Mind-Blowing AI Video Tools You NEED to See

there's a brand new tool for creating consistent Styles inside of mid Journey soon animators will be able to animate their subjects using only an audio track Google released a brand new image generation tool but is it better than other tools on the market and we curated an AI Art Exhibit that features work alongside Banky and Andy Warhol welcome to your AI film news of the week I want to kick things off by playing a game with you today I'm going to show you a clip and I want you to try to identify what inside of this scene is not real so let's go ahead and watch the first clip and now let's watch the second clip were you able to figure out what was composited in that scene well in the first clip it was a light pull and in the second clip it was the boom box on the Shelf now the reason why this is important is it came from from a white paper that basically is all about how you can easily add objects to your scene now traditionally objects are added into video using a process called compositing and compositing is a skill that takes a long time to develop but it seems like AI tools are now allowing us to drag and drop assets into our scene and it takes care of a lot of the compositing process automatically now I should note that this source code is actually available for you to download right away but I think that whenever this specific tool set or tools like it are integrated into compositing or video editing applications it's going to be an absolute GameChanger next up we have a brand new feature inside of mid Journey that I am super excited to show you so the tool is called style references and it's a little different than image references so for example let's say that we came across this image and we wanted to create other images in the sty style of this specific image now assuming we did not prompt or create this original image we may not know how exactly to get a result that looks like this well using the new style reference tag inside of mid Journey you can actually do it very easily I'll show you how it's done so I'm inside of a channel inside of Discord the mid Journey bot has been invited so let's go ahead and create our image I'll do slash imagine a dog wearing a hat and to use the style reference tag we just do D- sref so sref and then space and now we need to paste in the URL to our image so I have our image here I'm just going to go to the top copy it and we will paste it directly inside of Discord now if you want to pull from multiple image Styles like let's say there's two different images that you want to pull from all you have to do is put a space and then paste the next URL I think you can add up to three images to pull the styles from now you can also add the amount of weight that you want your overall styles to pull from so by default the style is 100 in weight but if we wanted to change that all we have to do is do-- ssw space and then you put in a number it could be less than 100 it could be all the way up to 1,000 it just depends on the amount of style that you want to pull from your reference image in our instance I'm just going to keep it to default so I'll just say the image itself as a style reference and we will click enter I should note you need to make sure that you are running the version 6 version of mid Journey or above if you run it with 5.2 for example this will not work okay so let's take a look at our results here we have this dog wearing a hat and all of the photos look pretty darn good and to put that in perspective this was our original uploaded reference image now you might be thinking well doesn't this mean that we could upload an image in in a certain color grading style and generate multiple images with that same style of color grading and the answer is yes let me show you how it's done so we have this image in this very specific style from the internet you can see it has this green and pink color grading there's some fog kind of these light leaks it's a very specific style that might be hard to prompt again and again so I'm going to go ahead and grab the URL for this image so we'll say a cinematic still of a woman in a western film we'll do -- sref and then paste in our reference image and then I want to add in a little more style weight so I'll do --w and we will do 200 in this instance and we want to make it 16 by 9 so we'll do D- a 16 colon 9 so that it is a widescreen format and one of my favorite tags d-r space 10 to repeat the prompt 10 times so we'll get 10 different versions of this prompt and because I have one of the more advanced versions of mid journey I can actually run those 10 prompts all at the same time so it's super fast so let's go ahead and run this and click yes so our images have been created and you can see that they all kind of have that same color grading style applied to them now obviously there are a few notable exceptions where it really pushes it a little far but for the most part they have that same color grading aesthetic which is really helpful if you're trying to get the same style again and again so I think that if you want consistent Styles inside of mid Journey this is a very helpful tool and it beats the heck out of the style tuner Google also came out with the second version of their image effects image generation tool this last week and the results look pretty dang good so you'll find a link below this video so let's generate an image here I'll say a curious dog finds a red ball at the beach on 35 mm film so let's go ahead and generate the image so we have the these four Images here of this dog that basically found this ball on the beach and they look pretty darn realistic especially this last one here so I'm going to go ahead and download that image so again this is our image from Google looks pretty realistic this is the exact same prompt inside of Dolly and while the composition is interesting and the dog definitely looks curious It's a bit more expressive it doesn't quite come across as realistic and this is the image that was created inside of mid Journey very very filmic now the dog doesn't kind of have that curious face but he is checking out the ball and it generally is the exact image that we were prompting for so you can see there are different styles depending on the image generation tool that you choose and that brings us to our game of the week so I'm going to show you four different images of a Renaissance painting featuring a man wearing a VR headset and I want you to tell me which image generation tool it was created in I used mid Journey Dolly Google and Leonardo so let's take a look at the images here's number one here's number two here's number three don't count the fingers on this one and here's number four if you think you know which image generation tool was used to create each specific image let us know in the comments below the first person to get it right will win an AI film making Club sweatshirt from Curious Refuge a team at Apple shared a white paper where they're using language to actually Photoshop images it's really incredible and I think it really does highlight the future of photo manipulation so the research model is called mg e and basically the researchers are able to type in instructions and change the output of certain images so they have some fantastic use cases over on their website here's a picture of some glasses the researcher said make them red and the result was these glasses in this instance they said turn this image of Central Park into the Vatican and it basically did the tool is also helpful for practical image editing so we have this image of a skier and we basically can say edit out the skiers on the right and the skiers are taken away and it even does color GR so in this image we can say lighten out the yellow tone and it gets rid of some of the yellow tone there's also a hugging face demo where you can actually try this tool yourself there's a link below this video so I uploaded an image of me and in the instructions I said make him Dumber so I think we're going to see a lot of the make it more memes begin to pop up with this new photoshopping tool and I went ahead and hit submit so the image generation time does take a while so I'm just going to show you the final result here here so this is the image that it created and it is absolutely terrifying but also kind of hilarious there are more examples of this tool in action over on the white papers website it's a really interesting glimpse into the future of image manipulation some researchers from China created a facial animation system that only needs reference audio to create the animations on the faces it's similar to The Meta tool that we demoed about a month ago but this one has much more realistic face animations so there's some pretty interesting examples over on their website I should note that this example that I'm about to show you was generated with just audio alone there's no video driving this animation star so obviously this is going to have a bigger impact on video games in the immediate future but it is conceivable that in the next couple of years Animation Studios might be using tools like this to create scratch animation bases for their animated characters there's a really interesting example where they had various actresses singing the Frozen Let It Go song and the face models were animating [Music] automatically and funny enough we have a Disney animator that worked on Frozen inside of our workshop and they were saying how this tool does look interesting and that in the future we may be using tools like this there's also a brand new image to video model that is completely free from China it's called dynami crafter basically you just upload an image you type in a prompt and then it generates a video there's plenty of examples of the videos they're only about 2 seconds long now so definitely a lot of room to improve but I did think that some of the results look fairly realistic now you may be wondering how this video generation tool compares to Pika and Runway well I got you covered so here is a video that was generated from an image of a cyborg using dynami crafter you can see that there's some weight behind the robot step and his arm moves back it feels pretty realistic so here's the result we got from P Labs you can see that there's some Distortion in the chest of the robot there's that very classic Pika slow push in and I should note that I did prompt for a robot walk walking and this is the result that we got now the same thing over on Runway prompting in a robot walking and we can see that it starts out kind of interesting and then it just decides to go out of focus and the robot kind of looks like his weight is Shifting in a weird way so actually in terms of the overall animation of the subject itself it seems like dynami crafter actually is doing a better job but in terms of cinematic realism I think that Runway still is probably a better choice I'd also like to extend a thank you to the pop and Contemporary Art Museum in Talon Estonia we actually had the pleasure of curating an AI Art Exhibit and we were able to feature dozens of artist inside of the exhibition now the exhibition is going to go on for the next four to 5 weeks so if you're in the area I highly recommend checking it out it was really awesome to be able to feature incredible AI artists and students inside the Curious Refuge program inside of this exhibition I think it's incredible that their work is being featured alongside Banky Warhol and some of the most talented artists of the last century the European Union also decided to bring forth legislation to put limits on generative Ai and the use of artificial intelligence inside of society some of the highlights basically you're not able to use AI to do emot recognition inside of the workplace or in educational institutions you can't do social scoring based on artificial intelligence and AI systems are not allowed to manipulate human behavior to circumvent their free will now most of the points inside of the legislation totally make sense and are good for society however there are some gray areas that will definitely be argued about for example AI algorithms already power most social media platforms and it could be argued that those platforms do interfere with your free will by making them more addictive so how will this new legislation affect old systems and algorithms that were already in place only time will tell we also came across a text to 3D model tool this last week that is probably the best one that we've seen to date they have some great examples over on their website so this is a fullbody lion you can see that there's fantastic Fidelity in the texturing it looks really good we have this alligator reptile man character looks really interesting we have this axe here that looks borderline game ready it really does have enough Fidelity for some of the lower resolution 3D games now to use this tool it's actually really really easy all you have to do is log into your meshy account you get 200 credits completely for free which is awesome and we'll do text to 3D and earlier I typed in a dog wearing a hat which I think is absolutely fantastic and it gave us this quick result here now you can see that the mesh is a little low res right off the bat but all you have to do to refine it is just hit the refine button and here is our refined dog wearing a hat and you can see that it looks pretty darn good he even has his tongue sticking out out there there's texturing on the ears the Hat has a different texture and it really is a good start and so as we've talked about on this show text to 3D models really is a very important part of the future of 3D asset creation I'd also like to wish the best of luck to everyone who joined us for the February session of our AI film making course and for everyone who signed up for the beta session of the AI advertising course the AI advertising course sold out in less than 24 hours but will have a bonus enrollment at the end of February you can sign up for the wait list to be the first one to know when the courses go live meta also came out and said that they are going to start labeling AI generated images on their platform so that includes Facebook Instagram and threads it'll have a little tag that says AI info where it basically gives people more insight into how the image was created the team at ROBLOX is using AI to translate live chats basically in real time now the next step will be able to translate the actual voice in real time we've already seen announcements like this with the Samsung Galaxy s24 and I can imagine that the future of Competitive Gaming online will likely feature voices in your native language which knowing how toxic Gamers can be might be a good thing might be a bad thing only time will tell a researcher this last week took a look at all of the open-source language models and determined that smog 72b is actually the best and it outperforms open ai's chat GPT 3.5 the reason why this is important is this language model is open source you don't have to pay money to use it and hypothetically it will be free forever to use this specific tool now if we're going to do a quick comparison using the MML test as a reference chat GPT 3.5 got a 70 smog got a 77 and chat gbt 4 got in 86 so the open source tool is really interesting but it's not quite the industry standard at this point we also came across this really interesting demo of someone using Luma lab's AI 3D model generation tool on the Apple Vision Pro and the cool thing is they were able to generate the 3D model inside of the Apple Vision Pro download it and then integrate it into their actual room so they're able to look at the 3D model look around grab it move it around position it and it's pretty obvious to see how using a tool like this in the future will allow 3D modelers to literally model as if they are a sculptor and seeing the runway telescope Magazine on aars table there reminded me that Runway was super kind it sent me a copy of telescope magazine in in the mail I totally missed out on grabbing the magazine and it sold out online so thank you so much guys for sending this along it has fantastic stories and artwork and some of the best graphic design I've seen in any magazine let alone in AI generated one so I highly recommend trying to grab a copy when the next batch goes on sale Google's Gemini Ultra will be dropping this week hopefully by the time this video is live you should be able to start messing around with the model we'll see if it's better than chat gp4 next week but I have some High Hopes because Google Gemini Pro was already pretty dang good now Google's marketing says that Gemini Ultra is better than chat gp4 but we'll let you know for sure next week we also saw some news from Microsoft where they said they will train 2 million people in India on using artificial intelligence this is a huge investment inside of the workforce inside of India and we've had many curious Refuge students from India inside of our program that have gone on to have awesome opportunities using the skills they developed and that brings us to our AI films of the week I want to kick things off by showing you this project created by JC photo feed it's called the cyclist and it really is one of the most high-end 3D animation projects that I've seen up to this point they actually created this project inside of our course and when we saw the submissions we were absolutely Blown Away the sound design the animations they really do look very professional so I highly recommend checking out the film there were also a ton of really awesome submissions inside of the runway gen 48 competition that happened just this last weekend and I came across this one called KO by EEG it was a really fun film it basically is kind of in the style of studio jibi and it has that kind of charm to it and it has this very Whimsical rural Japanese vibe to it and a kid basically gets a VR headset and starts going into a bunch of different worlds so it's really cute and I think they did a fantastic job the next film I want to highlight is called come and take it by Taylor Graham the film features some really impressive curation I love some of the shot selections inside of the project in many ways it feels like a high concept fashion film which is really interesting especially considering they put it together in less than 48 hours and our student film of the week is be yourself by Renee Seuss Renee put together a basically a fashion concept that has some really interesting shots in curation in it and the cool thing is I was talking with Renee and he said he published his assignment online and over the last 24 hours he had three job offers from this specific project so I highly recommend sharing out any homework assignments or projects that you work on because you never know what opportunities may open up and our last story of the day comes from archaeology where some researchers were able to decipher a scroll that had been damaged by a volcanic eruption at Mount Vesuvius and they basically used artificial intelligence to read the scroll and do you want to know what was inside of the scroll what this ancient wisdom was well drum roll please basically they said scarcity does not inherently make something better thank you so much for watching this episode of AI film news of course you can subscribe to get AI film news directly to your inbox by clicking the link below this video and you can subscribe here on YouTube to get notified when we drop a brand new video or tutorial thank you so much for watching we'll see you next time

Channel: Curious Refuge

Views: 23,169

Keywords: AI News, AI, ai generation, apple vision pro, google image generation, Runway, AI Video, AI Film News, AI Film, AI Filmmaking, AI Advertising, midjourney style reference, consistent color grading, eu ai laws, new video generator, ai film news, AI Animation, AI Apple Vision Pro, Google Image Generator, Google Gemini, Banksy, Warhol, AI Art Museum, AI Photoshop, Style Reference, Style Reference Tutorial, AI 3D Models, Create 3D Models with AI, Curious Refuge, Caleb Ward

Id: kBWAOAEkin0

Length: 21min 41sec (1301 seconds)

Published: Sat Feb 10 2024