorn]

Pog

[weak parcel]

woah

levels

[acoustic acorn]

add Pog emote

[grave acorn]

might disable levels

depends on what you want

[acoustic acorn]

to twitch

owo

[forest forge]

if you want to donate a nasa computer to me ill take it

[acoustic acorn]

dont worry

[wide grotto]

GG @acoustic acorn, you just advanced to level 1!

[acoustic acorn]

its only 15million

[urban fossil]

@forest forge

https://www.marinescene.co.uk/product/2936/nasa-weatherman/?gclid=Cj0KCQjwtr_mBRDeARIsALfBZA7n9ngBhzsGBQe5PH_qc9eItCJ8eUC8Ii0B8bjNkyh2QkYK0W6CRVcaAm4JEALw_wcB

ez

[dark ivy]

@forest forge https://i.imgur.com/dSGTwUU.png This should be a good circle for the skin, it's just WWW's circle with the number erased.

[honest mason]

even though it has been done before, chess

[gusty rain]

Try in a beatmap which only have 1 note and the 5s long. Train untill he catch it.

[rugged sonnet]

give neuro-sama a better framerate maybe

[forest forge]

neuro-samas frame rate is hardware limited

[wide grotto]

GG @forest forge, you just advanced to level 1!

[quaint badge]

Remove section fail / pass maybe?

[winter shell]

had a thought, maybe when analysing frames of auto playing you could add the grayscale values of every pixel in a frame to get a "total whiteness", then only actually backpropogate and adjust weights and biases if there is a certain amount of "whiteness" on the screen

so it doesn't bother analysing frames where there are no circles

could potentially make it faster to train and also avoid useless training

[forest forge]

yeah that's a good point I was thinking of doing that

[honest mason]

Make it less confusing for ai: count circle appearance time against what ar the map is. Instead of approach circles. This could also aid im click training

Thpugh I haven't elaborated enough, seems like a good idea yes?

[forest forge]

what?

[honest mason]

No approach circles
Instead let ai know the ar and.how long the circless have been up.

[forest forge]

how would I let it know how long the circles have been up?

that defeats the purpose of it

[honest mason]

True.

[wide grotto]

GG @honest mason, you just advanced to level 1!

[forest forge]

does anyone have any suggestions for games neuro-sama can play after osu?

[honest mason]

What is the game max complexity?

[forest forge]

well nothing openai level

[grave acorn]

tag

and also dance dance revolution

[forest forge]

build a robot that plays dance dance revolution

[covert mirage]

osu!ctb

[forest forge]

I was thinking of doing mania

since like 3 people play ctb

[honest mason]

pacman?

[forest forge]

ooo interesting

I've already made pacman ai

could do it again but better though

[honest mason]

Tru

Tohou

[forest forge]

now you're talking

[honest mason]

Mount your friends

Uno maybe.

[covert mirage]

maybe something like super mario world 2: yoshi's island

[honest mason]

Super hexagon

Geometry dash

SMB

[acoustic acorn]

maybe improve the setup first

and then play minecraft

[grave acorn]

i want to play minecraft with neuro sama

[honest mason]

Halo wars (at a stretch maybe)

[acoustic acorn]

terraria

ok seriously

mania/taiko/ctb

[honest mason]

non arcade games would be hard to implement imo

With super hexagon. 2 output neurones

Mania, 4k = 3 output neurones 7k = 3 output neurones

[forest forge]

Minecraft with a simplified texture pack of solid colours sounds cool

[wide grotto]

GG @forest forge, you just advanced to level 2!

[forest forge]

could have a server as well

[honest mason]

how would you go about implementing it?

[forest forge]

so does super hexagon

[acoustic acorn]

i doubt she would do anything

super hexagon hard to read in black and white

too many colours

[honest mason]

For minecraft I'd suggest a server plugin

[forest forge]

could have colour though

[honest mason]

the walls are very discernable too

[acoustic acorn]

i forgot what it even looked like

[forest forge]

hmm

[honest mason]

the walls

[wide grotto]

GG @honest mason, you just advanced to level 2!

[honest mason]

Tend to.ve brighter

Than the background

So I mean

[acoustic acorn]

obvious thing is osu!mania or other game modes

like ctb would be cool

[forest forge]

yeah true

[forest forge]

well rn I'm going to add gpu implementation

[honest mason]

aye

[forest forge]

and try and stream it

[honest mason]

about time xd

[acoustic acorn]

pog

[forest forge]

might not work though

[acoustic acorn]

time to pass bigblack hdhr

[honest mason]

so it wouldn't be that hard

[acoustic acorn]

is there an auto

[honest mason]

to read

you don't need auto

[acoustic acorn]

or will it have reward/punishment like learning

[honest mason]

you can use a reward-punishment system for super hexagon

[acoustic acorn]

unlike current

[honest mason]

the timer can be used as a judge for how well it did

or amount of obstacles avoided

[acoustic acorn]

timer

[honest mason]

at first the ai would obv do nothing for the timer

since sometimes that leads to the best results

but after a while it would try to avoid the obstacles

[acoustic acorn]

if it moves it will learn moving = reward

[honest mason]

xd, that's not really how you'd award an ai.

there are times in super hexagon where you DON'T want to move

[acoustic acorn]

to start with

moving will save it from death

but then it will have to avoid

[wide grotto]

GG @acoustic acorn, you just advanced to level 2!

[acoustic acorn]

Pog

[honest mason]

you have a point

though sometimes the level can give you 20 seconds of living time when you stand still so

non arcade games are a lot harder to implement learning into

[acoustic acorn]

it will have to learn

[honest mason]

like, unless you design the game yourself and input checkpoints which reward the ai

then it won't always work

also training an ai from just having it watch can take a really long time and can be inefficient. there are better ways to teach it sometimes

[acoustic acorn]

neuro-sama can everything

[honest mason]

neuro-sama isnt

the most efficient algorithm in the world.

so some things could take millenia

[acoustic acorn]

owo

[honest mason]

and could prooobably not happen too

[acoustic acorn]

slow learner

[honest mason]

the problem with some ais that learn from watching, if you don't have a way of punishing it, then it could end up regressing.

as you've seen with neuro-sama going up to 50% accuracy

then down to 3%

[acoustic acorn]

it seems to be on short maps

[honest mason]

like

one thing can fuck up

and it will ruin progress

if you have something that stops

the fuck up

it can just ignore the new input and be unchanged

with genetic algorithms

you can implement this with copying over the best from the last generation onto the new generation

so even if the new generation fucks up badly

the best is still the best so the progress won't receed

[acoustic acorn]

evolution

[honest mason]

essentially yeah

https://www.youtube.com/watch?v=BOZfhUcNiqk

this video should explain genetic algorithms kind of okay

[acoustic acorn]

what if i have already watched it lol

[honest mason]

not surprised xd

[wide grotto]

GG @honest mason, you just advanced to level 3!

[honest mason]

code bullet is quite gut

it's only to be expected

[glad blade]

Have it read the results screen on its own plays and teach it the 100x/50x, this is how it was done for CTB except only misses

[acoustic acorn]

that may be a bit hard

[glad blade]

Its simple, just make the font readable for a image->text python library and teach it misses

[acoustic acorn]

plus its on rx

and its not on a reward/punish system

[glad blade]

Misses show on results screen don't they?

[acoustic acorn]

yes

[glad blade]

Thats what I am saying for a reward/punish system.

[acoustic acorn]

for that sure

she would try and play the result screen like it was osu

[glad blade]

This is what I used for CTB and worked fine

[obtuse hatch]

if possible add map requests

would be neat

[narrow pagoda]

^yes this

[rugged sonnet]

how tho

[obtuse hatch]

I was thinking something like mikuia, except not mikuia because she's playing on a private server.

[rugged sonnet]

also 1. we have alot of pepole to do that and 2. bet someone would troll and put on centipede

[obtuse hatch]

yeah I guess

[narrow pagoda]

Only ranked maps?

[rugged sonnet]

then someone would put on the big black

[narrow pagoda]

Only ranked maps that neuro-sama has never seen before

[obtuse hatch]

I don't feel like it would be an issue if someone put on big black

if requests were added after she's been trained a bit more

[honest mason]

there is actually a chance that it will come on

most of the maps that have been mapped are just 1-4*

[forest forge]

yup

[covert mirage]

Someone requested a Fps counter.

I kinda agree.

[obtuse hatch]

if you have room on your overlay somewhere a pp counter would be cool

[wide grotto]

GG @obtuse hatch, you just advanced to level 1!

[urban fossil]

Try a fullmod play REEE @forest forge

[obtuse hatch]

it can't do HD though, right?

[acoustic acorn]

more anteaters

[covert mirage]

[obtuse hatch]

dude what if you got one of those generic osu! overlays, but the background is just a bunch of anteaters

[covert mirage]

That would be interesting to look at .

[obtuse hatch]

If I had any clue how to make an overlay I would do it

[honest mason]

make it 1920x1080 or whatever the stream res is

then leave everything transparent

except where you don't wanna be transparent

[covert mirage]

Dont forget to have a wifu in the corner.

[honest mason]

^^

[covert mirage]

Animated wifu is optional.

[honest mason]

tru

I'd draw her if I could honestly

but I suck at drawing.

[covert mirage]

What would neuro-sama look like?

[honest mason]

hmm

2 options

computerised cyborg head

[covert mirage]

I keep thinking sofia the robot.

[honest mason]

or like

literally making it out of neurones

[covert mirage]

Hmmm...

One thing is for sure. She has to have a stuffed animal that is a Anteater.

[honest mason]

yup xd

it'd be pretty cute

[covert mirage]

Cyborg head. Then anteater chilling on top

[honest mason]

YESSS

[covert mirage]

My wife is ok at drawing. Im going to see if she is willing to draw something like that.

[wide grotto]

GG @covert mirage, you just advanced to level 3!

[covert mirage]

Fan art.

[honest mason]

yesss

50 hours old and already getting fanart xd

[covert mirage]

XD

[glad blade]

Like the fanart in the bottom right?

[brave trout]

a Neural Network chat might be a good idea where we can adress improvements to the AI.

[quaint badge]

Isn't that the point of this

[wide grotto]

GG @quaint badge, you just advanced to level 1!

[honest mason]

Ikr

[minor magnet]

Wouldn't it be possible for neuro sama to learn from reading perfect replay files rather than having to see each map be played in front of it and then process it?

I suppose it's quite a re-code but might speed up the process a bit

[rigid gate]

I think it'd be rather tedious to obtain a perfect replay file for a large number of maps, but that is a valid idea

Unless there's some hidden way apart from watching auto play the map and saving the replay afterwards

[minor magnet]

There's this guy on YouTube that made a new auto algorithm, so it can also 100% aspire maps and 2B maps

He made a modified client of the osu client

In the options there's a new option that allows you to make a replay without even having to first let it finish

[vast shard]

this is from osu!fx leak?

[minor magnet]

Ofcourse you can just select the normal osu auto algorithm as well. Ans then just let it make some replays for a bunch of maps

I haven't heard of osu!fx before

It might be in there as well?

[wide grotto]

GG @minor magnet, you just advanced to level 1!

[vast shard]

https://osufx.com/

[minor magnet]

I googled it

But I am not on a pc so I can't test it

https://github.com/YaLTeR/osu-auto-plus

This is the client I meant however

[vast shard]

"osu!auto+ is a rewrite", the osu!fx is a original stable osu modified after a leak

but ty for share it 😄

[minor magnet]

Osu auto+ is also from the 2016 leak source code though~

But yeah if fx has that feature too then you might as well use that

[pure sphinx]

add some sort of skip function to either discord or twitch chat to make neuro-sama skip a song if it's bugged or if F2 gives her taiko

[forest forge]

donations = map request?

[wide grotto]

GG @forest forge, you just advanced to level 4!

[covert mirage]

owo yes

[lean wing]

https://www.dropbox.com/s/ikxuhna0bjcvlai/Grayscalereadable.osk?dl=0
I think NN can recognize circles a lot easier with this skin

[minor magnet]

I can also make a very basic skin for it as well

I don't know how NN looks for circles but it's not hard to create your own "super readable for ai" skin

[stuck cairn]

Since you can't use the pc while the nn is training, woudn't running linux and windows in a kvm with the 980 passtrhough to it be useful so you can still do stuff in the host machine while the kvm has it's own mouse captured?

[forest forge]

I suppose, would need a bunch of changes though and could be hard to get cuda to work but maybe worth a try

[stuck cairn]

i can help setting up some of those stuff, im running one rn with an nvidia card, and cuda works fine from the pair of benchmarks i took

[forest forge]

oh interesting, I see

well I'm alright for now but if I need it I know who to contact

[obtuse hatch]

new cursor for you

@forest forge

I think the last one might work best

@forest forge ^^^

[obtuse hatch]

no

[stuck cairn]

bigger?

it's grey btw

looks white coz of contrast with discord

@forest forge

[golden osprey]

posting here cuz buried

in grey

[wide grotto]

GG @golden osprey, you just advanced to level 1!

[stuck cairn]

Make her use Score V2 and use that as a mesurement of how well she is doing?
Since V2 lowers score when missing and increases it when hitting, you could use that as feedback.

Improve the replay capturing by using more frames with ffmpeg so she has more data?

[obtuse hatch]

dosen't score v2 also fuck with slider leniency or something?

[stuck cairn]

if V2 doesn't work, maybe changing every hitsound to the same sound and use that as a way to know she hit the note

it would need sound implementation though, and you don't have that

V2 fucks with a lot of shit

and last time it didn't work at all with relax, it crashed the game (back when it was just implemented, so it should be fixed)

[glass stone]

yeah with sv2 you need to acc sliders too

[wide grotto]

GG @glass stone, you just advanced to level 1!

[glass stone]

not like it's an issue with rx

[stuck cairn]

well, she has relax, so doesn't matter much

there's an issue

she can't use it with relax

https://i.imgur.com/1nG7eDf.png

score is 0

always

doesn't increase with rx

[glass stone]

oh really

that's weird

[stuck cairn]

https://osu.ppy.sh/ss/13175319

yeah...

[glass stone]

in sv1 it just gives you 300/100/50 per hit coz 0x mult instead

[stuck cairn]

happens the same with ap

[wide grotto]

GG @stuck cairn, you just advanced to level 2!

[glass stone]

is osu!sync dead now

just reading acc from memory might suffice

[stuck cairn]

she doesn't have memory though

[glass stone]

osu's memory -_-

[stuck cairn]

oh

[glass stone]

like what the pp counters do

[stuck cairn]

use the pp counter source

the coutner doesn't read the memory afaik

reading memory is too hard for that

[glass stone]

pretty sure sync/rtpp did, idk about the stream companion one

[stuck cairn]

i think they use msn live status

[glass stone]

wtf

[stuck cairn]

or they could specate localy, but i have no idea how they work

[glass stone]

surprised that would have hit counts

[stuck cairn]

btw #suggestions

[glass stone]

oops

[stuck cairn]

https://discordapp.com/channels/574720535888396288/574725266346344470/575868477953277982 to go up the conv

[frank sun]

legacy oppai is python compatible

[covert mirage]

I am not a programmer and I have absolutely no idea how Neuro-sama works (other than neural network in a nutshell), but I was wondering if she could get assist with feedback by enabling hits (300/100/50/miss), but only skin 300 to a black/gray cross or star, while missing would show nothing

Would Neuro-sama go like "wait, I am not seeing the cross/star, should I readjust my aim" if that was the case?

[stuck cairn]

idding image recognission would make her a lot slower in training and playing, so the best way would be to add something outside the picture and procecss it with async

[forest forge]

working on it but don't have a reinforcement learning algorithm at the moment

[pure sphinx]

https://stats.stackexchange.com/questions/181/how-to-choose-the-number-of-hidden-layers-and-nodes-in-a-feedforward-neural-netw

[grave mural]

Make her make a toaster ding sound or some cute anime noise when she finishes processing a full autoplay, I really love the idea behind the project but it's annoying to tab in and out to see if she finished processing it so a noise would make it awesome

[atomic sorrel]

"I'm finished "

[gritty comet]

generation number?

[honest mason]

Eh?

This isn't a genetic algorithm

This is like the 4th or 5th iteration

But thay's because of vedal making changes @gritty comet

This is a different algorithm.

[plain agate]

Would it be better for her to learn from a replay that instantly moved to the next circle after the current one has been clicked? I feel like auto is causing the problem where she jiggles just outside the hit circle

[pure sphinx]

did you try applying changes to the hidden layers as I suggested before? noticed any difference?

I see you reduced the size of the network overall, and now your GPU is no longer loaded 100% (at least it isn't getting as hot)

[stuck cairn]

the temperature doesn't matter as long as it is not over 100C

[tight swift]

How about this to make the cursor invisible to the network (making the Neuro Cam images cleaner to the network, and the cursor is still visible for the people watching the stream):

Remove the green from the hitcircles (making them red, or magenta), for example 255, 0 , 0.

Make the cursor 50% transparent pure yellow, 255, 255, 0

That way, the cursor in the background will look like 122, 122, 0. Over a circle, it will look like like 255, 122, 0.

Then, take only the red channel, and reduce the image to a 2-bit color image (1 for red values above 150, 0 for for values below it, for example).

If you make the hitcircles and sliderball pure red, you could also make the slider borders magenta (255 0 255), and you feed the neural cam a 3-bit color image instead. (2 if the pixel has blue, otherwise 1 if red is above 150, otherwise 0).

you could color code different elements by varying the amount of blue that way.

[pure sphinx]

did some tests, you can't specify alpha, and the cursor overrides whatever is behind it

[tight swift]

did you try by changing the alpha in the png image?

[forest forge]

@pure sphinx I did apply some changes thanks to your suggestion, seems to be working better now

@plain agate you are correct but then the AI wouldn't be very entertaining to watch if it just snapped to circles, might do it as a last resort though

[plain agate]

maybe mix it in with the auto runs?

[forest forge]

@tight swift the cursor has very little impact on the network and is still needed for grayscale tracking when watching auto (to train from), but otherwise it's a good idea

@plain agate hmm interesting, might try it but it's not that easy to implement

[tight swift]

the tracking in auto could be done by watching the green channel alone (the cursor always has a value of 122 green)

[pure sphinx]

@tight swift checked and.. it's average

[wide grotto]

GG @pure sphinx, you just advanced to level 2!

[pure sphinx]

the cursor is the yellow thing

the red channel is pretty much untouched, green channel is average of 0 and 255

[tight swift]

that works too

[pure sphinx]

suggestion: remove the anti-ban ant-eater from the "playing" screen

wanna see those RPMs

remove or put it somewhere else :v

[stuck cairn]

calculate them with score over time

[pure sphinx]

there's no score tho 🤔

[stuck cairn]

the big numbers on the middle of the spinner

[covert mirage]

that is super roundabout and also assumes it actually will get points on spinners which it usually doesn't lol

[pure sphinx]

not all spinners are made equally :)

[covert mirage]

yeah diff spinner lengths and diff difficulty settings change that anyway right?

[pure sphinx]

yep

[crisp whale]

suggestion: have a channel dedicated to what changes you do to neuro-sama so people know what has changed

[plain agate]

Autopilot has similar movement that I was referring to in my earlier suggestion, could try it as a simple test to if it makes a difference over auto

[forest forge]

@plain agate I did try autopilot originally

[river silo]

If you take away the skip button, then I believe you can solve the offset problem. If you look at the plays which have a long intro, then you'll notice it having a hard time deciding where to go. That distance that the cursor is from the skip button is the same amount of offset that it has with the circles as it play's the song. If you make a macro to just spam space at the beginning, then I think it will be fixed

I believe it is treating the skip button like a circle, but the cursor is not there during auto so it has no idea what to do

[glass stone]

easy solution: replace skip with blank file

play-skip.png

[glass stone]

suggestion disable cursor capture in obs

[covert mirage]

Replace section pass, skip with 1x1 blank file

[obtuse hatch]

same with section fail for consistancy

[pure sphinx]

Suggestion to improve accuracy: during training don't give her the coords of the mouse for the current frame, but for the next frame

[forest forge]

interesting idea

[pure sphinx]

because she sees the current frame, but when she decides where to put the cursor, that frame is already gone

so there's a little delay she needs to account for

[opal canyon]

it would have to stay there for an extra frame then, wouldnt it?

[pure sphinx]

why?

btw, have you benchmarked to see how long it takes for a single neural network call ?

also, start training her taps as well

[opal canyon]

if it aims a frame early it might leave a frame early too

[pure sphinx]

not really

[covert mirage]

change sliderborder color to white

[rugged sonnet]

try to make her click

[azure ivy]

idk why it has been removed but please display an eta while training

[forest forge]

uses different training method now that doesn't let me output a training time directly but I'll try

[lucid owl]

how about pp counter?

[covert mirage]

Off ingame pm's

[forest forge]

idk how

[covert mirage]

in options

[azure ivy]

faq expand:

• why anteaters?
• why is neuro cam so low res/low fps
• why always hr?
• why is she shaking so much? (will be obsolete eventually)

[wide grotto]

GG @azure ivy, you just advanced to level 2!

[covert mirage]

made a new overlay for twitch.
@forest forge couldn't decide if he should use it so i should ask you guys what you think about it.

pls dont hate for no ant eaters

ps: i know it is very pinkish idk lol

[winter iron]

it's v nice and clean

[covert mirage]

thx ❤

[queen shard]

wait a sec

[wide grotto]

GG @queen shard, you just advanced to level 2!

[queen shard]

where's the Neuro-sama's face

[covert mirage]

her face?

wut

[queen shard]

:D

[covert mirage]

ohh

[covert mirage]

i dont have that image sooo

[queen shard]

just cut it lol

[covert mirage]

low res

[covert mirage]

fast boi

[queen shard]

add a skin link in desc :D

[covert mirage]

thank youuu

@queen shard i guess it's alright?

[queen shard]

yea

[gaunt token]

@covert mirage what's that font

[covert mirage]

devil breeze and futura extended

[gaunt token]

thanks

[pulsar sparrow]

so cute :0---

[covert mirage]

ikr