OpenAI FINALLY Added Video to ChatGPT & More AI Use Cases

Wow, what a week for AI. We had more innovation in this week than probably in the past two months combined, at least when it comes to the biggest players releasing new tools. On the one hand we have OpenAI continuing their 12 days of OpenAI Christmas with some new announcements that we'll be reviewing here. But then on the other hand, Google made their massive swing with Google Gemini 2.0 that can see your desktop and your camera. Also we had the Sora Release and Devin AI finally dropped at $500 a month. All this and so much more. In this week's episode of AI News, you can use the show that reviews all the pract AI tools that have been released this week, filtered for the ones that you can actually put to work today. Let's get into it starting with 12 days of OpenAI. So I'm recording this on Thursday night. So we had four days of releases so far. On Monday we had the Sora release. On Tuesday we had ChatGPT canvas rolling out to free users and upgrades being made to it. On Wednesday we had the Apple Intelligence integration which was a little underwhelming. And first day we had them ship the camera and desktop recording access of the advanced voice mode Big. So let's talk through these chronologically starting with Monday's announcement. The Sora release Let me tell you, the reception on this release was not good at all. A lot of people are mad at the price tag because as I pointed out in a dedicated video on the Sora release. So if you want to learn more, then check out the video, we'll link it on screen right now. But the main issue there is, as I pointed out in my release video, the $20 plan is just sort of a preview. You have 50 generations per month with watermarks. You don't get to make full HD files. And not just the limitation, but also the capabilities limited because you cannot do image to video of humans. Meaning that if you really want to use sora, it costs $200, which is absolutely a crazy price tag. And yeah, it is an unlimited plan. So if you use unlimited generations it can be a great price if you're a company that puts this to work. But as I've said many times, most people don't need AI video yet. Plus they were facing a lot of issues with signups. Like heck, I still didn't get to create an account and I'm paying for the pool subscription at $200 a month and I wasn't even able to use Sora yet. Bit of a mess there. I still stand behind the opinion that they nailed the platform and the video generated, while maybe not being state of the art, is really good. Okay, again, if you want more details on this, check out the separate video that I made just about the Sora release. Moving on to Tuesday, we had ChatGPT Canvas releases. And not just that, they also overhauled the interface a little bit. So if you head on over to my computer here, you can see that there's a new button here where you can use various tools. And these really have been piling up. Not everybody was aware of all the things hiding in here. I mean, there's a code interpreter behind all of this that people don't even know about. I guess that is down here. The point is that they now upgraded the Canvas, and not just that, they gave free access to all users, even without a paid account, which I think is fantastic because Canvas is one of my favorite features in here. But they also added the ability for ChatGPT to run Python code, meaning if it writes a little script or if it uses some package, it can actually run the code. And if there's bugs, they pop up in the terminal and you can debug right from this web interface. No need to copy paste this into an IDE that would need to run the code. For anybody working with Python, this is absolutely amazing. And while it doesn't match something like artifacts inside of Claud yet, that can also run HTML and therefore build complete websites like dashboards or portfolio websites and many more, this is definitely a step in the right direction. If you haven't tried Canvas yet, well, now you can do it on the free plan too. And I would say this, especially if you're a student, the ability for ChatGPT to go in and to comment on things like so, and then you can just apply these edits to the comment. By the way, this also works if you're doing it for creative writing, like essays or research papers is one of the most powerful ways to work with this. Because if I write an essay about penguins in Canvas here and I want to edit something about it, well, I don't need to copy paste in something like Word and then bring it back into here. No, I can make my edits right in here. And maybe even more importantly, I can use the AI assistant I have here with ChatGPT surgically by selecting certain parts and prompting on top of it, rather than prompting on top of the entire text, which is more akin to hitting the entire text over the head with a mace or something like that. So those are Canvas updates. Next day we got these Apple Intelligence launch and this was more of an announcement that, hey, this is available in the iPhone now and if you're in the US you can make your Siri call ChatGPT. I saw one comment on this video which essentially said that this announcement could have been an email. And I kind of agree. They did show off how this works and how you can upload files through Siri now. But yeah, I would just say if you're not a Siri user already, just use the ChatGPT desktop app for this. And we had this for months, so not the biggest of news. And then today on Thursday, we've got something that has been long overdue. The ability for advanced voice mode on the phone to actually see through the camera. Let's check out our clip here and.

Then once it connects, I'll share my video. Hey, Chat, how's it going?

I'm doing great, thanks for asking. I see you're wearing a Santa hat.

I am. And do you see what I have in front of me?

Yes. I see a coffee setup with a kettle and a dripper. Are you planning to make some coffee?

I'd love to. Do you think you could walk me through the steps?

Sure, I'd love to. First, place a filter in the dripper and rinse it with hot water to eliminate any papery taste.

I think this is pretty amazing, especially in the light of the fact that yesterday Google announced Gemini 2.0 with similar capabilities and they showed off a preview of their project Astra, which is essentially this. It's a voice assistant that can use your camera. So while Google did release their new model that can see your screen and use your camera, they have not released this full application. This assistance that can use your camera and OpenAI took less than 24 hours to actually update their advanced voice mode, where the voice assistant, by the way, is just way better. I mean, in the Google video I mentioned that maybe the interruption on Google's product works a bit better, but everything else, the quality of the voices, all the styles that it can emulate is just better about advanced voice mode and if you're on A plus teams or pro plan in the us this should be rolling out to you as soon as possible. As this just came out a few hours ago. This has not rolled out to me or anyone on the team yet, but trust me, we will be following up with coverage of this either in next week's news you can use or a separate video because this is a really amazing feature that everybody has been waiting for. It's a true assistant that can actually see and it works for your phone, which just gets my mind spinning. This obviously opens up a whole new bag of use cases that are possible here. Very cool. And then in addition, I guess I should mention they also have a Santa Mode which all ChatGPT users have now that immediately rolled out to everybody. Bring a little bit of the Christmas spirit into the ChatGPT app. And that's it from OpenAI for now. We'll be following this as there is six more days of 12 days of OpenAI. Most of the things that were expected have been released by now, so I wonder what next week holds. I guess you'll just have to tune into next week's episode of AI News you Can Use and I quickly just want to point out that today we're actually celebrating the one year anniversary of AI news you can use. I started this show last December 9th and have not missed a single Friday. And at this point I just want to say a big thank you to the entire team that works on these videos. Robert Lucas for editing this, and Daniel and Ariadne helping with the research and preparation for all of these. Every single one of these episodes takes somewhere from 30 to 40, 40 hours of labor. If you add up all the different people working on it and we've been doing it for years straight every single Friday. We didn't miss a single one, we didn't upload on Saturday a single time. I think that's quite something. And thank you to all the viewers of this for tuning in regularly. This wouldn't be possible without you. And with that being said, let's move on to the next piece of news that you can use. One really exciting thing we're seeing in AI happen right now is this rise of AI agents and people throw this term around a lot. So I just quickly want to clarify what an AI agent really is. According to me, it's a program or a system built with an LLM in the background that carries out actions on its own. It's really not more complicated than that and obviously if executed well, agent like this can be very useful as it can take over your work. And this is where I want to introduce you to Lizard Agent Studio. Created by the team at Lizard AI, the sponsor of today's video, this platform takes AI agents to a whole new level by making it easy for anyone, yes, even non technical users to build and deploy custom AI agents in just minutes. Lizard Agent Studio is an all in one workspace for building AI solutions that are custom tailored to your exact needs so for example, you could automate customer support, optimize existing workflows, or even build tools for HR and finance departments within your company. And one interesting thing is that these agents don't just solve a task, they also evolve over time as they get used. And this is really what sets Lizard apart from a lot of the competition. These agents evolve as they are used and they learn all who have focused on safe and responsible AI. Lizard Agent Studio is built on the Lizard Agent framework, the only agent framework that natively incorporates safe and responsible AI into the core agent architecture. It even includes modules where you can toggle these features on or off based on your needs. So if you're just getting started, you can use these pre built agents for various use cases like social media automation, knowledge search, sales outreach, newsletters, and competitor analysis. Or you can even build your own custom agents from scratch, all without needing to write any any code at all. Whether you're a developer looking for full stack control or a decision maker trying to scale enterprise automation, Lizard gives you the flexibility to achieve your goals securely and effectively. So go check out Studio Lizard AI from the link in the description below and start building your very first agent for free today. A big thank you to Lizard for sponsoring this video and let's get back to the next piece of AI news that you can use. Okay, next up we have a bunch of announcements coming out of Google and actually they had so many announcements and most of them you can't even use today, with one exception which is the new model that can actually see your camera, see your screen and you can interact with it. Anyway, there was so much there that I had to create a dedicated video just on the topic. So I'll put a link up on screen right now and there's also a link in the description. But the gist of it is what I pointed out here. There's a brand new model that can see your screen, use your camera, it interacts in real time, it can input audio and video in real time and output audio in response to that extremely low latency, just like advanced voice mode. But hey, this thing sees your desktop so it can assist you. And then Google built a bunch of these experiences on top of that which they gave different names like Project Astra and Project Mariner that either uses your phone as an assistant or takes control of your browser to get things done, all based on this new Gemini 2.0 flash model. Again, if you want all the details, check out that video. It's actually quite interesting. And this experience of sharing your desktop and Working with the assistant is available today, and we try it out in that video. Okay, on to the next story, which is Midjourney Canvas. And this just got announced, and it's just a new thing that I quickly wanted to show you here. And it's basically this new way of working with Midjourney. The focus of this is entirely on world creation. So characters, locations, atmosphere. They call this patchwork. And it uses the power of the Midjourney generator and the power of consistent styles with the Style reference feature for you to generate a plethora of assets all within one board. So here, right here, we have Amara Vihan, a resilient and resourceful historian obsessed with unraveling the mysteries of Simulate Krumhalo. And then it just generates a bunch of images of her to represent that. So, for example, if you're writing a book and you need visuals or you're creating a comic strip, this is a fantastic way to tell a story, to build a world with multiple characters, and it's all coherent, right? You set your one style reference, and everything happens in a style. Same thing for locations. Look, a dilapidated opera. I can't even read that. Let's move on. In the video right here, Zaina finds the weird abandoned theater in the woods, and now it combines the character with the location you created previously, all bringing it together, making it easier to tell a story. I find this quite interesting as AI video is moving forward, and we're getting features like Storyboard Inside of Sora. Mid Journey is also moving forward, and you can try this out today. If you have a Journey subscription, that is. Okay. Next up, we have Devin finally launching, if you remember this story, this went Giga viral on its first announcement, and today it's finally available. I think the big headline that everyone is talking about is, hey, it costs $500 a month, which is crazy to most people. It's probably the most expensive AI tool that we have seen yet on a subscription basis. I mean, heck, one week ago, ChatGPT Pro with $200. Now Devon with 500. With these subscriptions, they're out pricing most of the world. So that presents a question. Is this even worth it? Because with ChatGPT Pro, that's a whole separate discussion that we could hold, but for most people, it's just not worth it. 01 Pro doesn't deliver much value to virtually everybody. That is not a coder. And Devin Pro has so much competition in the $20 range that why should you use this? Well, let's talk about it, and I have to say I'm saying this from the perspective of somebody who did not purchase access to this and hasn't played with this yet, just because I personally don't really see the value. And I'm also not clear yet on how I would even test this in a context that makes sense for this channel. Because what Devin essentially does is connects to your Slack and it acts as a junior developer that you hired and you can text it inside of Slack and let it do various things. It will figure out how to do them by itself and let you know that it saw the various messages. But at the end of the day, you're relying upon an AI agent to do things for you. And the state of AI agents today is as such that you cannot fully rely on them even if you achieve 80, 90% reliability. Imagine an employee that completes his task 8 out of 10 times and 2 times comes back with something that is completely disjointed from reality. That's not an employee you're going to keep around. And that's why in the development space, ideas like Cursor have seen such massive popularity, because they give you the ability to enhance your existing skills. They don't promise something that's currently impossible, in my opinion, which is doing all the tasks from A to Z. But Devin has exactly this promise and that's why it also comes with this high price tag, which obviously if that would work, a junior developer for $500 would make sense for a lot of people. But from everything I've seen on the Internet, and that's why also this launch seems to be a bit underwhelming and it's not getting so much hype so far. I'm just reporting what I'm seeing here. It's not perfect, it's just built on the state of the art LLMs, which also are not perfect. And no matter how well you orchestrate these agents in the background, which Devin seems to be doing extremely well, still going to be a little inconsistent and you're going to send the agent out to do a task and it's going to come back 15 minutes and meanwhile you're just hoping that it gets it done correctly. So, interesting release and I love this category and I really love the vision they're going after. It just feels like the state of LLMs currently only supports a workflow that enhances human ability and doesn't fully replace as a junior developer on most tasks. But again, that's just me echoing the opinions of what I've seen on the Internet and maybe reflecting on some of the experiences I had with some of the Deving competitors that are currently available. All right, next up, let's talk Grok announcements. I always like what they do. If you're following the channel, you might know that I'm kind of a fan of their philosophy, which is, hey, keep the AI truthful and let grownups make their own decisions. I like that. But what they did here, well, I don't know, maybe it's a step in the wrong direction. They released their own image generator. Okay. And before this, you might know that they integrated the Flux 1.1 model, which was state of the art. And that's why GROK was extremely good at generating images. Quite unhinged ones in some cases. Now they replaced this image generator with their own one, which they called Aurora. Now, unfortunately for Grok, we actually went ahead and tested this versus some of the other state of the art image generators, as we usually do, and here's some of the results. So this is what Midjourney puts out from this logo prompt. This is what Flux 1.1 Ultra puts out. And this is what Aurora puts out. Clearly, Midjourney and Flux are just better. Then we have a different portrait photography prompt Midjourney, Flux Aurora over here. I guess it's okay. I still think Flux is best here. Here's a cinematic still prompt. Again, I think Flux and Midjourney win, although it's closer than I would have imagined. Drone shots. Look at that. It's not bad. It's just not the best. And then on text, I was surprised. Look at that. It actually did super well. Aurora kind of killed it on this one. It might be even my favorite. I mean, okay, this. The artistic value of these majority outputs here are insane. But if you just want that image, another album cover or movie poster, this is perfectly fine. And here's one more test prompt that we usually run of a superhero character on top of a moving train. And yeah, just nails it here. Midjourney is so good at these. This is arguably the worst result for Aurora, but overall not so bad, but definitely not state of the art. So instead of Grok, you can access this now, which goes along nicely with a second announcement, which is the fact that GROK is now widely available to everybody, even to free users. Before this, it was only for premium user. Now, as mentioned before, it is not a S tier model. It's just inferior in certain ways, like prompt adherence, writing style quality, code generation quality. It's kind of just a bit worse across the board, except of the fact that it can pull together data from Twitter and now everybody can access that and I'm sure they're going to keep iterating on this over time. Stigo new image generator in Grok and it's freely available to all Twitter users up to 10 prompts every two hours and 10 image generations every two hours. Okay, next up. This one will be super quick, but it's a brand new website that I ran into. It's called countless.dev and it ranks all of the various models based on whatever you want. Actually check this out. So if you want to see which model has the longest output lengths, you could just rank it by that and voila. You can see right here Jamba 1.5 and Google vertex AI and Mistral code straw Mamba can actually put out a quarter million tokens. Impressive. You could do the same thing for input length where Gemini 1.5 Pro clearly wins with 2 million tokens of inputs. Look at that. You also have a pricing calculator here and you could compare different models to each other. Just a free website that I thought was kind of cool. And if you ever ask yourself which model does a specific thing best, this might be a good bookmark. And now let's move on to the next story. Trellis, the most popular hugging face space of this week. And we'll cover this quickly. But essentially this is a brand new image to 3D generator that performs pretty well. We've seen many of these pop up across the past year or two, but as somebody who doesn't work in 3D a whole lot, I can only judge my subjective impressions. But this is not bad. Look at that. Honestly, when I saw this, I just thought to myself, wow, this is probably the highest quality. I have seen it. You can input custom images, but I just wanted to include this to show you that from what I can see, these are getting better and better. And these models live in between the two other options that people who build games or work with 3D have. One of them is just modeling from scratch, which takes forever compared to doing something like this, where you can have a Dragon in seconds or the other end is buying prefabs, which can be pricey. And if you want to customize, you'll have to edit this anyway. This kind of falls in the middle. You get something custom immediately and then, yeah, you might need to clean it up, you might need to improve it. But with these image to 3D models getting this good, it should be too long until we see this implemented in some consumer friendlier way. I don't know. Leave a comment Below, if you know of a use case for image to 3D, I'd be interested to see something like that in practice. Practice. Okay, next up we have this little tool. This is completely free. Also not sponsored. I would always mention that, but I just thought this was sort of cool. It's an AI journaling app where you can log in for free and just kind of give this a shot. So I'm going to do exactly that right now and just journal for a second and see how this goes. Daniel on the team that has been researching this told me that this really positively surprised him and that AI powered journaling experience might be the little push that even he needed as he decided to start journaling recently but wasn't able to maintain the habit. But as per usual with a new habit, sometimes it's a bit tricky to get into the routine of it. And here you have an AI to reflect on it. So let's have a look. Let me describe how I feel today. Okay, so there are some of my thoughts about this day. This was a very long work day with all of these announcements. Wednesdays are always busy. AI news you can use always gets done. And see, you can just say reflect. And we'll come back with a question here. How does spending such a long day focused on AI and internal systems affect your overall well being and work life? Balance? Well, there is no balance. It's all work as AI goes crazy. Okay, let's reflect on that one too. Would you like another question? How do you plan to recharge and take care of yourself after such events? Okay, interesting. So I guess now this is sort of acting like a therapist kind of. But then also up here it analyzes what the mood of the day was. And when you come back on the next day, you can kind of keep track of how you were feeling, how you were doing. You can get suggestions, you can have a little conversation in here. Oh, look, it even says anxiety. I think it might actually be right. I didn't even consider that. I will go for the walk in the morning and read in the park for a bit. Reflect on that. How do you feel your productivity and creativity are affected by such long, intense work sessions? Focus on AI and internal systems. That's a good question. So there you go. So this is just a different way of journaling. You can try this out. You can come back a few days in a row, see how this goes. You can track your moods like this. I think it's sort of interesting to have this creative collaboration with the AI here. I don't know. It's completely free. Thought this was interesting. Wanted to feature it. All right, that's all we got this week. Don't forget to check out the Gemini and the Sora release videos that we uploaded this week. And with that being said, I'll see you next week, and I hope you have a wonderful day.