de blur this image not bad not bad remove her earrings pretty impressive all right let's try this what can be eaten replace it with pasta I'm starting to get a bit mind- blown here let's try this using her pose make a boy studying in the classroom no way how about this take the man on the right of this photo and the man on the left of this photo and have them eat dinner together oh my God this is insane so this might be the most intuitive and Powerful AI generator you've seen so far it's called omn gen and I think this could very well Define the future of image editing if you want to edit an image you don't need to adjust the brightness or contrast or saturation sliders anymore you don't need to do any white balance or color correction you don't need to paint over an image or mask things out or do any other manual stuff all you need to do is prompt it on what you want to edit and it would edit it for you and best of all this is free and open source so in this video I'm going to show you how to install this and run it locally on your computer you can run this offline for free for unlimited times plus I'm going to test it out on a series of tricky tasks to show you all the cool things that it can do so let's Jump Right In first of all let me just show you a few cool examples from their research paper it can do so many things for example if you upload one photo and you ask it where can I wash my hands please help me find the right place it will actually highlight the sink for you plus it's so easy to create deep fakes or face swaps with this tool you no longer need any third party or external plugins so for example let's say you want to get wo Kong and Iron Man in a kitchen together well you can say in a cozy kitchen man one which is Wukong is chopping vegetables while Man 2 which is Iron Man is sitting nearby and this is what you get you can also just upload a blurry image and then type in deblur the image and this is what you get the same works for brightness so you can also upload a really dark image and then write brighten the image and this is what you get it understands context really well so let's say the prompt is woman one which is Queen Elizabeth here is holding a cat and sitting on a bench in a park woman one is the person smiling in image 1 which is here and then the cat is the one in image 2 which is here so you can just upload an entire image which isn't just the cat and it knows how to identify the cat in this image and then generate an image of Queen Elizabeth with this cat sitting on a bench in a park here if you ask it to remove the butter on the cake that's exactly what it does it removes the butter on the cake here's another really impressive example so here's the input image and then if you prompt it with delete the rain in this image that's actually what it does it actually gets rid of the rain droplets on this window pretty well and it can carry out multiple editing tasks in the same prompt so for example if this is the original input image and you prompt it with add sunglasses to the man's face and change the color of the clothes to Blue that's exactly what it does it not only adds the sunglasses but it also swaps his shirt to blue super power ful tool now enough showcasing their demos let's actually try this out so I'm going to link to their GitHub page in the description below on this page there is a link to a hugging face demo so if you don't have a powerful enough GPU you can try to use this space online to generate images now you may or may not be able to run this on the free daily credits you probably need to top up more credits to use more compute but what we're going to do today is actually install this on my computer so we can run it for free for unlimited times now they don't specify the vram or Hardware requirements explicitly on their GitHub page but it seems to only work with Cuda gpus right now and some people have reported to successfully run this with as low as 12 GB of vram by the way I'm using a Dell Precision 5690 you can integrate a powerful RTX 5000 Ada into this so this is a perfect compact combo for running AI tools locally huge thanks to Dell and Nvidia for sponsoring this anyways on this GitHub page if you scroll to the middle it will contain the instructions on how to install it so you can install this wherever you want I'm going to install it on my desktop so I'm going to open up my desktop folder and then in this top bar up here I'm going to type in CMD which will open up my command prompt but in the desktop folder and then opening this back up the next step is to get clone this repository now note that you do need to have git installed before you can get clone anything so if you don't have git here's how to install it if you already have git installed feel free to skip to the next section so all we got to do is download the latest release for whatever operating system you're using so I'm using Windows so I'm just going to click on download for Windows I'm running 64-bits so I'm going to click on this to download and it's now downloading the cxe file so once that's completed all we got to do is open that exe file and then follow the steps so I'm going to click on next I'm just going to go with the default install location which is program files slit so I'll click next for that and then I'm just going to leave this at the default and then I'm going to click next again and click next here we're just going to use the default settings for all of these there's a lot of settings that you need to go through so I'm just going to click next for all of these all right and then it should go ahead and install all the files so this might take a few minutes perfect so now we we have G installed anyways assuming you have get installed let's copy this first line and then paste it in our Command Prompt so this will clone the omn Gen repository into our desktop folder all right perfect so after it has cloned the repository if you open up your desktop you should see a new folder here called omn gen and if you open this up it will contain all the files that are in the repository so these are the same folders and the same files as you can see here all right so the next step is to go into this folder so let's paste this line in our Command Prompt and this basically means change directory to Omni Jen because before we were still in the desktop folder right now we are going one folder in into the omn Gen folder and then the next step is to pip install everything you could run this right away however what I prefer to do when installing each of these open source tools is to create a virtual environment first so this step is optional but it keeps things from breaking so what a virtual environment is think of it as like a separate hard drive on your system and you're going to download all the packages and dependencies of this tool in that hard drive and this is separate from everything else you have on your computer and the benefit of this is if this requires some dependency or requirement that conflicts with an existing tool you can can download everything separately for this and it won't affect anything else you have on your computer now in order to create a virtual environment we're going to use cond if you don't have cond installed here's how you install it if you have it already feel free to skip to the next section now I'm just on anaconda.com and actually what I'm going to do is install miniconda this is a minimalist version of Anaconda if you install the full Anaconda it installs a lot of packages and dependencies that you might not need this just takes up more room on your computer and of course the installation time is a bit longer but with miniconda it's just a barebones package and you can always install additional packages and dependencies afterwards so I'm going to click on latest miniconda installer links by python version and I'm using Windows so I'm going to install one of these now for free and open- Source AI tools usually they do not support python 3.12 so it's better to install the python 3.11 version so I'm going to click on this which should download an exe file to your computer once it's finished downloading simply double click on this and then follow the steps to complete the installation so I'm going to click next and then agree and then let's set this to all users I'm going to go with the default destination folder and then I'm going to check this as well clear the package cache upon completion this just gives you back some more dis space without affecting functionality all right once that's completed let's click next and then we are finished now we aren't done yet so if you open up the command prompt and you type in cond D- version you're still going to see that cond is not recognized this is because we haven't added Anaconda to our path yet so let's exit out of this and then to add it to our path we simply search for this function edit the system environment variables we're going to click on this and then click on environment variables and then click on the one that says path and then click edit and here's where you add in the path of anaconda so it depends where you installed Anaconda for me I installed it in program data so it's going to be in program data/ miniconda and then if I double click on scripts you can see that cond is here so this is the folder we want to paste in so I'm going to right click on this and then copy as path and then back in the environment variables window I'm going to click new and then paste in the path here and then click okay and then okay and then okay again now if you open up command prompt again and type in k-- version you should see that we are running 24.5 point0 so this shows that we have successfully installed Anaconda all right so assuming you do have cond installed back to our Command Comm and prompt where we are in this omn gen folder let's type in condac create DN and then omn gen which is what we will name our new environment so what this means is we are using cond to create a new environment called omnigen so let's press enter and then let's press enter all right so it has been created now the next step is we actually need to activate the environment before we can use it so let's type in cond activate and whatever we named our environment in our case it's omnigen and you can see that your environment has been activated if you see omnigen in parentheses before every line all right so going back here the next step is to copy and paste this final line so let me paste this in here and let's see how long this takes all right so you can see it's loading up all the dependencies that it needs to download and then it's downloading all the files so for example accelerate diffusers torch so depending on your internet connection this might take a few minutes all right and if all goes well you should see this line again with no errors above so next step is we could create images using all these lines of code but it's not really intuitive so instead let's open up a gradio interface which looks like this and it's just easier to navigate and adjust things so in order to do that we first need to pip install gradio and spaces let me copy this line and paste it in here so again this is going to download another set of requirements which might take a few minutes all right perfect and again if all goes well you should see this line again without any error messages so the only thing left to do is to run this app.py file so let's copy this line and then paste it in here and press enter all right and the first time you load this up it's going to install additional things such as this model safe tensor file which is basically the model that is required to generate the images and note that this is 15 GB so it's going to take a while for this to download and one thing I forgot to mention is you might be wondering well how do I start this from scratch the next day so all you have to do is go to your omn gen folder and then in the top here type in CMD to open up your command prompt in this omn gen folder and then type in cond activate omnigen or whatever you named your virtual environment and then you should see this omnigen in parentheses at the start of every line and then after that all you have to do is type in Python app.py to load up the gradio interface so we've already installed the model. safe tensor file the first time we loaded this so This should only take like a minute to load up all right so if everything loads up successfully then you should see this link here so if you hold down control and click into it it's going to open up a gradio interface in your browser so note that even though this is on your internet browser this is completely offline you can run this without internet for unlimited times so let's try this out right now it's pretty intuitive to use I'm going to show you all the settings right now so for the prompt let's enter in a young woman sitting in a cafe she is wearing golden earrings and a white sweater in front of her is a table with a slice of cake and let's just leave all these images blank for now I'll show you what these do in a second and then for the height and width let's also leave it at 1024x 1024 guidance scale is how much you want the image to follow your prompt So Right Now the default is 2.5 if you drag this all the way to the right then it's going to interpret this really literally whereas if you drag this all the way to the left it's not really going to follow this as much and it might give you some random result but this gives it room to be more creative anyways I'm going to set this back at the default of 2.5 and then image guidance scale is how much you want these uploaded images to affect your result so if you drag this all the way to the left then these images would have more influence if you drag this to zero then it's not going to have as much influence I'm going to leave this back at I believe it was 0.5 and then inference steps this is just how many steps it takes to generate the image so in general the more steps you have the higher quality and the more detailed the image will be but at a certain point if you have too many stepss then you just get diminishing returns and it's just not worth it so anyways let's just keep it at the default for 50 for now and then for this checkbox check it if you want to reduce the memory cost same with this one if you don't have a really good GPU with enough vram you can also offload some of the processing to your CPU so check this if you also want to reduce memory usage now of course if you offload some processing to your CPU this is going to take way longer I'm going to leave this one off and then finally this setting is pretty self-explanatory so this is if you want your output image to be the same dimensions as your input image which I will leave off for now and then let's click generate and see what that gives us by the way I'm using an RTX 5000 Ada on a laptop which only has 16 GB of vram so this is less powerful than a 4090 for example which has 24 GB of vram and you can see the processing time down here so it does take quite long to generate one 1024x 1024 image all right perfect so this is what we get and if you open up our Command Prompt you can see that this one image took around 1 minute and 47 seconds so pretty slow compared to stable diffusion or even flux but we do see a young woman sitting in a cafe she is wearing golden earrings and a white sweater and in front of her is a table with a slice of cake let me tell you about this this awesome tool called chat llm by our sponsor Abacus AI this allows you to use the best AI models out there allinone integrated platform this includes the latest 01 preview 01 mini as well as other state-of-the-art models like GPT 40 and Claud Sonet 3.5 plus they have a new route llm feature which automatically picks the best llm to use based on your prompt they also have a search llm feature which searches the web to find the latest and most accurate information for you you can also generate images right from the chatbot and they're using the best generator out there flux Pro Plus you can even generate videos with a single prompt how cool is that plus it has a really cool artifacts feature so if you're coding or building something with it it allows you to view and interact with the app side by side you can also drag and drop PDFs and documents into here for it to analyze this is great for generating reports and analyzing data easily you can also create custom agents to automate specific tasks this is just a really powerful and versatile way for you to use the best AI models out there all in one platform try it out via the link in the description below now to download this simply click on this button to download it to your computer all right so what we're going to do is actually I'm going to drag the downloaded image onto here to use as an image image reference here's the cool part about omn Jen let's say I want to remove the earrings well all I have to do is reference this image by typing in this over here so it's an image tag followed by image1 and then a closing image tag so I'm going to just paste this in and then just to keep it more organized I'm going to press shift enter and write remove the woman's earrings now you don't have to start a new line this is just to keep everything more organized but that's how you would edit an existing image and then there's one more toggle we can play with which is down here this's use input image size as output so let's say you're uploading an image and you just want to edit the image so you want to keep the same dimensions as your uploaded image then you can just check this and it would basically ignore the height and width you set here and just use the height and width of your input image so all right everything looks good let's click generate all right perfect and as you can see this is indeed the woman but now she has no earrings so you can see why I think this is such a powerful tool you no longer need to manually select any objects in the image or in paint anything all you need to do is prompt it what you want to edit and it will automatically identify the object in the photo and edit it all right let's try another example instead of just replacing objects in an image you can actually prompt it to highlight something in the image so for example let's let's try this what can be eaten in this image highlight it in blue and in theory it should highlight this cake in blue so I'm going to keep all these settings the same in fact let me even reduce the number of steps to 40 just so this image can load faster there's going to be some sacrifice in quality but I think it's minimal and then let's click generate and see what that gives us all right and this is what we get and indeed it highlights the cake in blue very impressive now instead of just highlighting the object we can also ask it a question and then replace whatever object it identifies so for example if you ask it what can be eaten in this image and then let's type in replace it with pasta what it's going to do is for your input image identify what object can be eaten in the image and then replace it with pasta so let's keep all the settings the same and click generate image and see what that gives us all right and this is what we get and you can see it indeed identifies the food in the image and replaces it with pasta how cool is that now on the note of replacing things you could also use this as a clothes swapper so using the same image if we type in replace the white sweater with a red bikini and then we'll keep all the settings the same and click generate let's see what that gives us all right so here's what what we got you can see it's not perfect especially her left arm there and even the bikini is not perfect but this is a very tricky image to do a clo swap first of all because for this input image she's wearing a very thick sweater so it's really hard to guess her body shape and also there's a table and a piece of cake in the foreground which is obstructing some of her body so it's quite a tricky task in this example for it to make her wear a red bikini however if you just have a photo of a woman standing with no obstructions in the foreground it should be easier to swap her clothes for that image anyways instead of just swapping clothes here's another cool thing you can do so you can also get it to generate a depth map of this image so I'm going to scroll down keep all the settings the same and then click generate and voila here we have a depth map of this image how cool is that so you no longer need like a control net pre-processor or whatever to generate pose Maps or depth maps you can just upload an image and prompt this AI to generate whatever map you want based on an image how cool is that all right here's an even crazier example so again I'm going to take the same photo but this time for The Prompt what I'm going to say is using the same pose as this image generate a new photo a boy studying in the classroom and again I'm going to keep all the settings the same and click generate and see what that gives us and and here's what we get how cool is that so you can see it preserves the same pose as this original image but it turns it into a boy studying in a classroom a super powerful tool before in order to do this again you need control net you need to extract the pose of this image and then apply it to this new image which requires multiple steps and it's just a hassle to do so many things but with this you just upload an image and then prompt it to copy the pose and that's pretty much all you need to do here's something even more impressive you can use this as a face swapper so what I'm going to do is upload this image of Bill Gates and another dude and then for image two I'm going to upload this image of Will Smith and his wife now the reason why I'm uploading two people in an image is because you don't need to crop out each individual face and then upload it so here's what you can do let's say I want to generate an image of Bill Gates and Will Smith eating dinner well the prompt can be two men are having dinner the first man is the man on the right of image one so this is the man on the right of image one and then the second man is the man on the left of image 2 so this is the man on the left of image 2 so you can see I don't even need to like crop out his face or his face I can just upload the entire photo and then specify in this prompt all right only use the face of the man on the right here and the man on the left here and make both of them eating dinner together now in this example because I don't want the output image to match the dimensions of these two images I'm going to uncheck this option and then up here is where I can set the height and width of the output image and in this case just to make it load slightly faster I'm just going to set it to 768 by 768 and then let's click generate and see what that gives us all right and this is what we get you can see it is indeed taking Bill Gates and Will Smith and getting them to have dinner together and it even copies the outfit that they are wearing so you can see Bill Gates is wearing this black sweater with a blue and white color and that's what you see here same with Will Smith you can see he's wearing this white shirt with a red jacket now notice that there are some flaws here this part is obviously not correct but other than that I mean this is pretty good and it's so intuitive you no longer need to download any face swapper plugin like PID or instant ID you just need to upload two images and then prompt it with whatever you want to generate and omn gen can understand everything and generate it for you super powerful tool all right finally let me get rid of these two and I also want to test out different styles of images so let's test this one out uh girl in a black dress at home anime style so let's see if this generator can generate anime style images I'm going to keep all the settings the same and click generate all right and here's what we get so this is kind of anime style it's not great though definitely stable diffusion and flux do a lot better than this in terms of anime all right finally let's also see if it can do Disney Pixar style so I'm going to type in Disney Pixar style let's see if they can figure this out so again I'm going to leave all the settings the same and click generate and here's what I got for Disney Pixar style this looks more like a realistic photo to me so let's give it one more shot I'm going to type in like 3D animation style and let's see if it can pull it off but I have a suspicion that this could currently only do well with realistic images okay so this is a bit better so instead of just inputting Disney Pixar style if you input 3D animation style then this actually works all right so let's save this image and then I'm going to drag it here and then same as before I'm going to paste in image1 wrapped in image tags and then for organization I'm just going to start a new line and then I'm going to write replace the woman's black dress with a red dress and then let's scroll all the way to the bottom this time because I'm using an image as a reference yes I want to use the input image size as the output and then I'll leave all the other settings the same and click generate all right and this is what we get and it indeed replaces the the black dress with a red dress now as you can see it's not perfect you can see a slight outline here where the black dress used to be plus the hands and fingers are not accurate but it is a really easy and intuitive way for you to edit anything in an image so it seems like it doesn't work too well for 3D animation style images in terms of like replacing objects or clothes in an image it works the best with realistic images all right let's try something else so I'm going to drag and drop this blurry photo of a woman this is a photo which is quite blurry and what I'm going to do is I'm going to paste in the image one tag in order to reference this image and then I'm going to write deblur this image now I'm going to keep all the settings the same but we are going to use this checkbox down here use input image size as output because I want my generation to be the same size as my up loaded image and that's pretty much it let's click generate and see what that gives us all right and this is what we get so you can see that it does deblur it to some extent but this is not good quality and actually I have a suspicion on why that is because the uploaded image is very blurry so if it's just referencing that image then it's going to still include a lot of blur from that image so I think what we actually need to do is drag this image guidance scale slider all the way to the left and what this basically does is it reduces the influence of my uploaded image so it's going to reduce the influence of the blur of the image and hopefully that gives it more room to actually sharpen the image so I'm going to leave all the settings the same and click generate perfect and this is what we get so it's slightly better than the generation we had before and for your comparison here's the before and after so you can see it has indeed deblurred this image all right next I'm going to upload this really dark image of an owl and then instead of De blur this image I'm going to write brighten this image so I'm going to leave all the settings the same and click generate and there you go it brightened the image how cool is that this is such a powerful tool I mean traditionally in Photoshop or wherever you need to manually adjust the brightness and contrast sliders maybe you need to adjust the levels and the exposure there's just so much manual work involved whereas for this you just prompt it with brighten the image and it automatically brightens it to the best amount for you and note that for this really dark image there's already a lot of data that has been lost it's really hard to restore or even guess the colors of such a dark image sometimes so you probably can't just use Photoshop or a regular image editor and then just bump up the brightness slider to get this result but because omnigen is using some AI algorithm in the back end it's able to use AI to actually guess or extrapolate what the colors should be and then increase the brightness appropriately so this is a super powerful tool so anyways that sums up my tutorial on omn gen let me know in the comments what you think of this again I don't think the quality of the images are as good as flux or stable diffusion in some cases but this is not intended to be an image generator this is more of an image editor and this is super intuitive in instead of going to comy UI and dragging all these nodes and noodles around just so you could edit a specific object in an image or copy the pose of a character onto another image with omnigen all you have to do is upload an image and then prompt it with whatever you want to edit so everything is a lot more intuitive and a lot easier to use anyways let me know in the comments what you think of this and if you run into any errors while installing or running this let me know in the comments as well and I'll try to help you troubleshoot as much as possible as as always I will be on the lookout for the top AI news and tools to share with you so if you enjoyed this video remember to like share subscribe and stay tuned for more content also there's just so much happening in the world of AI every week I can't possibly cover everything on my YouTube channel so to really stay up to date with all that's going on in AI be sure to subscribe to my free Weekly Newsletter the link to that will be in the description below thanks for watching and I'll see you in the next one on