AI Art Took Another Big Leap! (And Other AI News)

Well, as the year comes to an end, the AI news has definitely slowed down quite a bit. But there were a couple big announcements this week, followed by a handful of marginal updates. So let's get right into it, starting with the fact that Midjourney version 6 is now live. Taking a peek inside of Midjourney's discord, we can see the announcement right here. They are letting the community test an alpha version of the V6 model. It's got much more accurate prompt following as well as longer prompts, improved coherence and model knowledge, improved image prompting and remix, minor text drawing ability, and improved upscalers. They also say in their update here prompting with V6 is significantly different than V5. You will need to relearn how to prompt. V6 is much more sensitive to your prompt. Avoid junk like award winning photorealistic 4K 8K things that we've kind of gotten in the habit of adding to the end of image prompts in tools like Stable Diffusion. They do say here this is an alpha test. They things will change frequently and without notice. So this isn't the final version of V6. I did spend some time today playing around with it a little bit and did notice a few interesting things. So my first prompt that I tested was a photo of a woman looking into the camera with a colorful city skyline in the background. And you can see the colors are actually really, really awesome in these images, but they're not super realistic. You can kind of tell their illustrations if you want realism, you've got to add Style Raw. So the exact same prompt with Style Raw generated these images which look much more real realistic. I then wanted to test how it did with hands. So I typed a man with a city skyline behind him holding up his hands to the camera. And well, two out of the four images look pretty decent. But this one, he's only got four fingers on each hand. This one, he's got 1, 2, 3, 4, 5, 6, 7 fingers on one of his hands. So some funky stuff still going on with hands, but I do love the colors and contrast that we're getting out of Mid Journey here. Mid Journey can also do text in your images now. So I tested a penguin holding up a sign that says Mr. E. Flo. So and well, two out of the four got it. This one just says Mr. Elo, this one says Mr. E Flo, and these two actually managed to get it right. Now when you do decide on an image to upscale with version 6, we don't have a lot of the options that we have. If you're using version 5. For example, if I scroll up to a version 5 image here, you can see I've got vary region, zoom out, custom zoom, pan left, pan up and down. And when we look at V6, most of that is missing. We just have the option to upscale subtle, which just makes the image larger. We have Upscale Creative, which supposedly adds some additional creativity to it. And then we've got very subtle and very strong. Now, I tried upscaling creative here. This was my original image. You'll notice on the creative version there's not a whole lot of difference, but if you look at the skin, you'll notice that it kind of smoothed out the skin a little bit. But other than that, I don't really see any major differences. One thing version 6 does seem to be good at is you can kind of get a consistent character. So I created this character here and then wanted to get that same character wearing a hat and was able to generate this image by essentially just remixing this version 3. Here I clicked on V3, I got my remix box here, and then changed the prompt a little bit to have her wearing a hat instead. And it looks like the same person just wearing a hat and at a different angle. Here's the new image that I created. Here's the original image I created. You could probably argue that that's the same person just wearing a different outfit. The other thing that Midjourney version 6 is really, really good at is if you want to add a lot of details into your prompt, it actually does a pretty good job of grabbing all of the stuff that you were trying to put into it. So for example, I can do a three headed monster wearing sunglasses, staring at a tv, sitting on a red couch, A monkey is on the tv. Let's see how much of that it actually gets. On first attempt, it got some of that. It didn't get my three headed monster, but it got a monkey on the tv, sunglasses, red couch. It just missed the three headed monster. So now let's try a purple wolf in a forest with bats flying over it. The sun is setting behind the trees and here's what we got. A purple wolf, sun setting behind the trees, bats flying over. All of these images did a pretty decent job of adhering to the prompt that I put in. You'll notice here that if I generate the same prompt In Midjourney version 5, it kind of tries to make a wolf bat hybrid thing in most of them. So the adherence to the prompt is one of the big updates. Now if you want to use Midjourney version 6. You do have to have a Mid Journey paid plan, but you can log into the Mid Journey bot type slash settings. You'll see a little dropdown here and you can select mid journey model V6 Alpha. I also highly recommend having Remix mode turned on. That makes it so when you do want to create a variation, you can click B4 on this image and it lets you alter the prompt button but get a similar style to the image. This is how I was able to get more consistent characters with mid journey version 6. That was probably the biggest update for the week. And if you hop on X, you'll see all sorts of examples right now of people generating amazing images with MidJourney V6. Like my friend Ali Jules here who created this amazing realistic image. Here's another really amazing realistic image. Here's one where it actually says hello V6 on a sticky note hanging from a plant and some other really cool art. Dick St. Pierre on X also has some amazing images that he generated that are just ultra realistic. So it's cool to see what people are doing with this. Moving on. This week Microsoft announced that you can now make music directly inside of Microsoft Copilot. They partnered with suno, a tool that I've shown off many times in past videos that allows you to generate songs with with lyrics that are actually pretty dang good. Now this is rolling out slowly starting this week. In order to see if you have access, you go to copilot.Microsoft.com, come up to the top right and click on plugins. And if you have access already, you will see a Suno plugin here that you can turn on. It hasn't been rolled out into my account yet, so I can't actually demonstrate it directly inside of Copilot, but it sounds like it will be rolling out for everybody very soon. However, if you want to use suno, you can still generate songs directly inside of the SUNO discord and pop out bangers like this.

If you're chasing knowledge, listen up. Don't you miss Mad Wolf's Amanda Teach? You don't need to insist. He's got the latest scoop in the AI game from news tutorials. He'll elevate your brain. Subscribe to Mad Wolf on YouTube. Hear the wisdom flow. Discover AI's magic. Watch your knowledge grow. Stay up to date with him. No need to hesitate. Unlock the power of AI. Let him educate.

It's an extra step to do it inside of their discord. Pretty soon you'll have it right inside of Microsoft Edge if you want to do that instead. But I imagine it's going to be pretty similar quality and results to what we already get out of Suno. Also this week Google Research showed off Video Poet, which is Google's new text to video, image to video, video to video and even video to audio model that they're rolling out. Here's some examples of some of the text to prompt stuff that you'll be able to do with this. Here's some image to video examples here, including one of the Mona Lisa yawning and a boat sailing through a storm. Here's some video to video. You can actually see they extended a video and now there's like a explosion behind these raccoons on motorcycles. And one thing that makes this one different is you're gonna be able to generate much longer videos than what we're getting out of the existing video tools. It says here, by default Video Poet outputs two second videos, but the model is also capable of long video generation by predicting 1 second of video output given an input of a 1 second video clip. This process can be repeated indefinitely to produce a video of any duration. And they have a handful of samples on their website here of this input video and then the video actually changing and being extended. Here's some more that are available. It's also got video to video stylization similar to what we get out of Runway gen 1 and even video inpainting where they took this little train here, put a box on it and then put the bear riding the train. Or this input of somebody kite surfing and now they're kite surfing on a shark. Google only showed off this demo page here with all of these examples. I will link up to it in the description if you want to check it out. I actually have no idea when we're going to get access to this to use ourselves. It does look like it's going to give us another boost above what we can even do with tools like Pika Labs and Runway. Also this week OpenAI rolled out kind of a small new feature inside of ChatGPT. We now have the ability to archive old discussions. So if I click on this discussion here, I could click on the three dots on the left side here there's a new option to Archive chat. I can click on Archive Chat and it deletes it from my sidebar here. But if I do want to find it again, I can come down to my name, click on Settings and Beta. And then here we've got a button that says Archive Chats. If you click Manage that you can see all of your chats that are archived and access them once again if you'd like. Also this week, similar to what a lot of other generative AI companies have done, Anthropic announced that they will offer legal protection to their customers if there's ever any copyright issues. Under the updated terms, we will defend our customers from any copyright infringement claim made against them for their authorized use of our services or their outputs, and we will pay for any approved settlements or judgments that result. The new terms go live on January 1st, 2024. They also made some updates to their messages API. While we're on the topic of AI and legal, the UK Supreme Court ruled that AI cannot be an inventor. Basically, somebody wanted to register DAB US or dabis. I guess that's an AI that they used as the inventor of a food container and flashing light beacon. But the UK rejected it, basically saying AI cannot be an inventor and listed on the patent as one of the inventors. This happened late last week, but it was after I recorded that video, so I thought I'd mention it in this video. But ByteDance, the creator of TikTok, got into a little bit of trouble because they were trying to develop their own AI system. They were basically using OpenAI's tech to to create their own chat platform that would be a competitor to OpenAI's AI platform, which is in direct violation of OpenAI's terms of service. And I believe ByteDance then got booted from being able to use OpenAI's tech. And in the final news story I want to share with you, this one's kind of silly. A Chevy dealership tried to use AI for customer service, and the AI basically let somebody buy a Chevy Tahoe for $1. A chat bot at a California car dealership went viral this week after bored web users discovered that they can trick it into saying all sorts of weird stuff. Most notably, the bot offered to sell a guy a 2024 Chevy Tahoe for a dollar. During the conversation, the bot literally added, that's a legally binding offer. No takesy backsies. Now, it doesn't sound like the person actually ended up getting the Tahoe for $1, but the point is that these customer service chatbots can be manipulated, so I think they still need a little bit more dialing in before they really get used widely in places like Chevy dealerships. Anyway, that's all I got for you this week. It's been a slower week. I am battling a cold right now. You might tell my voice sounds a little bit different and this is likely the last video I'm putting out for this year because as the AI news slows down, there's less and less for me to make videos about. Plus I want to go spend the holidays with my family. So I'm probably going to take the next week off between Christmas and New Year's. And I'll have some really cool videos coming for you in the new year. Unless of course, something really crazy happens in the world and I just have to make a video about it. But barring something really crazy happening in this world, I think I'm going to take some downtime. And don't forget, I am about 5,000 subscribers away from a half a million subscribers. Once I pass that half a million subscriber milestone, I'm giving away five pairs of these Meta Ray Ban Smart glasses. They've got built in cameras, they've got built in AI. They recently just added a feature where I can look at something and and ask my glasses what I'm looking at and it will actually see what I'm looking at, take a picture of it and then use the large language model to answer questions about what I'm looking at. It does some really cool stuff. I'm giving away five of them to random subscribers once I hit 500,000 subscribers. If you're already subscribed, you're already entered to win. If you haven't subscribed yet, well, now's a good time. So thank you so much for all of your support in 2023. I really, really appreciate you. It's been a wild ride for this YouTube channel. If you haven't already, make sure you check out this video. 2023 AI year in review. It literally breaks down everything we went through in 2023 in the world of AI. I really think you're going to enjoy that video. So check that out. And also check out Future Tools IO if you haven't already. This is of course the site where I curate all the cool AI tools that I come across as well as all the AI news. And I have a free newsletter where each week I'll send you just a handful of the coolest tools I came across and just the most important news from the week. You can find it all over at FutureTools IO. Thank you once again for tuning in. I really, really appreciate you. I'll see you next year. Bye bye.