@@toringmort4231 it should be the norm imo, saves a ton of time on troubleshooting, this is coming from my ai powered video project in skyrim about 4k mods deeps lol so much time as been lost just trying to figure out crashes
People put too much emphasis on graphics rather than the gameplay. People seem to forget where games come from. I think these tools are great, it means a single developer can make something that was impossible to make. Perhaps we will go back to the days were a single developer can make some of the most beloved games. Sure it's not gonna be AAA, but then again most AAA studios are not AAA anymore.
Making a text-based game is probably the easiest type of game to make the genre is known as Interactive fiction and it doesn't need graphics just text this TH-cam video has a nice history of them. When Infocom was still around the vid is called. 1985: Making TEXT ADVENTURE Games | Micro Live | Retro Gaming | BBC Archive
Don't you think this is a biased and very subjective opinion? Games today have become not only entertainment but also media, which is much broader than "pong" from which it comes. In my opinion, the roots of games are not important today because it is the media that gives us the potential to be multi-level art. Visual, narrative, musical art. And it is a great success and an underestimated - also by the creators - luxury. This multi-levelness and depth of some titles gives us the opportunity to immerse ourselves in our favorite aspects of the title. So saying "People put too much emphasis on graphics" is an irrelevant and biased opinion, with which you do not respect what is important to the other person. If someone enjoys traversing beautiful-looking lands without interesting gameplay because it does something to them, it gives them something, then it is none of our business. And this definitely does not mean that anything should be assessed as better or worse.
I think you’re underestimating how powerful AI is going to be in 10 years if you think a single dev won’t be able to make AAA games with AI. This is going to lead to a second renaissance.
Very nice video 😊 I really like you put all the important details as well as the used tools in the description. Calm and clear voice. Good use of memes.
It's interesting how AI naturally wants to make stuff with that "Unreal Engine" look. Guys, it might be time to switch to a more niche art style if you want to avoid company layoffs. =\
AI naturally genrates the most average, most common crap of which there are most examples out there. It is not the AI's fault that the training data is full of what this guy's idea of a game is.
@@clray123 Well, yeah. It's not a human-being. It can't be truly creative. You take the artists who created the art style for Psychonauts, for example. A.I. can mimic what's already there, but it has a heckin' hard time doing what comes easy to the original artist: Creating new ideas.
@@HTMangaka The human artist does not create anything 100% new. What he does is use his library of knowledge in the field, and use that knowledge (assembling it together in an ever new way) sometimes changing something. Whoever invented the computer did it because someone had invented transistors, transistors were invented thanks to electrical energy etc. etc. This is exactly what AI does, the only difference is that a human has the ability to understand if a mistake is really a mistake or if it is a cool thing to work on... I don't know if I'm explaining myself. Revolutionary ideas often arise from mistakes, but you have to be able to intuit that that mistake can be "beautiful" and developable. For now AI can't but it will soon be able to.
@@smokysmoka No, the difference is that the Human was THERE to experience the things in his mind. With Five senses. Sight. Sound. Taste. Touch. Smell. Everything in an A.I's brain is Second-hand information. It's the same thing as asking someone raised in a pampered high-class society to imagine poverty when all they have to go by are stereotypes.
But then what happens when there are options to tell the AI to change to various styles and it begins to be able to do it well? Just remember this is the worst it will ever be.
As if gamers like me cared one bit about that 😂 Most gamers only care about whether or not a game is fun and interesting. It could even have been made by a donkey for all I care. I have played a lot of “hand crafted” crap, so that doesn’t mean much to me.
That isnt how that would work... more games =/= cheaper games. Compare today from just 10 years ago. More games come out every other month than they used to. It used to be that youd only get a handful of games a year, maybe 1 or 2 big games a year, now we get a lot more and the price of games hasnt changed much, in fact theyve gone up for most triple a games. There will always be a market for games, regardless of how many come out, and theres more factors than just the saturation of the market.
@@etherealfields You can't compare now to 10 years ago, there was still powerful UE4 back then. 3d model generators and gpt5 will let people with 10 weeks of experience catch up to people with 10 years. There won't just be vast saturation, there will be a ton of more high quality games, raising the bar at unprecedented levels. Literally nobody will pay 20-30 dollars for an indie game when unreal engine itself is becoming the game
@@prodev4012 i absolutely can. the whole point is that there are more games being released today, than even 4 years ago. regardless of the tools, or how many there are being released, prices wont drop like that due to a "vast saturation" as you say. AI is a tool, it isnt some magical being that will allow everyone to make a game, there will always be need for a creative/human element behind it. if anything, opening up game dev like this with things like 3d model generators and people using gpt to write their games, it will lead to even more low quality games, regardless of the quality of the code or the models that get made, if the person behind the games has no experience, doesnt understand the fundamentals of game development, level design, etc. it will lead to a bad game 100% of the time. even if the bar raises to unprecedented levels, the floor wont, there will always be shitty, low quality games. edit: this isnt to say that what this guy did with AI and other use cases of AI ISNT IMPRESSIVE, this is to say that if you think ai will lead to anyone and everyone being able to just generate a game, youre sorely mistaken, and youre a fool if you think this will tank the market. a flood of high quality games will always be a good thing.
What's wild is that as soon as I learn blender and programming, I'll be able to do every single thing that you were able to do with ai and I'm so excited!
As an indie game dev that does 3D modelling, "rigging", animations, C++/C# coding, environment design, vfx, sfx, cinematics, gameplay design, networking/server management, and database management I am not threatened in the slightest way. In fact I welcome it because I have made an incredible influx of people asking me to help clean up their AI garbage to be realistically useable in a "finished product" not just a quick demo in engine. The most common believe it or not is code. AI code is still worse than fresh college students that have a strong understanding of code but never did game design with unreal engine. It creates so much redundant/junk code that usually is not meant to communicate with other code being made because most users do not think about that issue since they are not coders themselves for the AI model to follow (I don't consider someone who is copy pasting AI code "coders"). This is where someone can bust out a quick "test model" like the video says, but as far as finished product, if you think Ubisoft was putting out garbage unplayable games, you just wait lol. Same with the 3D models, the poly count and geometry is fucking horrible to have to fix where over half the time I tend to just remodel the model myself with proper geometry. Weapons, armor, and the worst of all armpits have geometry that makes my 11 year old son's 3d work look professional. Everyone here in the comments that are impressed and thinking "real devs" are in danger don't have the slightest clue what it actually takes to develop a playable game that people can enjoy. As Rafal said, its fine for prototyping, but in no way will anyone "actually want" an AI game. Again I am not shaming it because it is actually creating more work for devs like myself cleaning up and fixing projects that maybe 5% of them will ever see final production usually due to motivation and understanding what it takes to put a game out. Steam already robs devs of their money take 30%, plus the government tax taking another 30% (USA, differentiates from country and is usually worse in others), and then smaller %s from programs like Unreal Engine and Blender/Maya if your game is making over $100k (before the % reductions). So if you publish a game on steam a dev is usually seeing about 20-30% the purchase and that doesn't include purchases that are being made in countries where the VAT winds up making the dev pay Steam for their purchase if the game isn't priced high enough (yeah look this up for a fucked up loophole). Enjoy that word wall of knowledge. Feel free to ignore it. Reality doesn't care about what people choose to not listen to.
I'm a data engineer who uses chatgpt all the time for code, it's about scope. If you ask it, "Create a function that does X" it can do that pretty easily. If you ask it: "Create a fully fledged data pipeline that does A to Z" well then you're just asking for trouble. Chatgpt is 1000 times better than a freshman cs major. You can also give it cryptic error messages and it will pull out a solution. You can ask it to clean up your code and make it usable. You can get it to write you doc strings. It can find potential issues in your code. AI is a wonderful tool but, at the moment, it is not replacing anyone. What I'm excited about is that art has been a massive barrier for me in wanting to make video games. I usually buy an art pack or just make little squares. Any even then with a wife, work, and kids, I don't want to waste my time on things. So if there are options to have ai automate a lot of the work, I'm gonna do that. Leave to professional artists to do professional work, I can't afford 1000 bucks for a few assets.
@@6ixlxrd Somebody working in game dev (mainly 3D art) here - fairly sure there was no insecurity involved. He was just explaining why using AI exclusively won't make good games in the near future, and to explain that to somebody without knowledge about what it takes to make a game it takes more than a couple short sentences. He's absolutely right - when I look at the 3D models meshy created, I was impressed by how far it has come, but it's still in no way a replacement for real artists, both on a technical and aesthetic level. Who knows, maybe within a decade or so things change and AI actually gets closer to a useable-out-of-the-box level, but as it stands, realistically AI tools for creating assets aren't much more than easy to use tools to block things out for further refinement afterwards. And that's not very threatening but rather helpful.
@@tobiasfrey2793 As someone who also works in game dev (primarily animation), I’m positive that there was insecurity here. “I am not threatened in the slightest way” was the biggest tip off…no one mentioned anything threatening. I develop games and I’m impressed by how much AI can immediately do what takes me ages to do. Being impressed isn’t wrong, but being intimidated by people who are impressed and going on the defensive is factually insecurity.
it's crazy to think about how far AI how has gotten, and the fact that it''s still rapidly evolving. I will definitely be trying out some of these resources in my game
There's definitely still a lot to be desired, but it's good to mockup prototypes. In terms of the asset shaders, you could use the material shaders imported from Megascans to help polish the look.
People will just need to adapt, as the Luddites had to adapt when the printing press was invented. One new technology comes out which makes life easier, there's no putting the genie back in the bottle.
the topology on AI models is a disaster. Editing is more of a hassle than creating yourself. I would definitely call it an idea generating tool. I think an AI tool that cleans up models would be more useful. Including rigging and weight paint. There are very few reliable tools for this. It's very time-consuming process that doesn't give personality to a character/object.
Absolutely. I think it's a great time saver for blocking out characters (just dynamesh the bad topo and go from there), mainly for organic ones without much cloth and/or armor. Then again, blocking out a character shouldn't take that long, either, if you already have a concept ready. Also, agreed on the rigging part.
My philosophy with AI is simple. The internet currently has all the information required to launch a Rocket into orbit right now, yet SpaceX is the only private entity to do it because it takes serious talent and expertise to pull that off. The same applies to games, only talented and passionate people/teams will launch worthy projects, the rest of people online will launch Soulless(excuse the pun) lukewarm junk that no one will ever play. These tools are fantastic and can only get better...
When you sit and feed prompts and images for hours and still end up with flawed Ai models then you realllly understand how far there is to go. When your client says add this, fix that, and you can’t do it. What will you do then?
By prototypes, I mean the bare basics of a game can be put together to show studios what the concepts and game play would be. After others have seen what the plan is, and discuss what could be, then the game would be made.
I love your dev logs !! they provide impressive insight on how big studios tackle problems , also I have a request could you please make a video on how you should start learning Game development ?
I Think the best way to use AI for models is to have an image generator make the art of the character, then get an AI model for that art, then bring that into blender as a reference to make your OWN model. This way you dont have any of those wierd AI artifacts in the model.
I'm not sure AI will ever be able to replace hoomans until we reach a point where AI itself becomes conscious and knows what it's doing, why it's doing it, is able to look back over its work and fix issues, build upon existing work, etc... Until then I think it'll just be incorporated into gaming engines as ways to speed up development like with the UE5 plugin you mentioned. I'm not a game dev, just a software/web engineer... but this was a very interesting watch, so thank you for sharing.
Thank you for sharing your experience of using AI! I'm sure I'll be able to benefit from it. Recently, I've been thinking about making a game, but as someone who isn't very good at art, I'm finding it hard to get assets with a consistent visual style. I can't wait to see your next video on developing a cartoon style game. Hopefully, it will help me again as this video did!
The cool thing about AI is that it might be able to make the concept/prototype idea much faster and more accurate. Imagine AAA studios hiring a guy to make concepts just like this and then the production team takes it from there.
@@Circled123 thats the future I want but nobody has yet made a retopology tool. I think its because Topology is harder to market than making models from scratch, most people dont know about the retoplogy or LODs step in model creation.
The "cleanup" is sometimes just as labor-intensive (or more so) than simply doing it right from the start. The way I see it is that AI produces bad quality prototype assets that have to be thrown away and replaced by good quality assets. But many greedy companies or small devs will simply try to sell the bad quality, leading to an overall enshittification of the game market - where barriers to entry are lowered, product quality must suffer.
@@clray123 it depends because ai can take away a lot of the guesswork in the concreting phase and I can use the terrible AI mesh as a base mesh to sculpt on in zbrush But I agree it will flood the market with terrible 3d models. I’m still grappling with the morality of using AI But I think it’s better for Artist to make something with it than AI bros to profit off of it, Artist will always do a better job with AI than someone who exclusively works using only AI. I just wish someone would make a AI retopology tool that would even the playing field a lot
Would nice to see more info about using AI with UE5 regarding guidance and Blueprints optimization. Materials +Ai could be useful too. Thanks a bunch, great video. 👍
I would *love* a video on how to use AI for game logic. Oh please, yes! 🙏🏻 In fact, I’d love to learn to make a game by making an AI prototype first (within a layout, progression and gameplay mechanics I define myself), and *then* learn the skills to hand edit each asset and feature after the fact. Oh, that would be amazing!
15 years fast forward .... you will tell AI how you exactly want your VR game and it will create a totally real looking virtual world for you ... Holodeck similar experience is coming closer and closer ... But therefore hand and bodytracking has to be improved and some customer friendly lightweight gloves with tactile feedback and good walking pads have to be available. This would make a huge impact to the future of VR. Modded Skyrim VR already allows talking via microphone with the characters and AI slips into their roleand let them answer as they were real. What's soon coming is the replacement of 3D modeled characters by AI created real life looking characters ... There is a huge new artificial world coming toward us.
Its really interesting what nowadays AI can do, but still some of the AI tools requires to be "pushed" to receive good reply or model, but its capable to help much if you know how to use it. Before it was quite hard being as solo developer to do on your own everything, but now you can try asking for help from AI and it will actually help and if you don't understand, he can explain in different ways. Also haven't heard few tools that was shown in this video and will be nice to explore them. Great video Rafal! Also starting to wonder when we will see developed games with AI help and how people will be reacting to it :)
The thing these models look bad when you have freedom to look around them Also you should have built the level with ai models, like 99% of this ue5 scene and game isn't Ai I already can see textures not mixing together right when placed
Very impressive stuff, great algorithm pull! Had to give Meshy a try after this video, as I hadn't realized Text to 3D was already so advanced. Of course, the shovelware bloat is inevitable with the advancement of AI, but if even one artist can make good use of these tools, be it for themselves or for the gaming community, then it is well worth it, in my opinion.
The problem with this analysis is that it suffers from recency bias. It doesn't account for AI technology's ability to create even more detailed 3D models in the future, matching or surpassing the capability of a human artist. Right now, the human artist is considered the partner and AI is the tool, but one day, AI will become the partner and the human artist will become the tool.
The only problem I see with Meshy is the lighting that is integrated directly in the texture, it's why it's stand out so much when it's put inside a game engine. To fix it maybe do the texture yourself or in the img to 3D use an image of the 3d model you want with neutral lighting
Ai is capable of matching your creativity if you are willing to dump the time into learning AI as a complete skillset. AI is iterative, and requires small clean up. It always will, just like real art. But refining what prompting works with what models etc, comes from time spent doing it. How good a model is depends on how clean and concise your prompt is, and how many iterations of a model you're willing to run through generating. The thing about AI is that it will never do the complete job flawlessly. Again, just like real art, it requires clean up. You will need to have SOME skills in order to make effective use of it.
I don't know, man, I am currently making a project for uni and I kinda don't have time to learn to model and animate for it, so I went to the internet to download some stuff just so I can show something more proper there, which would all be changed when I wanna actually release the game. But using AI feels so damn empty, even if it gives the needed results. If a person did not do it, it feels kinda lacking in a way. But anyway, good video for its purpose!
saddest thing is people using whole templates from UE or Unity and not changing anything and release it, looking rly bad and they thinking its gonna sell..
Thanks for this (subscribed). Maybe raytracing will help with the integration of the prop models into the environment. The model detail itself will improve over time in Meshy (I guess), so that the texture map will not look so high detail onto low poly, or maybe even eliminate the need altogether for UV texture maps.
Great video, very interesting stuff. I used to work in games development (over a decade ago now) so my opinions are probably not as up to date as some, but I feel like the current situation of studios growing bigger and bigger, making gigantic games over the course of years, simply isn't sustainable. And maybe AI is a tool that can step in to help? If you have a real human artist creating textures - for example broken floor tiles - then can AI produce variations in a few minutes that might have taken the artist hours or days? Similarly, Can they use AI to fill an environment with variations of trees, rocks, and ruins originally built by a human? Each example like this might only save a modest amount of time, but repeat that thousands of times across the development cycle and suddenly budgets and timescales might become a LOT more realistic, especially for smaller teams working without the benefit of gigantic publishers footing the bill.
Thank you very much for this video. I'm trying to create my game solo, and it was more of a "learning" than "creating" experience. I don't know why I was ignoring AI tools all this time. Thanks again! And a quick question - what's the style or who's the original author of the illustrations like the one in 5:32? Love how it looks.
I'm less worried about the details and more wondering about the topology and efficiency. An artist will put the details in the right place because they understand how a model will be animated, where the players will be looking, and so on. An AI will just throw random clumps of pointless geometry all over the place, eating up resources, and causing animators an absolute nightmare. AI: It can do things. If all you want is "anything" - AI can probably do that. It can never do them very well.
I can bet my house that it didnt took this guy 2-3 hours to make that game. The dude is obviously a professional game dev, but stating that it took him 2-3 hours is just ridiculous.
11 วันที่ผ่านมา +1
If you know what you aim for and have enough skill and past experience, it's more than doable. If I can judge by different lens - with right framework you may set up a web app prototype in such timeframe, but you must know your tools 😉
Thank you so much for this inspiring video. I can see some individual developers are aiming to match top notch latest games provided by multi hundred million if not billion companies and that seems to have improved their output finely. If they would only compromise some high end quality for a more causal game they might be able to find some lucky millions around the corner. Candy crush for example made about a billion USD on its first year, it was a very trivial match making game which compared to some individual developers would probably be easily implemented with current AI tech and fully satisfied. The answer to the apparent question about AI capabilities does not seem to have a yes or no answer (on whether it can or cannot support in implementing games) but rather (how good exactly will the implementation be or how much can AI support in accomplishing)
I'm not sure, that we will always need a human touch. AI tech is not there yet and won't be for a while longer. But there have been massive improvements in really short amounts of time and there is no sign of the tech slowing down it's progress. While we might need the human touch for a few more years, that may change in a decade. What about two decades? Looking at where AI was just a few months let alone years ago and where it's now, that future where we could get high quality assets that are indistinguishable from an experienced artist in seconds may come sooner rather than later.
if I'm honest I'll only use ai for the art and that's it cuz I enjoy coding and solving bugs in the code because it gives me a feeling of accomplishment
Any thought about ai could help at the end of the dev to optimize a game? To push down the minimum reqs from the fancy x3d cpus and xx90 rtx cards to "peasant" level hardware, and you get reasonable performance+quality.....
But why do you need AI-generated 3d models for greyboxing where you can simply drop in free premade models for the same purpose. One could say the AI models are somewhere in the middle ground between production quality and test dummy... using them is like lying to management that your game is in far more advanced stage than it really is, the only "reward" you will get is your salary cut because "it looks ready" to the morons in suits.
Really!!! No c coding knowledge required? I feel its too overwhelming? Does unity at all require c coding knowledge before to learn? For easier learning curve?
From one side 3d generation come a long way, and it's looks amazing from the technical standpoint, but still far from production ready. So as you said it's may be good tool for like prototyping. But same time how this differs from just buying premade 3D assets? And as with just using only premade assets before "AI slop" we had a lot of "Asset flip" games
The way it's going I'll probably learn to use AI to make games before I learn to code properly, which I've done a bit with python. I've got a project in mind that is specific enough I could probably give AI clear instructions for making it, but I lack motivation, and I'm not really sure how to start.
I wish the AI tool can become much polish and advance. Such that even I don't have any skills related to game industry, I can still create my own game with my imagination. The concept is that AI will not replace our works, but people who learn advance technology like AI tool can replace us.
While AI is seen as controversial, its benefit for making quick prototype ideas into gameplay cannot be understated, making use of this allows what could typically be a hundred hour endeavour to become a dozen hour mission to see if the idea you have is even worth investing. Of course though, it can also be used incorrectly especially when a company is run by someone who doesn't understand the requirements of a game!
@MartinDlabaja so you created another universe in this one with AI? Great, we are ready for the next level of gaming then. Btw sell it as Artificial Engine 1, I am sure people would pay billions for it.
jak na dwie godziny używając tylko AI jest to niepokojące, jeszcze długa droga żeby ludzie zostali zastąpieni przez AI w game developingu, ale dalej jest to niepokojące
Yo Rafal! That was an epic journey into the AI game dev world! 🌟 I was ready to slay some dragons but ended up vibin' with some serious creativity instead! 🎮🔥 Your visuals were on another level, and the AI antics had me laughin' like a toddler with Legos! 😂 Keep up that fresh approach, 'cause you've got somethin' special cookin'! Can't wait for the next one! And just remember, team up with AI but don’t let it skip leg day! 💪💖 #ReactionQueenApproved
I've tried to use ChatGPT to get me started with blueprints but it will often tell me to create nodes that don't exist. I assume what you would need to do is scrape the unreal blueprints documentation into a json datasheet and finetune a LLM with that data.
There are a few tricks you can try - for example tell ChatGPT to „only rely on actual documented nodes” and tell it „if you do not know the answer - look it up in documentation but do not guess”. Another thing I noticed is that once you correct it a few times, it starts getting things more accurate
There are a few tricks you can try - for example tell ChatGPT to „only rely on actual documented nodes” and tell it „if you do not know the answer - look it up in documentation but do not guess”. Another thing I noticed is that once you correct it a few times, it starts getting things more accurate
What also works for me is: Copy paste the part of the documentation to a separate chat and solve the problem there. That is the most accurate practice I've found so far. You can also upload pdf documentation if it has one...
using ai to debug crashlogs would be a godsend
Bethesda did that with Starfield I think
@@toringmort4231 it should be the norm imo, saves a ton of time on troubleshooting, this is coming from my ai powered video project in skyrim about 4k mods deeps lol so much time as been lost just trying to figure out crashes
People put too much emphasis on graphics rather than the gameplay. People seem to forget where games come from. I think these tools are great, it means a single developer can make something that was impossible to make. Perhaps we will go back to the days were a single developer can make some of the most beloved games. Sure it's not gonna be AAA, but then again most AAA studios are not AAA anymore.
It's like watching movies in low quality. I mean it's not the main point, but you know...
Making a text-based game is probably the easiest type of game to make the genre is known as Interactive fiction and it doesn't need graphics just text this TH-cam video has a nice history of them. When Infocom was still around the vid is called. 1985: Making TEXT ADVENTURE Games | Micro Live | Retro Gaming | BBC Archive
Don't you think this is a biased and very subjective opinion? Games today have become not only entertainment but also media, which is much broader than "pong" from which it comes. In my opinion, the roots of games are not important today because it is the media that gives us the potential to be multi-level art. Visual, narrative, musical art. And it is a great success and an underestimated - also by the creators - luxury.
This multi-levelness and depth of some titles gives us the opportunity to immerse ourselves in our favorite aspects of the title. So saying "People put too much emphasis on graphics" is an irrelevant and biased opinion, with which you do not respect what is important to the other person.
If someone enjoys traversing beautiful-looking lands without interesting gameplay because it does something to them, it gives them something, then it is none of our business.
And this definitely does not mean that anything should be assessed as better or worse.
I think you’re underestimating how powerful AI is going to be in 10 years if you think a single dev won’t be able to make AAA games with AI. This is going to lead to a second renaissance.
@@MildlyAutisticApe sadly not, because renaissance was about humans. Maybe you meant industrial reevolution.
Very nice video 😊 I really like you put all the important details as well as the used tools in the description. Calm and clear voice. Good use of memes.
It's interesting how AI naturally wants to make stuff with that "Unreal Engine" look.
Guys, it might be time to switch to a more niche art style if you want to avoid company layoffs. =\
AI naturally genrates the most average, most common crap of which there are most examples out there. It is not the AI's fault that the training data is full of what this guy's idea of a game is.
@@clray123 Well, yeah. It's not a human-being. It can't be truly creative. You take the artists who created the art style for Psychonauts, for example. A.I. can mimic what's already there, but it has a heckin' hard time doing what comes easy to the original artist: Creating new ideas.
@@HTMangaka The human artist does not create anything 100% new. What he does is use his library of knowledge in the field, and use that knowledge (assembling it together in an ever new way) sometimes changing something. Whoever invented the computer did it because someone had invented transistors, transistors were invented thanks to electrical energy etc. etc. This is exactly what AI does, the only difference is that a human has the ability to understand if a mistake is really a mistake or if it is a cool thing to work on... I don't know if I'm explaining myself. Revolutionary ideas often arise from mistakes, but you have to be able to intuit that that mistake can be "beautiful" and developable. For now AI can't but it will soon be able to.
@@smokysmoka No, the difference is that the Human was THERE to experience the things in his mind. With Five senses. Sight. Sound. Taste. Touch. Smell.
Everything in an A.I's brain is Second-hand information. It's the same thing as asking someone raised in a pampered high-class society to imagine poverty when all they have to go by are stereotypes.
But then what happens when there are options to tell the AI to change to various styles and it begins to be able to do it well? Just remember this is the worst it will ever be.
This is a blessing in disguise, actual hand crafter artist is going to be recognize so much more now
Ur mom 😂
Most people don't care about nuances of hand-crafted quality.
As if gamers like me cared one bit about that 😂
Most gamers only care about whether or not a game is fun and interesting.
It could even have been made by a donkey for all I care.
I have played a lot of “hand crafted” crap, so that doesn’t mean much to me.
This tendency to convert everything into art has made everything garbage
Idk man it seems like 2 more years and game dev will be so saturated you will be lucky to sell a game for 99 cents on steam
good game will always sell
@@TheMadKid91 yeah, for 99 cents...
That isnt how that would work... more games =/= cheaper games.
Compare today from just 10 years ago. More games come out every other month than they used to. It used to be that youd only get a handful of games a year, maybe 1 or 2 big games a year, now we get a lot more and the price of games hasnt changed much, in fact theyve gone up for most triple a games.
There will always be a market for games, regardless of how many come out, and theres more factors than just the saturation of the market.
@@etherealfields You can't compare now to 10 years ago, there was still powerful UE4 back then. 3d model generators and gpt5 will let people with 10 weeks of experience catch up to people with 10 years. There won't just be vast saturation, there will be a ton of more high quality games, raising the bar at unprecedented levels. Literally nobody will pay 20-30 dollars for an indie game when unreal engine itself is becoming the game
@@prodev4012 i absolutely can. the whole point is that there are more games being released today, than even 4 years ago. regardless of the tools, or how many there are being released, prices wont drop like that due to a "vast saturation" as you say.
AI is a tool, it isnt some magical being that will allow everyone to make a game, there will always be need for a creative/human element behind it. if anything, opening up game dev like this with things like 3d model generators and people using gpt to write their games, it will lead to even more low quality games, regardless of the quality of the code or the models that get made, if the person behind the games has no experience, doesnt understand the fundamentals of game development, level design, etc. it will lead to a bad game 100% of the time. even if the bar raises to unprecedented levels, the floor wont, there will always be shitty, low quality games.
edit: this isnt to say that what this guy did with AI and other use cases of AI ISNT IMPRESSIVE, this is to say that if you think ai will lead to anyone and everyone being able to just generate a game, youre sorely mistaken, and youre a fool if you think this will tank the market. a flood of high quality games will always be a good thing.
That's amazing work actually. I'm starting my own dev, so I know how hard to create that kind of visuals... Incredible job, man, keep going!
What's wild is that as soon as I learn blender and programming, I'll be able to do every single thing that you were able to do with ai and I'm so excited!
Yeah as soon as you learn blender😃😃😃…
@@4henn155 It's been only 12 years I'm still learning but progress is great feels like im closer to finishing every year!
bro if you learn blender and programming, you can literally make good looking/working games already without ai.
Go on I give up in years 3, because I have a lot of real life work to do, and with Ai in the future it's seems so helpful
@@4henn155 lmao I know right 🤣🤣🤣
As an indie game dev that does 3D modelling, "rigging", animations, C++/C# coding, environment design, vfx, sfx, cinematics, gameplay design, networking/server management, and database management I am not threatened in the slightest way. In fact I welcome it because I have made an incredible influx of people asking me to help clean up their AI garbage to be realistically useable in a "finished product" not just a quick demo in engine. The most common believe it or not is code. AI code is still worse than fresh college students that have a strong understanding of code but never did game design with unreal engine. It creates so much redundant/junk code that usually is not meant to communicate with other code being made because most users do not think about that issue since they are not coders themselves for the AI model to follow (I don't consider someone who is copy pasting AI code "coders"). This is where someone can bust out a quick "test model" like the video says, but as far as finished product, if you think Ubisoft was putting out garbage unplayable games, you just wait lol. Same with the 3D models, the poly count and geometry is fucking horrible to have to fix where over half the time I tend to just remodel the model myself with proper geometry. Weapons, armor, and the worst of all armpits have geometry that makes my 11 year old son's 3d work look professional. Everyone here in the comments that are impressed and thinking "real devs" are in danger don't have the slightest clue what it actually takes to develop a playable game that people can enjoy. As Rafal said, its fine for prototyping, but in no way will anyone "actually want" an AI game. Again I am not shaming it because it is actually creating more work for devs like myself cleaning up and fixing projects that maybe 5% of them will ever see final production usually due to motivation and understanding what it takes to put a game out. Steam already robs devs of their money take 30%, plus the government tax taking another 30% (USA, differentiates from country and is usually worse in others), and then smaller %s from programs like Unreal Engine and Blender/Maya if your game is making over $100k (before the % reductions). So if you publish a game on steam a dev is usually seeing about 20-30% the purchase and that doesn't include purchases that are being made in countries where the VAT winds up making the dev pay Steam for their purchase if the game isn't priced high enough (yeah look this up for a fucked up loophole). Enjoy that word wall of knowledge. Feel free to ignore it. Reality doesn't care about what people choose to not listen to.
ditto
“I am not threatened in the slightest way.”
*proceeds to ramble an entire bible to stroke the insecurity from his threatened ego*
I'm a data engineer who uses chatgpt all the time for code, it's about scope. If you ask it, "Create a function that does X" it can do that pretty easily. If you ask it: "Create a fully fledged data pipeline that does A to Z" well then you're just asking for trouble. Chatgpt is 1000 times better than a freshman cs major.
You can also give it cryptic error messages and it will pull out a solution. You can ask it to clean up your code and make it usable. You can get it to write you doc strings. It can find potential issues in your code.
AI is a wonderful tool but, at the moment, it is not replacing anyone.
What I'm excited about is that art has been a massive barrier for me in wanting to make video games. I usually buy an art pack or just make little squares. Any even then with a wife, work, and kids, I don't want to waste my time on things. So if there are options to have ai automate a lot of the work, I'm gonna do that. Leave to professional artists to do professional work, I can't afford 1000 bucks for a few assets.
@@6ixlxrd Somebody working in game dev (mainly 3D art) here - fairly sure there was no insecurity involved. He was just explaining why using AI exclusively won't make good games in the near future, and to explain that to somebody without knowledge about what it takes to make a game it takes more than a couple short sentences. He's absolutely right - when I look at the 3D models meshy created, I was impressed by how far it has come, but it's still in no way a replacement for real artists, both on a technical and aesthetic level. Who knows, maybe within a decade or so things change and AI actually gets closer to a useable-out-of-the-box level, but as it stands, realistically AI tools for creating assets aren't much more than easy to use tools to block things out for further refinement afterwards. And that's not very threatening but rather helpful.
@@tobiasfrey2793 As someone who also works in game dev (primarily animation), I’m positive that there was insecurity here. “I am not threatened in the slightest way” was the biggest tip off…no one mentioned anything threatening. I develop games and I’m impressed by how much AI can immediately do what takes me ages to do. Being impressed isn’t wrong, but being intimidated by people who are impressed and going on the defensive is factually insecurity.
Some cool visuals and big monsters to fight and I'm halfway to buying. Lol
it's crazy to think about how far AI how has gotten, and the fact that it''s still rapidly evolving.
I will definitely be trying out some of these resources in my game
surge of indie game devs incoming
There's definitely still a lot to be desired, but it's good to mockup prototypes. In terms of the asset shaders, you could use the material shaders imported from Megascans to help polish the look.
The fact that we EVER got to this point, and how QUICKLY the technology is evolving is concerning enough.
People will just need to adapt, as the Luddites had to adapt when the printing press was invented. One new technology comes out which makes life easier, there's no putting the genie back in the bottle.
the topology on AI models is a disaster. Editing is more of a hassle than creating yourself. I would definitely call it an idea generating tool. I think an AI tool that cleans up models would be more useful. Including rigging and weight paint. There are very few reliable tools for this. It's very time-consuming process that doesn't give personality to a character/object.
Absolutely. I think it's a great time saver for blocking out characters (just dynamesh the bad topo and go from there), mainly for organic ones without much cloth and/or armor. Then again, blocking out a character shouldn't take that long, either, if you already have a concept ready. Also, agreed on the rigging part.
Thanks for the video. I liked what you said about Ai at the end. It’s a tool and it’s not meant to replace people just help
AI is only a tool until it's not.
You wouldnt release it as a game, but a AAA game studio would.
My philosophy with AI is simple. The internet currently has all the information required to launch a Rocket into orbit right now, yet SpaceX is the only private entity to do it because it takes serious talent and expertise to pull that off. The same applies to games, only talented and passionate people/teams will launch worthy projects, the rest of people online will launch Soulless(excuse the pun) lukewarm junk that no one will ever play. These tools are fantastic and can only get better...
There may be a time when AI can be heavily used to create games, but right now it's value is in prototypes.
When you sit and feed prompts and images for hours and still end up with flawed Ai models then you realllly understand how far there is to go. When your client says add this, fix that, and you can’t do it. What will you do then?
By prototypes, I mean the bare basics of a game can be put together to show studios what the concepts and game play would be.
After others have seen what the plan is, and discuss what could be, then the game would be made.
I love your dev logs !! they provide impressive insight on how big studios tackle problems , also I have a request could you please make a video on how you should start learning Game development ?
Yes, I will make it :) thanks for the nice comment ♥️
Pick game -> pick engine -> just start
do you have a link to the movement animation sample that you downloaded from unreal engines asset store?
I Think the best way to use AI for models is to have an image generator make the art of the character, then get an AI model for that art, then bring that into blender as a reference to make your OWN model. This way you dont have any of those wierd AI artifacts in the model.
I'm not sure AI will ever be able to replace hoomans until we reach a point where AI itself becomes conscious and knows what it's doing, why it's doing it, is able to look back over its work and fix issues, build upon existing work, etc... Until then I think it'll just be incorporated into gaming engines as ways to speed up development like with the UE5 plugin you mentioned. I'm not a game dev, just a software/web engineer... but this was a very interesting watch, so thank you for sharing.
Able to reduce 70~80% workloads is very powerful already, I guess we could assure this means over 50% of humans will lose their jobs
@computervision557 Does it really reduce that much of the workload though? I would've guessed more like 30~40%.
@@macjabeth5073 After 5~10 years, there might be over 70%
Maybe in 5~10 years, we should ask a question
Why should the capital let us live if we are not exploitable?
@@computervision557 #badbot
Thank you for sharing your experience of using AI! I'm sure I'll be able to benefit from it.
Recently, I've been thinking about making a game, but as someone who isn't very good at art, I'm finding it hard to get assets with a consistent visual style. I can't wait to see your next video on developing a cartoon style game. Hopefully, it will help me again as this video did!
Thanks for kind words!
The cool thing about AI is that it might be able to make the concept/prototype idea much faster and more accurate.
Imagine AAA studios hiring a guy to make concepts just like this and then the production team takes it from there.
why not using cascadeur for animations ?
Will try ot out!
You are so smart. It was very interesting. Thanks.
I can see in the future a 3d artist could take the mesh into Zbrush to clean it up and add details, or even more Items.
Yep, that’s valid, such cleanup is partially what I mean by additional „human touch”
I think the opposite is quite likely as well where an artist makes the model and then AI cleans it up, generates LODs etc.
@@Circled123 thats the future I want but nobody has yet made a retopology tool.
I think its because Topology is harder to market than making models from scratch, most people dont know about the retoplogy or LODs step in model creation.
The "cleanup" is sometimes just as labor-intensive (or more so) than simply doing it right from the start. The way I see it is that AI produces bad quality prototype assets that have to be thrown away and replaced by good quality assets. But many greedy companies or small devs will simply try to sell the bad quality, leading to an overall enshittification of the game market - where barriers to entry are lowered, product quality must suffer.
@@clray123 it depends because ai can take away a lot of the guesswork in the concreting phase and I can use the terrible AI mesh as a base mesh to sculpt on in zbrush
But I agree it will flood the market with terrible 3d models.
I’m still grappling with the morality of using AI
But I think it’s better for Artist to make something with it than AI bros to profit off of it, Artist will always do a better job with AI than someone who exclusively works using only AI.
I just wish someone would make a AI retopology tool that would even the playing field a lot
Would nice to see more info about using AI with UE5 regarding guidance and Blueprints optimization. Materials +Ai could be useful too.
Thanks a bunch, great video.
👍
I would *love* a video on how to use AI for game logic. Oh please, yes! 🙏🏻
In fact, I’d love to learn to make a game by making an AI prototype first (within a layout, progression and gameplay mechanics I define myself), and *then* learn the skills to hand edit each asset and feature after the fact. Oh, that would be amazing!
Ai for the games logic is the most interesting topic actually.
15 years fast forward .... you will tell AI how you exactly want your VR game and it will create a totally real looking virtual world for you ... Holodeck similar experience is coming closer and closer ... But therefore hand and bodytracking has to be improved and some customer friendly lightweight gloves with tactile feedback and good walking pads have to be available. This would make a huge impact to the future of VR.
Modded Skyrim VR already allows talking via microphone with the characters and AI slips into their roleand let them answer as they were real.
What's soon coming is the replacement of 3D modeled characters by AI created real life looking characters ... There is a huge new artificial world coming toward us.
Great video! Would love to see more on the individual steps for each phase of this.
Video about generating logic would be absolutely awesome
It’s coming soon :)
it would be great if you can do a video about toolkits like ludu AI inside the Unreal Engine.
Damn, this was interesting to see lul. I never thought I'd see AI generated creation be into game developing unless it was for NPCs or enemies lol
Thanks so much for this opportunity. It is our favourite online classes. 🏭
Thanks! Out of curiosity - what online classes are you referring to? :)
Thank you for making this video! I have an interest in indie game dev and these tools can definitely help. Really appreciate your work, Rafal.
Thanks!
Its really interesting what nowadays AI can do, but still some of the AI tools requires to be "pushed" to receive good reply or model, but its capable to help much if you know how to use it. Before it was quite hard being as solo developer to do on your own everything, but now you can try asking for help from AI and it will actually help and if you don't understand, he can explain in different ways. Also haven't heard few tools that was shown in this video and will be nice to explore them. Great video Rafal! Also starting to wonder when we will see developed games with AI help and how people will be reacting to it :)
The thing these models look bad when you have freedom to look around them
Also you should have built the level with ai models, like 99% of this ue5 scene and game isn't Ai
I already can see textures not mixing together right when placed
Where are the shadows?
As usual has the uncanny "smooth" "smeary" AI look.
Yes! Please make a video about AI generation making logic in games with Unreal Engine 5!
Very impressive stuff, great algorithm pull! Had to give Meshy a try after this video, as I hadn't realized Text to 3D was already so advanced. Of course, the shovelware bloat is inevitable with the advancement of AI, but if even one artist can make good use of these tools, be it for themselves or for the gaming community, then it is well worth it, in my opinion.
Hi! Interesting devlogs and videos, subscribed here! Curious to see more on sharing experiences with Gameplay Ability tricks and suggestions.
Thank you!
The problem with this analysis is that it suffers from recency bias. It doesn't account for AI technology's ability to create even more detailed 3D models in the future, matching or surpassing the capability of a human artist. Right now, the human artist is considered the partner and AI is the tool, but one day, AI will become the partner and the human artist will become the tool.
the video in a phrase: a human touch, costs less than a human project.
The only problem I see with Meshy is the lighting that is integrated directly in the texture, it's why it's stand out so much when it's put inside a game engine. To fix it maybe do the texture yourself or in the img to 3D use an image of the 3d model you want with neutral lighting
This is awesome! I'm building a game in Rust/Bevy using Cursor. I'll be following. Keep it up!
Cool experiemnt. Did you work on cyberpunk ? I noticed the plaque in the background
Yeap :)
Story will always win. Team up with professional creative writers and you will remain relevant. Game mechanics will never trump story.
Awesome video, this is truly inspires
Ai is capable of matching your creativity if you are willing to dump the time into learning AI as a complete skillset. AI is iterative, and requires small clean up. It always will, just like real art. But refining what prompting works with what models etc, comes from time spent doing it. How good a model is depends on how clean and concise your prompt is, and how many iterations of a model you're willing to run through generating. The thing about AI is that it will never do the complete job flawlessly. Again, just like real art, it requires clean up. You will need to have SOME skills in order to make effective use of it.
this is awesome!!!
Thanks! ♥️
I don't know, man, I am currently making a project for uni and I kinda don't have time to learn to model and animate for it, so I went to the internet to download some stuff just so I can show something more proper there, which would all be changed when I wanna actually release the game. But using AI feels so damn empty, even if it gives the needed results. If a person did not do it, it feels kinda lacking in a way.
But anyway, good video for its purpose!
saddest thing is people using whole templates from UE or Unity and not changing anything and release it, looking rly bad and they thinking its gonna sell..
Does anybody know which page he is looking at in 1:15??
Thanks for this (subscribed). Maybe raytracing will help with the integration of the prop models into the environment. The model detail itself will improve over time in Meshy (I guess), so that the texture map will not look so high detail onto low poly, or maybe even eliminate the need altogether for UV texture maps.
Great video, very interesting stuff. I used to work in games development (over a decade ago now) so my opinions are probably not as up to date as some, but I feel like the current situation of studios growing bigger and bigger, making gigantic games over the course of years, simply isn't sustainable. And maybe AI is a tool that can step in to help?
If you have a real human artist creating textures - for example broken floor tiles - then can AI produce variations in a few minutes that might have taken the artist hours or days? Similarly, Can they use AI to fill an environment with variations of trees, rocks, and ruins originally built by a human? Each example like this might only save a modest amount of time, but repeat that thousands of times across the development cycle and suddenly budgets and timescales might become a LOT more realistic, especially for smaller teams working without the benefit of gigantic publishers footing the bill.
Thank you very much for this video. I'm trying to create my game solo, and it was more of a "learning" than "creating" experience. I don't know why I was ignoring AI tools all this time. Thanks again! And a quick question - what's the style or who's the original author of the illustrations like the one in 5:32? Love how it looks.
This video deserves so many more views! Great content and great quality, good job!
Thank you!
I'm less worried about the details and more wondering about the topology and efficiency. An artist will put the details in the right place because they understand how a model will be animated, where the players will be looking, and so on.
An AI will just throw random clumps of pointless geometry all over the place, eating up resources, and causing animators an absolute nightmare.
AI: It can do things. If all you want is "anything" - AI can probably do that. It can never do them very well.
meshy is realy cool, thanks
I want AI to be a great helper for anyone working in the field. It will not replace, but it can help with rapid development.
Looks like someone was heavily inspired by Dark Souls 2. Almost everything you said fits perfectly into the DS2 opening part
We will definitely see a fully developed game using prompts in 2025. Maybe 2026.But, we are there already. Same thing for AI based movies.
Keep making more videos like this!
1:14 what is that 3d model website called?
Meshy AI
I can bet my house that it didnt took this guy 2-3 hours to make that game. The dude is obviously a professional game dev, but stating that it took him 2-3 hours is just ridiculous.
If you know what you aim for and have enough skill and past experience, it's more than doable. If I can judge by different lens - with right framework you may set up a web app prototype in such timeframe, but you must know your tools 😉
Thank you so much for this inspiring video. I can see some individual developers are aiming to match top notch latest games provided by multi hundred million if not billion companies and that seems to have improved their output finely. If they would only compromise some high end quality for a more causal game they might be able to find some lucky millions around the corner. Candy crush for example made about a billion USD on its first year, it was a very trivial match making game which compared to some individual developers would probably be easily implemented with current AI tech and fully satisfied. The answer to the apparent question about AI capabilities does not seem to have a yes or no answer (on whether it can or cannot support in implementing games) but rather (how good exactly will the implementation be or how much can AI support in accomplishing)
I'm not sure, that we will always need a human touch. AI tech is not there yet and won't be for a while longer. But there have been massive improvements in really short amounts of time and there is no sign of the tech slowing down it's progress. While we might need the human touch for a few more years, that may change in a decade. What about two decades? Looking at where AI was just a few months let alone years ago and where it's now, that future where we could get high quality assets that are indistinguishable from an experienced artist in seconds may come sooner rather than later.
if I'm honest I'll only use ai for the art and that's it cuz I enjoy coding and solving bugs in the code because it gives me a feeling of accomplishment
did you tested some with the full paid meshy? the designs look great in the pics but i would like to see them in engine
Any thought about ai could help at the end of the dev to optimize a game? To push down the minimum reqs from the fancy x3d cpus and xx90 rtx cards to "peasant" level hardware, and you get reasonable performance+quality.....
can u make a video on ludus ai with unreal engine 5
I didn't even know that existed.
The assets created with Meshy heavily remind me of fables art style, especially the angel statue.
Those 3d models would be great for a quick greyboxing stage. I will have to purchase meshy now.
But why do you need AI-generated 3d models for greyboxing where you can simply drop in free premade models for the same purpose. One could say the AI models are somewhere in the middle ground between production quality and test dummy... using them is like lying to management that your game is in far more advanced stage than it really is, the only "reward" you will get is your salary cut because "it looks ready" to the morons in suits.
Can you add links to the tools you used in the video to the description plss
Done :) thanks for reminding me ♥️
@@rafalobrebskiThank you! ❤
Jakis kontakt do Pana mozna np. Email ?
obrebskirafal@gmail.com :)
Can you do this again but make it more indepth and explain the process?
Great video. Thanks for the insight.
jak stworzyłeś vfx? - 10:58
Really!!! No c coding knowledge required? I feel its too overwhelming? Does unity at all require c coding knowledge before to learn? For easier learning curve?
From one side 3d generation come a long way, and it's looks amazing from the technical standpoint, but still far from production ready. So as you said it's may be good tool for like prototyping. But same time how this differs from just buying premade 3D assets? And as with just using only premade assets before "AI slop" we had a lot of "Asset flip" games
The way it's going I'll probably learn to use AI to make games before I learn to code properly, which I've done a bit with python. I've got a project in mind that is specific enough I could probably give AI clear instructions for making it, but I lack motivation, and I'm not really sure how to start.
PLEASE do a video discussing how Ludus and even Chatgtp to create logic would work. thanks!!!
so glad youtube suggested this , great content ! ♥
Great video. Any tutorials/courses available?
I wish the AI tool can become much polish and advance. Such that even I don't have any skills related to game industry, I can still create my own game with my imagination. The concept is that AI will not replace our works, but people who learn advance technology like AI tool can replace us.
And why not get creative? or hire someone to do it?
While AI is seen as controversial, its benefit for making quick prototype ideas into gameplay cannot be understated, making use of this allows what could typically be a hundred hour endeavour to become a dozen hour mission to see if the idea you have is even worth investing. Of course though, it can also be used incorrectly especially when a company is run by someone who doesn't understand the requirements of a game!
That's it, I switching from UnITy, to UNREAL, Also I'm new here and for the first video I'm impressed
Can you do detailed how to on this
Tell this AI to create a game engine, come up with some different world and gameplay, invent a brand new art style etc. I'll see myself out.
it did, I coded a game engine with AI. It has PBR shading and works well. :)
@MartinDlabaja so you created another universe in this one with AI? Great, we are ready for the next level of gaming then. Btw sell it as Artificial Engine 1, I am sure people would pay billions for it.
Meshy looks great now, is this skeleton compatible with ue5 default sk? To use marketplace animations sets.
Thanks for this video!
It was quite straightforward to retarget :) I first use mixamo to auto rig, then retargeted UE5 skeleton anims 👍
@rafalobrebski Thanks for this tip, its really impressive.
jak na dwie godziny używając tylko AI jest to niepokojące, jeszcze długa droga żeby ludzie zostali zastąpieni przez AI w game developingu, ale dalej jest to niepokojące
Can u make video about using ai to make animation video same way you did
How to create mobile game?
Yo Rafal! That was an epic journey into the AI game dev world! 🌟 I was ready to slay some dragons but ended up vibin' with some serious creativity instead! 🎮🔥 Your visuals were on another level, and the AI antics had me laughin' like a toddler with Legos! 😂 Keep up that fresh approach, 'cause you've got somethin' special cookin'! Can't wait for the next one! And just remember, team up with AI but don’t let it skip leg day! 💪💖 #ReactionQueenApproved
This answer was generated automatically?
I've tried to use ChatGPT to get me started with blueprints but it will often tell me to create nodes that don't exist. I assume what you would need to do is scrape the unreal blueprints documentation into a json datasheet and finetune a LLM with that data.
There are a few tricks you can try - for example tell ChatGPT to „only rely on actual documented nodes” and tell it „if you do not know the answer - look it up in documentation but do not guess”. Another thing I noticed is that once you correct it a few times, it starts getting things more accurate
There are a few tricks you can try - for example tell ChatGPT to „only rely on actual documented nodes” and tell it „if you do not know the answer - look it up in documentation but do not guess”. Another thing I noticed is that once you correct it a few times, it starts getting things more accurate
@@rafalobrebski Ooooh, i will try this. Thanks!
What also works for me is:
Copy paste the part of the documentation to a separate chat and solve the problem there. That is the most accurate practice I've found so far. You can also upload pdf documentation if it has one...
This is insane for early prototyping, please do a video on ludus ai
Love the video