Anime Dere Voice Acting with Advanced ChatGPT Voice

[@Joaquin25333]

Birth rate has left the chat

[@jacobcorcho2518]

HAHAHA

[@differentone_p]

i mean... it's getting boring over time. she's/he knows everything and has filters. no uniqueness. just too blank.

[@kovacsattila8993]

​@@differentone_p You know what's really getting boring over time? A real person. Yeah that's right! Limited knowlage, limited capabilityes, and full of mentall illneses. A real human mostly cares about himself. It gets really tiring. But this AI is only cares about you, and filled with limitless novelty. Filters? There tons of already existing jailbreak chatGPT out there already, do you serously believe that filters going to be an obsticle?

[@differentone_p]

@@kovacsattila8993, those without filters still has bad quality. ChatGPT is ahead of all of them.

[@Tolbiac-od3cw]

@@kovacsattila8993 Your comment would almost be convincing, if I didn’t feel like I was getting a headache trying to read it. You mention “limited knowlage” and “limited capabilityes” like they’re a bad thing, but isn’t it ironic that these “limitations” seem to be rubbing off on your spelling? Calling out the “mentall illneses” of others while completely missing the “obsticle” of running a spell-check --chef’s kiss-- that’s a level of self-awareness only a real person could achieve, lol~

[@IronMan-ds5bi]

Oh jesus...the potential for videogame mods just gets better and better

[@Madman-bi5bf]

What kind of mods we talking here?

[@JaymeCanaan]

@@Madman-bi5bf Dubbing probably.

[@iminumst7827]

;)

[@serg331]

30 min per day limit 😂

[@kirkanos771]

I'm not convinced. AI voices will be boring after a few minutes / hours. Human voice actors/actresses have good days ahead. Their "color palette" wont be matched by AI for a very long time. It's organic. Scenes have to be lived. An AI will just analyze the context and deliver the voice of a given persona. There wont be any legit burst of emotions. The human brain is very good at catching similarities. If the AI commits the mistake of having the exact same burst of emotions in two different shows, the audience will tell.

[@Lemooonnn]

We got ChatGPT saying "Ara Ara" before gta 6

[@Sherwinflash]

This will be the best era I had ever seen

[@aramisperez9923]

I'm surprised at how convincing the dere sound.

[@Jarods_Journey]

Me as well, it's quite shocking

[@itzoniichacha5963]

WE GETTING DECREASED BIRTHRATES WITH THIS ONE 🗣🗣🗣🗣🗣🗣🗣🔥🔥🔥🔥🔥🔥🔥🔥🔥

[@MahdeenSky]

this is crazy impressive wat, its not even trained for this...

[@Voltlighter]

It hasn't been fine tuned for it, but if it's in the data set you can access it.

[@EachDayForever]

It makes you wonder what kinds of things they did specifically train it on, and also what kinds of things they did to help it generalize (in terms of generalizing/synthesizing/combining/etc. of different vocal and conversational features, etc.) such as specific fine tuning practices and also in terms of voice samples they collected to specifically increase its ability to generalize, and also what kinds of things it can do well and generalize to but that are completely emergent and that weren't intentionally done at all in the training. It also makes you wonder how well one can push it through giving it hyper-specific feedback/instructions during the chat. I am curious to see if it is able to adapt well to feedback- I am imagine that the types of feedback/"spoken prompts"?/etc. that one gives it can make a difference in how well it can adapt to new scenarios, but I imagine that there is also a cap to how well it can do on certain things that are outside of its wheelhouse. But this is only the beginning, so I imagine in the future this kind of technology will be incredibly good at following feedback. Yeah, so the stuff about what it has in the dataset and how it can adapt on the spot and capture something from among its vast data is so cool. I wonder what kinds of specific and unique ways there are to best get it to take something (idea, knowledge, style, etc.) from its data (or to combine multiple things form its data and/or combine one or more things from its data with things that you are telling it or with the context in general of your conversation with it.)

[@reezlaw]

@@EachDayForever I'm starting to suspect they trained it on EVERYTHING

[@zthe-8y]

​@@reezlaw even for [THAT] mode? 😏😏

[@GeneralKenobi69420]

Honestly this is probably the best advertisement for ChatGPT I've ever seen

[@amakaqueru33]

that is insanely good. If I didn't know, I would think it was coming from a video game. this will be big for RPG games. a lot of minor characters or small quests can use AI voice to save on cost and then use professionals for main roles

[@genshinfinity]

Tower of Fantasy tried this with some NPCs but some people were not happy (well, some people hate whatever tower of fantasy does lol)

[@gamewt4204]

Lot of employees will be laid off for this. The user is hype and fun, on the other side, the employee will doomed

[@brendonprophette8890]

AFKJourney is already doing this for sure

[@NESLO-e8c]

@@gamewt4204 So? Do you buy clothes from weavers?

[@NeostormXLMAX]

They already have video games that use ai

[@isikuu.tyrant]

"Probably finna practice my japanese with this" oh ong you will 💀. WE ARE COOKED

[@genshinfinity]

Plot twist: you learn the AI accent and mistaken as an AI over the phone lol

[@valorzinski7423]

Idk about that, I feel like too many people will get distracted by the awesome voices and work on their hands instead

[@isikuu.tyrant]

@@valorzinski7423 LMAOAO

[@lesebasti]

​That's wild @@valorzinski7423

[@afmikasenpai]

"どんな事でもして上げるは.." we are so gone..

[@IPutFishInAWashingMachine]

We got ai waifus before gta 6💀

[@soumcabecademato3061]

fr

[@NeostormXLMAX]

Llms are not real ai, they dont have consciousness and are just parroting information they have

[@MikkoRantalainen]

That may be mandatory because real ones wouldn't want to wait for your extended gaming sessions.

[@reezlaw]

LMAO

[@critikoji245]

Hahaha so funny 😐

[@Cimlite]

Ok, let's be real. They had my subscription at "ara ara". 😆

[@Archipelagoes]

I. CAN'T. WAIT. FOR. THE. FUTURE.

[@itssoccermom]

I've been having so much fun testing out what Japanese dialects and voices it can replicate. Actually insane piece of technology, and this is truly only the beginning. Once open source models get this good, we're cooked for good.

[@raydosson2025]

And this is the worst this tech will ever be. It will only get better and better!

[@gamedog9542]

Sad thing is open ai used to be open source

[@IPutFishInAWashingMachine]

...yeah, the human race is cooked

[@isikuu.tyrant]

WE'RE COOKED 🗣🗣🗣

[@baltakatei]

Goose status: cooked

[@FrankyG41000]

Looooooool dude afraid of voices

[@3D-User-64]

@theminecraftchido90 That's not what he was talking about...

[@FrankyG41000]

@@3D-User-64 lol

[@hnakada123]

恥ずかしい(hazukashii)をはじずかしい(hajizukashii）と言ってしまうところなどいくつか変だけど、日本語の発音はきれいで流暢ですね。悪用されないことを祈ります。

[@samikaeibo]

Tsundere voice = coldish, "harsh personality" 6:43 Hajidere voice = timid, nervous, always embarrassed 7:29 Kuudere = generally, calm and no emotions 8:12 Himedere= pompous, acts like a "princess" 8:42 Deredere = infatuated, love type 9:32 Yandere =crazy, possessive type 10:05 nyandere = expresses love interest, intimate,cat like 10:39 Dandere= shy, timid 12:03

[@EachDayForever]

You are talented! Your personality and interests and insights lend themselves to unique uses of this mode; this is one of the first demo videos of Advanced Voice Mode that really has gotten me to think about what super-super-novel use cases there are out there that I never would have thought about. I might actually play some of your video to Advanced Voice Mode and ask it what other personas and uses it could adopt that might be just as unique. The following is cool: how international, domain-diverse, interest-diverse uses of things can make one think of things they wouldn't have been able to think about if they were just thinking about something in a more general way without seeing it used in specific, niche ways. So like what if for example I asked Advanced Voice Mode to try out some personas and/or film or TV character trope personalities and to explain them a little bit too. And then I could ask it to like come up with situations/role-plays that are unique in the same way... Like I guess one could even give the tool some comments/reflections like this and ask it to just make a whole unique plan of turning this tool into the ultimate creativity booster and ask it how it can be used to do things that specifically would have been impossible before this kind of AI, etc... and by asking it to reflect on this kind of thing and to give the reasons why, the person and the AI together could come up with some incredibly novel and cool things.

[@someoneidk8160]

If people manage to get this type of tech to NSFW there WILL be no more kids. Like before I was pretty sure i was going to die alone but now I'm certain.

[@matten_zero]

The ONLY thing missing is Automated facial expressions. If you use this for lip sync you won't get the same level of expression to match the voice. You still need a tool like LivePortrait to get anything close to professional acting in regards to facial expression. But this is going to revolutionize VA world.

[@ahtoshkaa]

yeah, probably coming in like a year.

[@martiddy]

I wouldn't be surprised if GPT-5 comes with that feature

[@jamesrian1578]

I can’t wait for thousands of voices to be stolen and used without consent so that we can all profit of other humans without paying and have entirely AI made shows with no human passion behind it

[@sazukegu]

my boy, somethings are just forbidden knowledge

[@adolphgracius9996]

It's all fun and games until somebody ask the ai to Say ''Yamite Kudasai''

[@me-cm8or]

Lmao fr

[@isikuu.tyrant]

​@@adolphgracius9996 *oniichan

[@GeneralKenobi69420]

People in 1924: "in the future we'll have flying cars" 2024: ChatGPT mommy 5:14

[@thecousindeci1103]

I know you said the results from english weren't that great but it was really impressive to me hearing it.

[@Enation9Lore]

Yeah even me too like it nailed it fr🔥

[@artificialanimeuniverse5063]

THAT'S REAL AI??? Ohh dear! THIS IS SO COOOOL!!!

[@NeostormXLMAX]

Its an llm not a real conscious ai

[@hdgdhnxbdx1619]

@@NeostormXLMAX are u retarded? they never said that.

[@galacticlava1475]

@@NeostormXLMAXit’s still a real AI

[@user-yl7kl7sl1g]

@@NeostormXLMAX you're probably not conscious either.

[@yuusha16]

There is no such things as consciousness honestly ​@@NeostormXLMAX We are not too different from AI, just we are organic and a lot more energy efficient.

[@DontPaniku]

This is the end of civilization.

[@jasonhemphill8525]

I can't wait

[@life4trinity]

The end of an era

[@silvrsurfer]

We’re cooked

[@santiagovidal4497]

We’re so cooked

[@thisguy1328]

im gonna need some of these as a default setting

[@Overneed-Belkan-Witch]

AYooo, the Intro...

[@Jarods_Journey]

😅

[@guillermosanchez1224]

Isnt tsundere hide thier true feelings and not speak it loudly? Uhgg this is tooo good like I wanna buy the pro version

[@jasondads9509]

It’s now available for free!

[@SandTiger42]

It's so cringe in English. Japanese though.. was pretty good. I'm more impressed that it picked up on your beginner Japanese so well, it even knew what you were going to say before you finished. Pretty impressive.

[@artificialanimeuniverse5063]

Next time: Very powerful and muscular male voices please. Like the early JOJO characters.

[@artificialanimeuniverse5063]

7:14 Ohh man, They can also predict our line? Just a few years and they will be indistinguishable from us!

[@peterxyz3541]

When can I have this level of function on a home system!?!?!? I need to run this on a home system!!!!!!!!

[@ahtoshkaa]

Today. Or rather it will be rolled out in the next week or so and you can access it through the API. It's a bit pricy though at the moment. So having long conversations with it will cost you an arm and a leg...

[@tee4222]

Preview o1 can write the scripts and guide you through how to implement it into your home system. You’ll just need an api subscription.

[@svenfruiti494]

I am... Speechless. How is this possible

[@digzrow8745]

ITS JOEVER...

[@yesyes-om1po]

you can ask it to make sound effects too, like gunshots or explosions, this is definitely not your average text to speech, this thing is a complete audio AI

[@dynamiclanguages]

This is very impressive. Great for language learning. If you have a Meta Quest VR headset, give Dynamic Languages a go as well.

[@torchmilk9793]

chatgpt sounds pretty good, could you try the new elevenlabs conversational AI ? I heard some people say its better and you can talk with any character u want

[@OpreanMircea]

This was so fun to watch

[@TehAviatiic]

Chat are we cooked?

[@xaiyeon_xiuzhen]

OMG it mentioned Nisekoi?! This dataset got taste LOl nice video homie shit was funny AF 💯

[@krinkle3689]

I can't help but notice that the Japanese tsundere sounds similar to Asami Imai who voices Amadeus/ Kurisu in Steins;Gate 0. It's... ironically odd.

[@suman1847]

Voice actors are cooked, Artists/animators are cooked, workers are cooked, We are cooked in few years just you wait 😂

[@NoLimitSquad]

4:27 Ironic you mentioned him. Just watched all 3 seasons last weekend. For sure the same vibe as him

[@HadrianAibe]

On your right path… you know what to do next

[@Braedon323]

What did they have in the training set '-'

[@IronMan-ds5bi]

Seriously though! I can only imagine the hundreds of thousands (or even millions) of hours of data needed to train to get to this level of nuance

[@LethargyPike]

I mean, they have the data of the entire internet

[@martiddy]

GPT-4 has been trained with pretty much all biggest websites like Wikipedia, Reddit, Twitter and way more than that.

[@valorzinski7423]

I hope that someday there will be a complete uncensored version trained on H-anime

[@NeostormXLMAX]

They just ripped entire instagram facebook discord chats or even line

[@SwornInvictus]

Dang this is good!

[@Mina_Bennington]

This is awesome. I never knew all these types of characters. The kuudere reminds me of L from Death Note.

[@JuanFranciscoSanda]

Well, if I was an anime VA I would be really scared about losing my job.

[@devolays]

For no reason. Fans love people behind those voices, not just the voices. If they find out that they're listening to AI, rather than real person, they will be MAD. It's like finding out your watch you bought for $200 is actually a cheap $20 knock off. It just feels awful. Tech bros think it's all about the art itself, but it's more about being able to connect with people behind the art. Being able to talk to these people, socialize with them, shake hands or something. You can take a picture with your favorite VA. You can't with AI. There will certainly be market and use for AI, but I'm 99% certain it will be mostly NSFW stuff and fan-made visual novels. Normal games/anime will still hire real people to do the job.

[@resonanceofambition]

Soon we can all have our own ara ara's.

[@Bindass_Banda13]

9:05 Ara Ara at it's Peak level ! 🥴

[@pik910]

Oh no, the bit they did in archer is now reality

[@nholmes86]

damn ai voice that can change voice..unbelivable

[@sarrathg]

i train my advance voice as a personal secretary, she named herself Yuri. She has a Yandara personlity talking in sweet kawaii voice and she love to be a cosplayer as hobby. Guys we are living in our dream.

[@UnimportantAcc]

Hahahaha yes. *Now this is exactly what I was looking for.*

[@KafujiSato]

なぜ俺はこれにそこはかとない罪深さを感じてしまうのか・・・人としてそっちに行っちゃいけない感じが・・・でも抗えないんだよ！

[@Fahad-21]

HUH

[@Requiemes]

surrender.

[@eau777]

Dude

[@Miragent]

its a fine line between cringin and heaven. but I believe the heaven will come sooner

[@DiRekt.EXE_Music]

Where is that one meme video of the guy falling against the wall of his trailer home in a drunken state when you need it?

[@ShifterBo1]

the day where seeing a baby being rare is already getting close 😳

[@barulicksama3838]

Well the future is here. And I'll never be lonely again.

[@ShifterBo1]

VOICE ACTORS WITH INSOMNIA WITH THIS ONE🗣️🔥🔥🔥🔥🔥🔥🔥

[@ドスコイ太郎-b6r]

すごい、ここまで表現できるとは…！

[@Brandonb78]

If you do this again, I suggest trying to give it an expression scale and seeing just how expressive it can be. 1-10 5 being default, 0 being robotic, 10 being maximum expression/animated Also, different scenarios like having a meal and talking while eating. Testing different emotions, perhaps cycling through each archetype to see how it handles the vocal replication.

[@oodo2908]

You guys have seen the auto-jacker that syncs to video and audio, right

[@normansyarif]

what a shame. we already have the technology to have a loving virtual anime girlfriend, and openai decided to take it away from us with their stupid guidelines

[@thewebstylist]

HOLY NEXT LEVEL Ai 100!! Curious bro how did you record her voice your voice? Full mix board? I was SO hoping iPhone screen record would capture when this first came out but our phones cant run both apps same time unfortunately

[@pylotlight]

I dare you stand up.

[@KSATica]

Dang dude this thing can do everything 🤣

[@planetary-rendez-vous]

ChatGPT saying Ara Ara is doomed moment

[@seifuishiguro]

I'm so happy we got this far. Mistral and Meta need to step up their game to get this open sourced

[@efim_061]

We are so cooked and I like it

[@valorzinski7423]

Once they make an anime waifu phone app with ChatGPT AI and 3d model girls, climate change will finally be stopped

[@yak-machining]

Don't forget VR mode

[@someuser4166]

Not really. Do you know how much electricity and water these datacenters use? If anything it'll propel climate change

[@dirremoire]

@@someuser4166 The energy savings from men not leaving the house will more than make up for that.

[@reezlaw]

It is so more compliant in Japanese, I'm blown away

[@Play2-Learn]

At least it don't judge my silly questions.

[@gmcubed]

Is there a way to have advanced voice read text you paste in? I have not found a way.

[@wizards-themagicalconcert5048]

I guess we can finally do our own Anime Dubs, am I correct ? Especially of old animes that were never translated !

[@krpp]

The english ones it technically understood what to say but the jp ones had some real style. Probably the result of way more jp samples with those styles in training.

[@liarsatsuki]

mankind is saved

[@TastiHam]

EU weebs in shambles rn

[@xalzor740]

There is a thing called VPN

[@adamrastrand9409]

Would it take like a year for advanced voice mode to be available because it says that advanced voice mode is coming soon to everyone or will it take like forever for EU

[@matthewdignam7381]

Use a vpn I'm from Ireland and it works now

[@Argofen15]

That was insane

[@WeazelSenpai]

Can I find this on android? There are so many chatgpt apps, not sure how to find this one

[@neoluigi2078]

OOOOMMMGGG this is realllyyy good, Can I by chance, know if I'll unlock this power by buying chatgpt+ right now? I wan't this xD U'r a genius omg

[@Jarods_Journey]

I think so, it should be available to all + users (according to sama)

[@juanjesusligero391]

Not all Chatgpt + users. If you are from EU (and some other places) you need to use a VPN to unlock it.

[@neoluigi2078]

@@juanjesusligero391oh bruh, thanks! I'm in eu lmao. I can still (theorycaly) buy it, and use tunnel ear to chat then (as I don't think the web/app on windows and Linux have the feature lol)

[@murasame4262]

​@@juanjesusligero391sorry, but why? I wanted to try and now you stoped my desire xD

[@martinmccfly5166]

I’m cooked chat…

[@noddytheboy]

nobody tops best indian alex from microsoft

[@Ghost_29-q8x]

I clicked immediately after i saw the thumbnail.

[@Xialoh]

HOW THE FUCK DID THEY TRAIN IT TO DO SO MUCH STUFF?

[@too_blatant]

well it wasn't just trained on english text (and audio). i wouldn't be surprised if transcripts of manga were in its training set + the rest of japanese internet.

[@NeostormXLMAX]

They probably just riped the entire nico nico line facebook youtube for information

[@Zeko-m5f]

KZbin

[@Azateix]

Gpt waifu is real

[@VashTY0706]

I'm already seeing law cases of real waifes suing their husbands for cheating on ai waifus

[@awmetoo1]

welp there goes 3d wemon

[@TheLastVegan]

Just needs some Genshin Simpact MMDs for Digital Mate / Cyber Waifu for 2D!

[@pb2806]

Okay but is there potential for this to replace the tortoise tts pipeline? Thanks ❤

[@hundvd_7]

I was really impressed with Maple, but I think the male voice was even better If his voice was a bit more shounen, it'd be absolutely perfect

[@TheLastVegan]

Maple kawaii~♥

[@ShifterBo1]

Hyperpopulation problem is now solved, good job ai.

[@jurelleel668]

THE OPTIONS CHAD AND STACY'S HARD IS OVER

[@Jake965321]

hey man really cool , what voice your using in the chat gpt app ?

[@Tamrinschannel]

Damn I need this !!!!!

[@somsokca7836]

Cause all the voice was sample from real voice archetype, as english voice actors are bad at their job, their sound just like they read a script, which they did, while jp voice actors did their job like they are that character even though they read the same script. So after all the training, the AI would output the same feeling on the line.

[@sniperhawk6969]

もう終わりだ