we've had another intense week of AI releases ranging all the way from aeropic releasing a llm that can remote control a computer we have some use cases of that in today's video and various creative applications that we didn't even know were possible before as you can see I'm still on the road here I'm in Guru Brazil where the wind is absolutely insane and I've been using starlink to upload these video clips to my video editors which sit in Europe and Asia respectively so here we go a truly Global episode of AI news that you can use featuring all the AI releases of this week that you can actually put to work today okay so this update made a lot of users very happy because advanced voice mode finally shipped to All European users including Switzerland Iceland Norway and lonstein so now everybody has access to Advanced voice check out another video on the channel for 20 plus use cases that you can put to work today with it there's a lot you can do today including the universal translator next up we have 11 Labs coming out with an update and this is easily summarized as prompt to any voice you can imagine you can find this in their web app under voices and then if you go to add new voice the very first option is voice design now and voice design lets you design an entirely new voice from a text prompt let's give it a shot shall we maybe randomize a few times how about an angry old pirate loud and boisterous okay but let's add a little bit of a personal touch here how about this okay you think you can cross Captain match carel latte and LIF to share the insta post okay so this is a pirate with a valley girl accent why you so rude man she stop doing that can't help it this is my voice let's see if it can actually pull this off I imagine this would be quite challenging let's have a look I face storms that would turn your bougie hair white and sea monsters that would make your knees like totally Shake I face storms that would turn your bie I face storms that would turn your bougie hair white whoa okay interesting I don't know I feel like doing one more how about a massive evil ogre troll with a pirate accent let's generate your weapons are but toothpicks to me surrender now and I may Grant you a swift end your weapons are but toothpicks to me surrender now and I make weapons are but toothpicks to me okay granted these were super tricky let's give it one more that is quite simple let's do this movie trailer voice presets that they have in here in a world on the brink of chaos one hero will rise in a world on the brink of chaos one hero will rise prepare yourself for a story of Epic Proportions coming in a world on the brink of now that's pretty good at this pace we'll soon have text to literally everything we have some updates from canva with all of their AI features upgraded even further they're calling this drop toe and throughout October they're releasing a bunch of features and you can check out all of them in the link in the description below I'm going to highlight one that I find particularly interesting because minor improvements or a writing tool which is essentially a cat GPT wrapper is not something I consider worth featuring but they have this brand new whiteboard plus AI feature and I always thought that all of these whiteboard and mind mapping type features work really well with AI so what I'm going to do is open up canva here and just click on one of these whiteboard presets how about the SWAT analysis that looks good to me and as as you can see there's a lot here okay there's different layouts and you can fill all of these out zoom in and out this is nothing revolutionary but if I went ahead and used all of this and not just used I could also collaborate in here right I can add other users and we could all work on this collaborative SWAT analysis here's the new thing you can go ahead and select everything and now you should be able to say magic right create summary and you will summarize everything on this whiteboard with the power of AI and I think that's just pretty amazing look at this obviously there's no info in here so this content will be useless but I just thought it was an interesting workflow to have a collaborative visual workspace like this where multiple people can contribute and then you can use AI on top of that as a final step or as an addition not as a core feature and the summary you could sense to whomever might care about the results of it but not about the intricate details of the process as you were developing this SWAT analysis mind map site Maps heck you could have entire business plans in here and then just summarize it with AI in the end and for many visual Learners this be a better way to lay out things than simply going ahead and throwing everything into a Word document and as you can see there's a lot of these whiteboards so you could do customer Journeys too and also there's a whole lot more AI features so if I just select all of this we can do all of these presets or even a custom prompt on top of the visual elements pretty interesting and if you want to check out some of the other features there's a lot of minor features here that relate to both Ai and design so this is actually absolutely massive news you can news you might have heard that Claud released a brand new set of models the Sonet 3. 5 new and the Haiku 3. 5 5 is coming and also they have their brand new computer use API but as you might know I created a separate video on the channel going into all the details and showing you where you can access this even as a non-technical person so if you want to see all that you can check out the dedicated video but what I have for you here is the first use cases that have been popping up and as promised we'll do a dedicated video on this exploring all of them and not just showing you what we found while researching this but also what all of the internet has been doing with this brand new feature that is essentially a llm that remote controls your computer so here's a few while examples one of them would be this prompt go to YouTube find the video and Skip all the ads and then look at it doing it it full screens the video it finds a skip button it presses it and then you can get Rick Rolled without ads stopping you okay but admittedly that's not very useful how about this one where it goes ahead and fills out different job applications for you all with a simple prompt that says first scrape enr.
com with fir crawl next scrape to their career pages with fir craw and find a job navigate to the job page using Firefox and click the apply Now button until you see a form then find the why do you want to work at a propic text box and enter a great answer into the form box based on the scrape and look at it going to the correct page and applying to the job and using an llm to fill out this field and you can imagine that if you give it extra context in form of your CV it could take all the info and fill out all the fields for you send it and then you could use a prompt generator to generate variations of this prompt with different websites where you can apply to different jobs and then this thing would just go ahead and apply to different jobs for you all day long with your CV and custom answers and this is where the power of prompting comes from because if you prompt it really well it's even going to sound like you it's going to have all the context on you it's going to know which pages to go to and now all of the prompt generators that I've been teaching you for a while now with the various products I mean since over a year in the freeb get with our newsletters you get 10 prompt generators well now you could go ahead and repurpose those to create different anthropic computer use prompts and then this Dam thing goes out and does the work for you where are all the people claiming that prompt engineering will be completely useless in 2024 where where it's not 2030 yet you need to know how to communicate with the AIS to get things done today this is super super interesting to me I'll be playing around more with it than reporting back in a dedicated video just focused on various use cases of how to put this to work next up we have X releasing their Gro API if you're not familiar this is Twitter SLX large language model that has access to all the Twitter data that is its main advantage but to be honest I don't know many people that actually use it regularly if you do please leave a comment below mostly the story of this AI has been that they're catching up to the other players in the game they do have the unique data but just the quality of the outputs and the tooling around it haven't been there yet but now they have an API meaning you can build this into various applications and pay per use and people are trying things out with it like Daniel San over here uses it to generate code inside of vs code now why would you use grock beta over Sonet 3. 5 that is state-of-the-art at code generation especially with the new updates this week I'm not exactly sure but you can do it but then this use case might be a bit more interesting XI actually put on a hackathon and so here built a Chrome extension that allows you to bring your own Twitter algorithm to websites and it filters it using grock and they're using this Onix allowing you to effectively modify the algo on your own Twitter feed with this extension it checks out the different posts and adjusts them based on the topics that you picked in your preset I mean this is interesting but it should also be possible with other API I guess the advantage that you have here is that this does have all the Twitter data so it probably makes most sense to use it to moderate Twitter posts not exactly sure but nevertheless xcii is catching up there's an API now you can use it and now let's move on to the next story which is Runway act one and this is one that might not be available yet for you they claimed that they started rolling this out I don't know a single person who has this yet but this thing is super fascinating in a nutshell this is essentially motion capture without the crazy device you're probably familiar with some behind thes scenes footage of how Hollywood movies are made especially in the VFX Department wear these green suits or these suits with all of these different tracking points or tracking devices on a person so that when he moves around they can map characters on top of them perfectly now Runway is the first player in the AI video game to release a feature that is trying to mimic this without all of the technology and all of the extra equipment to track something here all you need is an actor performing a certain expression or moving ahead in a certain way and then so let me get this straight you came all the way down to the Department of Motor Vehicles and didn't bring your driver's license do I understand that correctly you're going to have to go in the uh separate line isn't that amazing I mean look at this there's a bunch of examples on this release page and again they're claiming that they're slowly releasing this but in all of these demos this looks absolutely incredible and this is something we haven't seen before now we had an interesting discussion with the team about what's next with this and obviously what's next is well the Avatar will be you and then somebody else can reenact you as you speak so I don't know this could be Ai igore and somebody completely else could be sitting here presenting and then I could just map my avatar on top of it use my levels voice to actually reproduce the voice I mean Heck if you check out 11 labs they do have the voice changer where I can pick Eagle AI advantage and then somebody could record audio and it just gets reproduced in my voice with this Tech doing the video it's about to get crazy I might even be able to take a week off for the first time in years because somebody else will be presenting news you can use and all of the itch will do the makeup of the voice and the Avatar now is this actually good H I don't know I guess it depends on the use case I personally think there's value to this human touch of the interaction that we're having right now but this certainly opens up some new opportunities that most people haven't thought of so far and I personally can't wait to try this myself once I get access okay when it comes to image generation there's a bunch of new releases this week starting with M Journey actually announcing something but this is only available to a exclusive set of users it's new image editing features and they're only accessible to people who are subscribed for on the yearly membership subscribed for the past 12 months or have at least 10,000 images my account actually does not fall in this category because I started using other tools next to my journey and I got to admit I canceled my sub around two months ago as I mostly go to flux these days if I need something but essentially they're adding some of these editing features that we've seen in Photoshop a while ago and that's pretty much the story here bringing me to the next release of this week which is ideogram canvas magic fill and extent this is very similar to M Journey's release they're adding these in andout painting features which allow you to modify only parts of the image or areas outside of the image but let me tell you all of these features that you see here in both ideogram and M Journey are things that we've had in Photoshop for a while and essentially they're features that you could do manually if you knew how to photoshop properly before this is just ease of use being enabled by artificial intelligence and what we're seeing this week is some of these feature and trickling down from the pro level apps like Photoshop into something like a deogam or M Journey making it accessible to most consumers so if you could benefit from something like extending an image into something wider or replacing a specific object in an image well this week we got multiple alternatives on how you can easily do that in the apps that you might already be using next up we have stability I releasing stable diffusion 3. 5 large now rather than me telling you about this let me just show you because me and the team actually went ahead and created this new Excel sheet that compares all the major image generators on a few prompts that we deemed to be quite useful book covers portrait photography logo design and some specialty techniques that we like as you can see you have the comparison of all the different models here ranging from mid Journey 6.
1 across flux 1. 1 Pro but also IDE 2. 0 and what I did here is I took some of these prompts and I also ran them through stable diffusion 3.
5 so we can compare the quality levels of this versus some of the other top tier image generators and by top tier I mean we have this monthly ranking that we freely publish we updated it once a month so you can stay up to date on what tools are the best in our opinion link below but now let's have a look at what stable defusion 3. 5 produces for some of these images first up we have this portrait photography prompt so right away I just noticed that these eyes are a bit off they just don't look very real especially when you compare to something like mid Journey or flux I mean this is just not on the same level fair enough that's one image let's not judge too quickly how about this logo okay really this is what it comes up with versus these results in M Journey flux kind of a magic Studio didn't get the text right but it's a little more detailed and actually really like these ones from ideogram versus again this okay that's not good let's give it one more chance how about this cinematic still prompt this is a technique that we originally saw from Tim from theoretically media and Matias then runs the events in our community absolutely loves this he uses it all the time and it produces stunning results across all generators arguably Leonardo does the worst but I suppose it's okay flux imaginary are super impressive here the other ones are okay and this is what I got from SD 3. 5 this is terrible this person doesn't even look like a person come on this is D two level humans so I don't know am I missing something here this release is just very underwhelming one thing that I should warn you about is that it is quite un sensored so if you go in here sometimes you will just get graphic images without a warning so you can generate all sorts of unsensored stuff here but other than that not sure why one would use this over flux all right then next up in AI video generators we have two new releases one of them fully open source and another one is version 2.
0 of hyper we went ahead and tested these two for you so here are the results from the fully open source Mochi One release this is the open source video generator by genmo We compare them to the meta movie gen prompts as metam movie gen seems to be the best thing we have seen so far maybe exora well not bad on this ghost prompt this is a tricky one next up we have this monkey prompt again we'll put up a comparison on screen so you can see the difference between metam mov gen and this but this is surprisingly good physics look realistic it handles the fog well the consistency on this monkey is super good eyes look realistic I mean it's a bit of a ridiculous scene but I don't think there's anything obviously bad about this and next up we have two more shots of a sloth chilling in a floaty I don't know there's something about this one that is extra fun and it sort of works there's not a lot of movement it's quite subtle but the Shadows the water the love with the glasses it all looks good now we did generate this one more time and this generation looks absolutely terrible so I also wanted to include this in here I mean this looks like something I would have made inside of Photoshop when I was 15 uh that's when I was learning Photoshop by the way this is just not good but fair enough just reran the prompt and all of a sudden it was great so there you go Mochi really impressive and this thing is available under an Apache tool license meaning you can use this for commercial purposes in your own project that's pretty amazing at this quality level and then we also have the hyper 2. 0 release and Hyper 1. 0 was actually the model that we were surprised by how good it was and they have a 2.
0 model so let's have a look this one we ran through some of the image to video prompts We compare it with you might be familiar with these and the results are surprisingly good even better than hyper 1.
