NVIDIA ACE for Games Sparks Life Into Virtual Characters With Generative AI
Vložit
- čas přidán 27. 05. 2023
- NVIDIA ACE for Games is a new foundry for intelligent in-game characters powered by generative AI. Developers of middleware, tools, and games can use NVIDIA ACE for Games to build and deploy customized speech, conversation, and animation AI models in their software and games.
Watch our NVIDIA Kairos demo, showcasing how NVIDIA partnered with Convai to help optimize and integrate ACE for Games modules into an immersive and dynamic interaction with a non playable character named Jin. The demo is also enhanced with ray tracing and performance multiplying NVIDIA DLSS 3.
Learn More: www.nvidia.com/en-us/geforce/...
Subscribe to @NVIDIAGeForce for more PC gaming content featuring advanced graphics, AI, and high-performance technology.
Subscribe to @NVIDIADeveloper to see more cutting-edge PC gaming development software and tools.
GeForce Twitter: / nvidiageforce
GeForce Instagram: / nvidiageforce
GeForce TikTok: / nvidiageforce
GeForce Facebook: / nvidiageforce
GeForce Community Portal: www.nvidia.com/en-us/geforce/...
#nvidia
#Omniverse
#RTXOn - Hry
This is the widest aspect ratio I’ve ever seen for a video, it’s so wide it just blows past the point of being cinematic
it looks much better on a 21:9 monitor, but yeah for a conventional 16:9 monitor its a really weird aspect ratio
Edit: Guys I get it, it’s 32:9
@@Flame1it looks very strange on a phone in vertical mode lol. Still looks cool tho
@@brohouse7882 oh jeez, I bet it looks awful on a phone lol
It's a 32:9. It looks best on 32:9 monitors like Samsung Odyssey G9.
I thought my monitor was broken.
Feels like an oblivion npc interaction lol
Stop Right There, Criminal Scum! has more emotion than this
@@GiuseppeNelva Peppino malding 😭 I would be angry all the time too if I were named Giuseppe.
It doesn't even say hello when the guy walks in, just stares at him. Not impressed.
@@b0bl00i In order to access kindness emotions in your AIs you must pay an additional fee.
@@KKahnwald Giuseppe is an awesome name dude wtf
Interesting showcase for sure. Jin definitely sounds a bit unnatural, but I can easily see how these technologies could change the gaming industry.
Got to have that perfect voice for marketing you know instead of just being natural.
This sounds worse than the skyrim mod.
@@southcoastinventors6583 then you’ve never played Skyrim and are just insulting this just because you don’t like it.
@@thefatbob3710 I am talking about the ChatGPT Skyrim mod and it sounds better than this demo by NVIDIA which it should not since NVIDIA is running many of farms used to train LLM.
@@southcoastinventors6583 the ai can be inconsistent at times so idk what you’re talking about there both like this.
It's hard to exhibit the AI generated element from this alone. This is a sample of every recorded conversation we skip, skip, skip in any given video game. I'm sure the AI generative side allows us to return to that bar and order food, have a unique conversation or take a different, radiant quest. Just hard to express that in a 2 minute UE5 showcase.
This is a half bake demo, who the hell gets a mission to destroy an underground by just asking like that. Id like if this was a more relationship building conversation before he actually gives you the mission. You could talk about [world lore] before he would feel opening up to you and giving you a mission , in which i would ask him "whats in for me ? " and the ai would find a nice dialog to satisfy my gamer reward brain
@@staberas It will be utilized in a much better way by developers with an actual idea for a story and dialogue im guessing
@@amcfluff1547 one hopes. But goodness, they thought this was a worthy "showcase"???
99% of the player base will ask silly questions.
copium
I think eventually we get full non-scripted NPC using generative-AI dialog. kinda like procedural generation. So each time you speak to a NPC, you will get a different conversation entirely.
Skyrim already got that thru a a few mods. So yes, that's where things will be going.
Yeah imagine being able to ask the NPC in more detail what I was supposed to do on my quest or something like that, sometimes you wonder around for too long in open world games and just forget what you were supposed to do haha.
That already exists, it isn't implemented into any released game yet though
Thats what this is, inworld also does this.
We already have that, it's happening now
Games are going to get so crazy so fast, when they can make AI NPC's have better human like expressions it will be beyond amazing
never did before, neither they will now
@@Blaz1n yes. Because people haven't been able to do something in the past they will never be able to do it in the future. By that logic we should still be stuck living in cages and hunting for food using sticks.
@@Blaz1n bro is talking like people who said that we never going be able to fly
@@archdemonplay6904 Are we able to fly? ... I still use airplanes , I did not create wings in my body
@@Blaz1n That's true. Also, we will never see real time reflections in games 😢
The Ramen shop looks amazing, but the interaction, while cool in its conception, was not very impressive in its execution. IMHO
Yep honestly sounded like a normal npc dialog
The underlying technology sounds great, and being able to give these characters backstories and history that inform their AI interactions is super cool. That said, this is a terrible demonstration of that tech. Aside from the quality of the rendering, it's indistinguishable from an NPC interaction in a 90s CD-ROM game. I implore you to please use storytellers to create scenarios for demos like this, to breath some life into the character, and show them doing a little smalltalk interaction before assigning a side quest. That's the whole point us using AI for characters in the first place. To make them feel and react like humans.
Exactly!
yes, small talk! or even say some out of pocket stuff like, like giving Jin a light insult, like a friendly little jab or something.
By the way this is just a showcase, things will change once this gets implemented
Thanks so much for this, as an actor...a game actor...every word resonates. Seeing that matter is meaningful on many levels.
Except, no human voice actor was involved... realtime dynamic conversation!
This showcases conversational AI in the same way your Redfall and Gollum ads showcased graphics.
I don't hear or see emotions... But if they answer with that speed I'd be quite happy
Oh trust me, there’s already AI that have fixed this easily. This is mostly just demonstrate potential. Just look at AI presidents dating things. Plenty of inflection and vocal.
This seems a lot less emotional than some other AI voices I have heard online.
@@Savitarax the AI presidents videos that have been flooding on CZcams have more emotion and humor compared to this haha..
Imagine quests being assigned on the fly by underlying AI instead of getting scripted ones.
Exactly what I was thinking!
It will soon be pointless if not implemented correctly.
I don't think it'd be fully unscripted, but I think it'd be good enough to have scripted quests and assisted with AI.
Not only that. Immagine just letting the world run on its own and it evolving without you. Soon ai games are going to be like real life. Civilisations rising and falling in matter of minutes when using skip feature and then coming back into the world to see how it changed. If what I say is true i cannot wait for the future gaming
The AI is inconsistent, letting it create content will end up inventing something that doesn't fit the story of the game.
Many individuals critiquing this might not fully grasp the significance of what's unfolding here. It's monumental! This represents the future of dialogue-driven AI, taking charge of every non-playable character within a gaming universe.
I often find myself being critical of the content shared on this platform, but this particular video compelled me to reply. At present, our gaming worlds are populated with hundreds, sometimes thousands, of NPCs spewing out predetermined dialogues, serving little to no purpose beyond their scripted roles. Let's consider GTA V - an incredible game but held back by the lack of substantial interaction with NPCs beyond a couple of repetitive lines. Not immersive at all.
Now, envision a scenario where every seemingly inconsequential NPC has a unique backstory, a set role, a distinct purpose. What if you could engage in meaningful conversations about the game world with any character you encounter? The scale of impact this could have is astounding.
Consider exploring the world of Cyberpunk 2077, where every interaction doesn't merely fall into the confines of a few prewritten dialogues. Instead, you could engage in diverse discussions about the intricacies of the world around you. This could revolutionize immersion in gaming like never before. I am excited!
It's relatively effortless to dismiss the initial versions of technology. Take the case of RTX, and see the remarkable journey it's undergone from its infancy. Remember the debut of DLSS? Today, it's universally used by anyone who can access it.
Allow the technology to mature and witness its potential unfold. It's always easy to hate online to get likes.
The significance of what's happening here is quite clear. No one who is on the fence was convinced, and the faithfuls and astroturfers are bending over twice to defend this ultra-mediocre showing. As expected.
It sounds almost like we're talking about NFTs and crypto. The arguments are pretty much the same.
Wat you describing is a lot of meaningless fluff adding bloat to a game with very little value. Cyberpunk 2077 is great as it is, with dialogue written by *humans* for *humans*
"What if you could engage in meaningful conversations about the game world with any character you encounter?"
I'd get bored half way through the first conversation and start skipping the dialogue entirely.
I play games to be entertained, not to be bored by every NPCs sob story.
@@GiuseppeNelva How does this have anything to do with NFT and Crypto? What are you talking about?
It's evident that you may not fully grasp the repercussions of this.
You probably also criticized RTX or DLSS when they were introduced.
As we converse, I believe game developers are exploring ways to incorporate this into their projects.
In a decade, this could very well be standard - numerous NPCs, each with engaging narratives relevant to the game world, helping the character advance in the story.
You appear to be an online troll thriving on garnering attention with provocative remarks.
It's typically straightforward to garner approval from hateful statements, we're on the internet after all.
It's fairly apparent that your understanding is lacking and you seem consumed by hostility.
It's quite hilarious to see this level of anger towards a promising new technological advancement.
The funniest part is that in 5-10 years you will be playing said games and you will like them.
@@cormoran2303 Dude or you can just not talk? it's optional to talk to NPC with AI.. ever thought of that? i'm pretty sure majority of haters here are just making up excuses to hate on AI
@@allen57 I’m pretty sure you can say the same with the functionality. It’s optional. Its also useless in a bigger picture. AI is great and all but this use case is pretty much useless. I mean unless we get Roy the game from Rick and Morty lmao
Correct me if I’m wrong, but I think the point of this is to show that developers could basically use AI to create entire scenes, characters and dialogue with minimal input.
Sounds about right. But tbh, humans haven’t been very creative lately now have they? We’ve been getting some of the worst triple a games we’ve ever seen. I think I’m ready to see if AI can do better. And if it can….f*** it
No, only npc. This doesn't work with cutscenes you need actors in motion capture for that. This would be lazy work or low budget stuff, it would look terrible (aside from the graphics)
Many do not fully grasp the significance of what's shown here. It's actually huge
This represents the future of dialogue-driven AI, taking charge of every non-playable character within a game.
Game worlds are populated with thousands of NPCs spewing out predetermined dialogues, serving little to no purpose beyond their scripted roles.
Let's consider GTA V - an incredible game but held back by the lack of substantial interaction with NPCs beyond a couple of repetitive lines. Not immersive at all.
The interaction is broken every time you interact with any NPC in the world.
Only the few main ones feel realistic because they are all heavily, heavily scripted.
Now, envision a scenario where every seemingly inconsequential NPC has a unique backstory, a set role, a distinct purpose.
What if you could engage in meaningful conversations about the game world with any character you encounter?
I am excited to see how game-devs will take this tech and use it.
I think in 10 years this will be the norm.
Also I don’t think people realise that humans will still be needed, maybe even more than before, to make sure that every character has a compelling, realistic story that works within the rest of the game universe.
From how I understand it, it's more showcasing "autonomous" AI NPC. The purpose is to have AI take control of NPCs by giving them a form of roleplay. "You are Kai, a Ramen Shop owner in a crime ridden district of city X" and AI builds the character around these instructions and engages in custom dialogue it makes up as the game goes on.
Think of it like this: instead of NPCs running on pre-written scripts with voice actors recording a few hundred lines, it will run on a neural network and use speech synthesis to produce near infinite possible results when interacting with the NPC.
If that tech gets to the right point, you could basically just invent human backstories (or even have AI create those in the thousands without much input) and then give each NPC in your game a backstory, put an AI and tell it "here, that's you, go, live."
Basically an attempt at simulating real humans via AI in a game environment. Still sounds like science fiction to me.
@@SolidBoss7 The AI is not going to be deciding/creating the game mechanics, and the lack of creativity is due to every developer trying the same sh*t, not wanting to take risks. There is plenty of creative people, they just get shot down. What the AI could help with is the script, missions for characters, etc.
Now for this very particular video what the AI would be doing is generating unique side quests. But I would expect both very bad and good quests. And there is the risks of devs getting to lazy, only doing the main path and leaving everything else to the AI.
We need wider aspect ratio and bigger black bars!
Cara isso é muito louco, isso sim eu posso chamar de nova geração. Imagina as possibilidades dessa tecnologia em jogos de RPG, SAO está cada vez mais presente.
é a nvidia vai dar um jeito de deixar essa porra exclusiva para as placas com a tecnologia de ia.
@NicolasOliveira-bp2dd yeap Nvisia is needing already some competence, they have the hardware monopoly of the most powerful technology, the AI, and really don't know what to do with it
"Hey Jin, how's it hangin'?"
"The Hanging Gardens of Babylon were a massive, terraced garden built by King Nebuchadnezzar II for his wife, Amytis. The gardens were said to be a marvel of engineering and irrigation, and they were one of the Seven Wonders of the Ancient World. However, their existence has been disputed by some historians, and their exact location is unknown. Worst of all, the powerful crime lord Kumon Aoki is causing all sorts of Chaos in the city."
It would be more inversive if the NPC would also greet the player and ask him what's up.
It’s like talking to Microsoft Sam in 1998.
Would love the same scene except with completely different dialog. I want to see how flexible the AI is. Will it always steer the conversation to a predetermined conclusion?
That's definitely interesting, and it can be extremely good when it's rolled out in games. However, I'd like to have heard the answers to some questions that didn't sound scripted; the dialog here was pretty much what you could expect with a multiple-choice menu. How about saying "Hey Jin, how much did you pay for that hat?" or "I found a booger underneath this bar chair, damn that's gross"... I mean, he should be able to respond within parameters - and it'd be key to making the NPC's more lifelike.
You nailed it. They need a better demo video
Yeah, Good point. I'll ask NPC to tell me a joke 😂
They need to keep to using keywords because AI will hallucinate and give bad answers based on it's training data. It's actually really amazing but there are practical limitations. For example, you train an npc to answer questions on politics in a game. Devs would have to write tens of thousands of lines to train the AI on, or just use real world data which is easier and will make the AI sound more realistic but runs the risk of the AI suddenly going on a rant about Trump in your medieval fighting game.
@@mike4402
Yes, it is a real problem. But one possible approach would be to use broad real-world data and heavily restrict it. That's what the chat GPT already does. If you ask it to create a bomb, it will refuse to respond. However, ChatGPT aims to be a comprehensive AI, so it's difficult for OpenAI to impose heavy restrictions. In a game, on the other hand, where an NPC's objective is more focused, it is easier to limit. At any sign of a player discussing real-world politics, the AI could respond that it knows nothing about the subject. Still, some players would certainly be able to "break" the AI, but it wouldn't be as easy.
Character ai site showcases some very decent characters - ask them any question you have and they usually reply in character, so that's not the problem. The voice acting, on the other hand...... I'm thinking they are better off sticking to text - for now - lol
Looks great. AI NPC sounds cool, this was a little monotone in this showcase but excited to see how games use this in the future.
This is very powerful. Think about the level of gaming addiction some are going to get through this.
Imagine trying to convince ai npc to give the keys.
Hey why is this video dimensions this small?
btw this should become a skyrim mode
"Hey Jin, can you finish my easy for me? It is due tomorrow."
I'll throw in a positive word since it seems a lot of people are missing the point. You can make a whole story with this concept in mind. If the game runs small interactions like this, it could also run all the characters in the city dependent on your interactions. There have already been simulations that have been running whole towns of GPT characters that interact with each other. This is just showcasing a concept.
Remember Project Natal...
This is going to end up the exact same way.
NPC just stares at the customer walking into his shop and waiting to be spoken to... a greeting would be nice
To whoever wrote the dialogue... that is not how a (ramen)shop owner would welcome you, let alone task you with defeating a local crime lord... he isn't even trying to feed the character... what the hell...
This is the first time my Samsung Odyssee G9 purchase has been validated. Thank you.
First Façade, now this
I'm excited for lifelike npc's
Love my 4080 but 4060 Ti should have 10 GB vram.
Thats the Nvidia treatment you get when you buy their middle of the pack products. If you wanna stay high end, always buy halo product.
Jin should have been like, "Don't worry, man. What can I get for you?", and when Kai pushed him for answers, Jin should have said, "This isn't something a single person can solve, don't do anything stupid."
chat gpt in a game nice
would like to try a demo of this.
Now we need to be able to use our mics, have the game convert it to text then have the AI respond based on what we actually say
"Nvidia Riva is the company's speech-to-text / text-to-speech solution. In the ACE for games workflow, a gamer will ask a question via their microphone and Riva will convert it to text which is fed to the LLM. The LLM will then generate a text response which Riva turns back into speech that the user will hear."
Why convert it to text, though? At that point, just use the sounds itself or hell. Why even need a character when you can give it your camera video too and interact with you, to you movement and expressions. Text isn't the only median that AI can learn with, the genie without the lamp and rules
@@roblesize that's how voice recognition software works. It transcribes audio. I can't even imagine any other way of doing it. Take audio, turn it to text, input the text. Otherwise you'd have to input audio, still analyse it and determine what is said. Then formulate a response. The best we can do is probably live transcribing. So the AI could in theory interrupt you as it transcribes each word as it's said and then can interject mid sentence. Turning the audio to text is just how it works.
Imagine not being able to use your mic because you live at your parents or with a gf and you want to play at night lol. Now you can't progress because you can't use your mic.
Yes, I know they'll offer text options.
@Vex But why go through that? AI doesn't need to transcribe text. It can literally listen to the audio itself via spectrograms or MFCC. Meh, it's whatever both ways work. I just can't wait to see what this does for gaming. If we aren't all dead or controlled
Is it available for download?
In ramen shops with meal tickets, you have to buy the ticket first before being seated.
I think this was more of an insider demo, the flex was how little work it probably took to set it up.
This did not sell me on it at all lol.
Proposition: Please add to this model options for learn english (with low english skills). For example: I didn't understand what the NPC says to me because I don't know some words. So I can ask NPC for describe these difficult words in other words, or translate it to my native language. Also great will be if I can ask him for more slowly repeat dialog for better understand. I want learn english in this way. (As you can see, it maybe will help me to avoid mistakes in future) Yes, my english isn't well.
Just being able to read scripts with artificial voices in any language will speed up development tremendously. There could be some funny Skyrim type quirks or misunderstanding too. This demo was pretty dry, monotone, but did sound real as far as the shop owner.
The interaction was ok, but it would be interesting if we could type our own text so they can interact in real time generating the conversation. Other important aspect would be AI generative interaction based on what happens around the place or other major events the player is involved somehow.
you do type your own text here already , except you speak it out loud. The main character was the voice of the player into a mic, not the voice of the character.
I think where AI will truly shine in future games is in more advanced collision detection.
No matter how pretty something looks or sounds, as soon as characters start to interact with their environment in current tech the illusion bubble snaps.
I like this idea
That's a good point. If the NPC were washing the plates or watching TV waiting for the customer, then he would notice the player and based on the experience with the player he would say something like hey Kai, wait for an answer then enter interaction mode or something...
Imagine dodging bullets with an actual AI connected to a neural network. I might smash my controller to pieces for the very first time.
@@aeternus80 You don't need such AI to use a neural network. There are already AI mechanics for npc that enables such "cool" features as bullet dodging.
All cool and well but I'm more curious how you guys managed to get 32:9 aspect ratio not looked stretched at the edges. What's the magic behind that. Clearly not panini projection.
what's better is.. the guy is talking to the npc using his microphone.. imagine being in a game like pokemon or metro.. damn it will be so fun
Generative AI NPCs and a player interacting witha VR headset that has eye tracking would be great. If you look rapidly elsewhere he reacts and checks too. Or he could tell you to look at him while he's talking to you if you look away. You might struggle with eye contact and it could effect how they perceive you and treat you.
I'd love to be able tell an npc, "hey look behind you" and rather than point or them look generally behind them, they infer where to look depending on where your eyes glance.
All the input possibilities paired with all the ai possibilities. Could be amazing.
This is absolutely incredible. Keep it up
They just pasted a lot of stuff together but if we consider only the conversation it's basic ai conversation you can have this day, simple and direct.
AI moves very fast, 90s PC hardware fast. I'm OK with it as a tool that gives guidance, which is what I think will eventually win out.
That deadpan stare and cold af delivery. 😵💫
This is far, far away from a shippable product.
This is exactly as unenganging and immersion breaking as I imagined an AI generated NPC conversation would be.
It's a tragedy that inestimable fortunes have gone into this in lieu of supporting artists and authors.
stop smoking crack it's bad for you
to be fair, I blame the dude talking to the jin AI. He seemed like a AI too the way he was talking.
This legit sounded like your average skyrim convo tbh
Finally a video for my 32:9 screen.
That did something similar with Socom on ps2. Using your voice to direct your team
Crazy. Can't wait till it's impemented
Its a bit stiff but with some refinement i can see this being a great tool for devs to create a much more organic experience and hopefully one day as compelling as the primary crafted content we see in the better made games. Think peak bioware storytelling.
Robo-monotone voice aside since that's constantly being improved, I'd have liked to hear more than just one player having that conversation.
The real time conversation is amazing to see, but I'd like to see how it would respond with different phrases and people.
Thanks you for posting 32:9 need more game to support ultrawide community xD
Ey! finally a youtube video in my screens native aspect ratio :D
is the questline pre programmed in or was it just created from the conversation.
Imagine GTA 6 with talking NPCs on street without scripting. You could ask them whatever and they answer, but have premade character so you could bump up to someone who is nervous or feeling happy and use mic to communicate with them....
I recall google displaying its phone calling voice assistant - who sounded very natural. Back then, I thought it was BS - still do. The fact that we haven't seen anything coming even close to that just proves it. It is possible to make AI sound natural, but only with a lot of post-editing. In real time - it sounds like Keanu Reeves trying to act - lol
Ai can massively change video games in the near future. Very interesting !
my gues is if integrating all off the recent nvidia ai will take a lot more computing power
I wish in the conversation, an odd question or statement was asked to show some sort of G-AI response. Oh well... Anyway, I love the ultra-wide screen they used. Even though my monitor is only 2560x1080 (21:9), it looked amazing!
it's like I'm watching from inside a mailbox
So expressive and emotional! Definitely only sounded about 75% as wooden as a generic text to speech engine! I could almost believe i was listening to a worried robot who had recently been shot at, rather than reading an overly expositional speech bubble! 😂
this is a demo you dimwit
Emotions can be easily added, but this video is not about emotions, it's about the speech being real time which is incredibly insane, because it used to take minutes to produce this in real time and now its seconds if not less.
Neat? Shitty wooden sounding instant text to speech already exists. Every twitch streamer sets it up to read resub messages. This is just slightly better wooden sounding TTS.
@@Pest789 you're not very bright
@@AAHAHHHHH I'm aware it's a demo. That makes it even less impressive. Demos are structured to showcase the tech in the best light possible. That means we can expect even less from it in a real application.
FPS RPGs are going to play out like LARPs, where dialogue can go anywhere but funnels down toward the story arcs
I am just wondering is this the working model for NVIDIA ace? because this seems like it was done in unreal with metahuman with voice over
This reminds me of the text based adventure RPGs of the old just revitalized with voice, AI, and crazy good graphics
what's with the super ultra wide screen ratio????
Great, this is what video games need. Quest givers with the emotional range of a marketing caller robot. The Ubisoft padding time is about to be shot through the roof. I would rather have 10 to 20 well-crafted side quests that tie nicely into the main story than this ostentatious busy work giver.
AI generated real-time reflections. AI generated quests. AI generated conversations. Art and writing drop to zero, but you will see a lot of paid AI accounts posting praises for all of these products, hoping to lure the unsuspecting customers into buying this lazy, uninspired horseshit - lol
If they do a good enough job - especially when it comes to spamming review sites with bots - companies creating this AI drivel will still be in the profit, since AI is so cheap (just hire a prompter - lol). At the end of the day, it's all about profit for the big companies. This here is the result when profit becomes the sole motivation for creating products.
This is just a concept, with this technology every single npc in a game world could have depth. Writers could still put time into well crafted quests as well.
Just like procedural terrain gen, this dialog generation will be a tool that makes massive virtual worlds. Think about it more like that and not so much replacing key plot dialog, the comparison is with NPCs that would otherwise have zero dialog.
With how good chat gpt is as a language model you could probably replace key plot dialogue too. Mass Effect style where your choices have impact but instead of having good response vs evil response you actually type/say what you want to and get a different reaction from the NPC.
Guys, notice how they never say this is real time. This is could be tricked somehow. Don't get me wrong, I know IAs for Speech to text, generative text, and Text to Speech exists, and are soo good. However: The time it takes for every one of this process are a few seconds and getting all of that to work as fast as it did here would be amazing, but I don't know if we are there yet. Just remeber that they haven't said this is in real time...the didn't even told the GPU... This is probably WIP
Well, it can be that fast if everything it very well optimized and work pretty good together, hard to tell. The Skyrim demo of this tech for example use stuff that was not developed to work in the best possible way together what ended with a 8 second delay. I bet the speech to text generation at the Skyrim demo started after the user stopped speaking and not run convert the text that was spoken directly after it as the user still talks. That would reduce a lot of lag. Maybe the same is on the text to speech side possible. But no clue if NVidia use such stuff.
@@Hoto74 Speech to text is fast, but you have to wait for the user to end always. In the other hand. I think the slowest process is IA answering and text to speech. Could me super optimized.. I'm just sceptical
If you want you can do a similar thing to what they did in the video yourself, there is already a mod for skyrim with it, there is usualy a couple second delay between the question and answer but it's not so huge, at least in the video that can be found on youtube since I haven't tried it myself yet
2005 called. They want their Oblivion NPC back.
Good: making the video ultrawide
Bad: rendering it with black bars absolutely nullifying any upside to the aspect ratio.
Strong Deus Ex vibes here
Me: Hey Jinn, how are you?
Jinn: As an AI language model, I do not have emotions and can not answer that.
Me: dafaq Jinn
Most of the cases in "real life" will be:
Human - "Just gimme the quest i gotta walk the dog later"
NPC - Hi "name", it's so nice to s...
Human - SKIP godammit
Quest added.
That's nice and all but how can you run a game AND a large language model at the same time on the average gamer's PC ?
A small model like WizardLM-7B takes already like 5GB of VRAM...
I would totally play this if it were a real game.
so wait, is it a pre-recorded dialogue for the playable character, or is he actually talking with his microphone? The mic pic popping up is giving me mixed signals as to what is going on
Finally something for my g9...
...
...
Now to watch literally anything else.
Why is aspect ratio ultrawide?
Not great for actual player interaction, but I could see this being amazing for background dialogue. With generative AI, every random NPC can be having idle conversation as you walk past. So many voices, so much chatter. It will make games feel so much more alive. Once this makes it into games, going back to something like Starfield and being in a populated city space where plenty of people are walking around, but nobody is talking will feel extremely jarring.
Look up Inworld Origins demo if you're interested in stuff like this.
Here's a wild idea, how about game studios just hire voice actors to voice npc? I'm sure the multi million dollar companies could afford that
They already did to train these AI, it’s a techndemo dude, it takes time to build a multi million dollar game
@@SinisterAnimationS Yeah they trained these AI with voices of people who didn't consent to being used that way. Imagine thinking voice actors would actually give their voices for an AI to replicate them forever. Are you stupid?
What about translating the voices for all languages on earth? This AI can be used for any language and thousands of secondary NPCs, this can be used for smaller game studios as well
@@Edu_RJR No small game studio (that I know) will willingly use this.
Using Ai to voice your npcs or non npcs for that matter takes away a job opportunity for actually human beings something that only big companies want to do so they can make even more money.
Source: I work for indie companies
why hire voice actors if you can use the whole web for free to train the AI?
This will be amazing for indie devs ♥ Also, I didn't even know CZcams supported super-ultrawide videos!
Yeah, as long as NGreedia makes it affordable for them. Knowing NVIDIA, they will probably charge an arm and leg for access to this technology.
this tech cannot run on consumer device at least in 2023. only big companies that have massive compute will make these tech affordable to players so that will be amazing for them.
This future is not the creation of a game.
This is the creation of a living world
Bad chatbot inside a video game? Whats the innovation?
Once they implement the use of contractions that will help with the woodenness of the speech.
Finaly first video that fits my screen. 😁😁
they gotta work on the inflections. sounds so robotic without em
"Do you visit the sky district often? What am I saying? Of course you don't!"
We have finally re-reached the conversational capability of Starship Titanic (1998).
Oh finally I can watch a video with the proper resolution on my ultra-ultrawide monitor
Holy papers !
Do not look where we are now, look where we'll be two more papers down the line
Can't wait until we get to a game where we just have some simulations and generate a game by keeping bad people bad after a point and then go on quests to beat them while simulating their Interaction in the background
Every futuristic game there's ramen. Is that what we have to look forward to?
Can't wait to buy my own T-1000 bodyguard in real life.