Mastering Consistent Characters in Leonardo AI: How to Make AI Influencers (Step-by-Step Tutorial)

[Music] hey what's up everybody B here from blog withth bin. comom and in today's video I'm going to show you how to create consistent characters with Leonardo Ai and the best part you won't need any additional software or coding skills everything we're about to cover in this video can be achieved within the Leonardo AI platform and in today's tutorial we'll explore two powerful methods to designing characters that are not only visually stunning but also carry a consist consistent theme across your entire portfolio this is crucial especially if you're looking to venture into the world of AI influencers a trend that's not only captivating audiences but also opening new doors for revenue and brand endorsements recently we've seen a rise in AI influencers that are bringing in some serious cash this particular AI generated influencer is pulling in almost $111,000 per month and has 300,000 Instagram followers that's literally insane but it's the wor world we live in now and you either adapt or you get left behind so by the end of this tutorial not only will you have the skills to create your own consistent characters and AI generated influencers but we'll also cover Pro tips on refining your designs and making your character style stand out real quick before we get started if you get any value out of this video I would greatly appreciate it if you would subscribe to the blog withb YouTube channel helps me bring you more resources and it keeps you up to date with all the AI and Tech Trends happening throughout the year but either way thank you so much for all your support all right with that being said let's get [Music] started okay so I'm going to show you two different ways to create consistent characters in Leonardo AI that way you can decide which one works best for you and your projects and before we dive in I want to point out that Leonardo AI has some really helpful documentation on this topic that you could access by clicking this little icon in the bottom right corner of the dashboard this will open their support chat and help docs and if you click where it says search for help you can browse all their support documentation and even search by keyword or key phrase just type what you want to search for in that field like consistent characters and you can find some really helpful tutorials that walk you through the process again we'll be covering all of these steps listed in this help doc but it's still a great reference to come back to if you need some extra clarification on how to create consistent characters with Leonardo AI also if this is your very first time using Leonardo you can definitely still follow along in this tutorial but there may be some terminology and steps taken throughout the video that may seem somewhat intermediate or Advanced if you're brand new to all of this so if you want an in-depth overview of how to use Leonardo before diving into consistent characters or if you still need to sign up for a free Leonardo AI account I highly recommend my YouTube tutorial mastering Leonardo AI a comprehensive step-by-step tutorial for beginners and I'll link to this in the video description below but there you'll learn the ins and outs of the entire Leonardo platform and will be an AI image generating Pro in about an hour so check that video out if you feel like you need a refresher on Leonardo all right with that being said let's move on and create our consistent character so for this video I'm going to be showing you two fairly quick and easy ways to create consistent characters in Leonardo the first will be in the image generator where we'll use prompts to create our character and then build off of that particular character with a fixed image seed then the next way will be to create what is called a trained model this is essentially you inputting data or images to train the model within Leonardo then with its Advanced Training algorithms Leonardo AI can create realistic images based on the given set of data and if you're brand new to all of this and scratching your head right now no worries I'm going to walk you through the entire process step by step right now starting with creating a consistent character with a fixed seed so first things first let's head to the image generator within the left sidebar menu under the user tools section go ahead and click on where it says image generation then from here we're going to create our character as we would any other type of image and that's with a descriptive prompt we'll then copy what is called the seed and use that to essentially guide the AI on how to create similar images with new prompts and if you're brand new to seeds in this context an AI seed is a series of numbers that tells the AI how to generate an image simply put the seed image is an AI generated image that you provide to guide the AI in generating new images aligned with your vision it acts as a starting point setting the stage for future images again if you're scratching your head right now no worries Leonardo makes this super simple to learn and Implement and let me show you how now really quick I should point out a few things before I enter my prompt and create our character which in this example will be an AI influencer so the first thing you should know is that I'm using a premium version of Leonardo in this tutorial and all that means is that I paid for a Leonardo subscription rather than using the free version so that also means that I'll have access to more tokens as well as the Alchemy 2 pipeline now if you're using the free version of Leonardo you can definitely still follow along and create consistent characters but the quality of your images may be slightly different than mine in this video plus you may not be able to create as many images in one setting due to the amount of tokens you get with the free account but either way just because you have a free account doesn't mean you can't follow along in this video and create consistent characters okay next let's quickly go over some configurations for this particular prompt so that you could see exactly what I'm using to get these results so first I'm generating four AI images at a time and as you'll see a little later on in the tutorial that will come in handy because the more images you have for the data set when training the model the better below that you'll see that I'm using the photo real and Alchemy pipelines if you don't have a paid Leonardo account then you won't have access to these but as I mentioned earlier you can still get by however these pipelines will enhance the quality of my images additionally within the photo real pipeline you do have the ability to configure the version and model being used I'm using version 2 and Leonardo's newest model Leonardo Kino XL next I have the public images option turned off again only paid subscribers to Leonardo can do this for now but keep in mind that if you have a free account all of your images you create will go into the Community Feed below that are the input or image dimensions and for this example I'll be using 512x 768 pixels but feel free to use whatever works best for you in your projects now I should point out and I may be getting ahead of myself here but if you plan on using these images that we're about to create to train your a a model Leonardo recommends images that are 512x 512 pixels or 768 by 768 pixels and this will give you the best data set training results you could still get great results with these Dimensions but I just wanted to make you aware of that before we moved on and just a heads up you can enter your Custom Image Dimensions here if you'd like okay then all the other settings in the sidebar I'm leaving as is and we will be configuring an advanced setting a little later on whenever we implement the fixed seed but I'll show you that whenever we actually use it next you'll notice that since I'm using the photo real pipeline I'm unable to change the type of model being used however I can change the style I currently have it set to cinematic but if you click this drop down you'll be presented with a handful of different styles to choose from adding another way to enhance and stylize your images then finally we have the elements and I'm not adding any elements to these images in this video but feel free to experiment with different elements here if you'd like they too offer a great way to enhance your images and make them really unique and with that being said let's create our consistent character okay so once you have all of your settings dialed in it's time to enter the prompt and you have a ton of creative freedom when it comes to how you want to go about this but within this field where it says type prompt that's where you'll type telling the AI what to create now we could do a whole other video on what makes a good prompt but what I've found is that less is more meaning you don't need to write a whole paragraph every time describing what you want I try to keep things concise less than 200 characters but I use descriptive words like peaceful Vivid breathtaking Etc or I'll be specific about physical details like skin color eye and hair color body type what the subject is wearing where they're at Etc things like that and as you get more practice with writing prompts you'll start to get a feel for what works best for you but either way this field is where you enter your prompt also I highly recommend revisiting Leonardo's consistent character help DOC for prompt tips they have some great Insight on ways to structure your prompts when creating consistent characters but for this particular example I'm creating a female Instagram model with white hair tan skin blue eyes athletic build so I'll enter that in the prompt field now the only reason I'm not adding much more detail to this prompt is because I want to get a good base that just focuses on the subject's face that way I could decide whether or not I want to use this seed for future images and let me show you what I mean so once you've entered the prompt go ahead and click the generate button and keep an eye on how many tokens you're spending on each generation it'll be listed on the button as well as below it just FYI then the AI gets to work and begins creating our image based on the text prompt now it usually takes about 30 seconds to generate four Images but for the sake of time I'm going to speed that up so you don't have to sit through this loading for 20 more seconds and voila check that out we have four great images I just love Leonardo's keynote XL model it does a fantastic job on human portraits I love it all right so I'm pretty happy with this result and I could go back to this prompt and continue to tweak it if I have a certain look in mind but again for the sake of time I like these results and I think this seed will work well for our consistent character an AI influencer okay so once you've molded your character it's time to access the image seed so that you could build off of these particular images again the seed will guide the AI on generating new images that align with the overall style of these current images and it's super easy to do trust me trying to create consistent characters with some of the other AI image generators out there is not as easy or as good as Leonardo and all you'll do is within the image set that you want to build off of click that 3. icon in the upper right corner and then from the drop down menu select copy seed this will copy the seed for this particular output and you should get a notification letting you know that the seed was successfully copied then within the sidebar on the left hand side of the screen scroll down toward the very bottom and click on where it says show Advanced features this will obviously give you access to the advanced features where we'll enter the seed so just make sure that that switch next to where it says used fixed seed is turned on and it will turn this purplish pink color when it's enabled then in the field provided simply paste the image seed that we just copied I'm on a Mac so I'm pressing command p on my keyboard to paste it next it's time to edit the prompt so that the AI will put our subject in a different setting close mood Etc but now that we've told the AI to use the fix seed it'll generate images of our character keeping the features consistent but giving you the creative freedom to change what they're doing for example I'm going to edit the initial prompt I'll keep all of the original text in place but I'll include black dress in a hotel lobby at the end of it so the idea here is that since we're now using the fixed seed from our original images the AI will use the physical features from our previous image generation from that fixed seed but put her in a black dress in a hotel lobby so let's see how it turns out so go ahead and click the generate button and again for the sake of time I'll speed through the image generation process so that you're not sitting here for 30 seconds watching a blank screen and check that out the fixed seed from our original image set has been used in the same character is now wearing a black dress in a hotel lobby again my prompt was super simple so that gave the AI a little wiggle room in terms of design and style but overall I'm really happy with these results the features and overall look of our Instagram influencers seem to be very similar from the fixed seed and I can definitely see a strong resemblance between all of these images now it isn't always going to be perfect but that's the beauty of AI you can tweak the prompt you can use different seeds and really mold the output until you get it to a place that you think works for your project okay let's keep going so I'm going to keep the fixed seed from earlier in place since I like how this consistency turned out but this time in the prompt field I'm going to keep all the physical feature descriptions in place but she didn't seem too happy in the hotel lobby so I'll add that she is happy and then I'll swap out the black dress hotel lobby with sitting at a candle lit table in a fancy restaurant downtown setting and then let's click the generate button to send that to the AI and we'll fast forward really quick and voila our Instagram influencer is now in a completely different setting and seems to be in a much better mood now there also seems to be some slight variations to the facial features and that's probably due to me adding happy to The Prompt which will happen from time to time but again that's okay keep testing new prompts until you get the output where you want it to be okay so let's keep working on this character this time let's put her at a beach party during golden hour in Southern California so I'll add that to the prompt and then click the generate button to send that through and again I'm fast forwarding through the image generation process to speed things up and perfect she resembles our fix seed and is smiling a lot more in these picks again there may be some slight variations in terms of hairstyle and whatnot but overall there is a really strong resemblance here between all of the images okay so that's the first method of creating consistent characters with Leonardo next let's go over the second method which is training your own model in this portion of the tutorial I'm going to show you the second method of creating a consistent character and that is by training your own Leonard model now if you're new to the concept all you're doing is uploading multiple images of your character to Leonardo's training and data sets and then from there Leonardo AI will basically use those images to build a model for you to use within the image generator again if you're scratching your head right now it's really super simple once you start going through the process so let's do that right now okay the first step in the process is you need to have multiple images that you want to train your model on and these images should all look similar to what you want the output of the model to be they should also be the same size sticking to a consistent size ratio example 768 x 768 pixels this ensures that your model can generalize effectively to new scenarios so what I'm going to do is I'm going to use the Instagram influencer images that we created earlier in the video that way Leonardo will have a good base for what the model should use when helping us generate a consistent character okay so that means I need to download all of these images and I should point out that as I'm filming this video in March of 2024 you have to download these images one by one however if you're watching this video in the future you can try clicking this 3 do icon next to the image set and from the drop- down menu you'll see the download all feature again as of March 2024 it isn't available but it is coming soon this will save you a ton of time in the future and maybe you're watching this video in the future and this feature is available if that's the case I recommend using this instead of downloading the images one by one however since I'm not a time traveler I'll need to download each image individually so to do that hover your mouse over each image and click this little download icon and this will download the image to your computer and you could then use it to train the model which we'll do in just a sec and again I'm not going to make you sit through me downloading each image one by one so let me fast forward through that really quick now I also also want to point out that Leonardo recommends using up to 40 images to train the model so try to aim close to that when uploading images to use for this data set all right once you've downloaded your images it's time to train the model so let's head back to the Leonardo home dashboard by clicking that little arrow in the upper left corner of the screen then within the sidebar left menu click on where it says training and data sets and this will take you to your model builder this is where you can create data sets view your job status and edit your current models as you can see I've already experimented with a model of myself and I got some pretty good results with it but it was for a cartoon character model so it probably wouldn't work for an AI influencer so let's train our AI influencer model okay so in order to do that click the new data set button next you'll be prompted to name your data set and give it a description this is for backend organizational purposes only it won't affect the visual output so go ahead and fill those two fields in so this will be my female Instagram character and then I'll give a little description here to help remind me of what the model output will look like there we go then click the create data set button once you're done filling those out next this will take you to the data set editor this is where you'll upload your images to train the model on and you have a few ways to go about this you can upload them from your computer here or you can drag and drop images from your personal feed and then add them up here and that will add the images to your data set it's whatever is easiest for you but if you recall I already downloaded each image that I want to use from my personal feed so I'm going to click where it says click here and then I'll do a bulk upload of all of those images so let me do that really quick and there we go the images are uploading to the data set and this is what we'll train the model with now I only have 21 images here at the moment Leonardo recommends 40 I apologize that I don't have that many for this example but again I would try to upload closer to 40 images for your model the more images the better but again for the sake of time I'm just going to use the 21 images that I have for this example okay once you've uploaded your images click the train model button towards the right hand side of the screen there and next you'll get to fine-tune the metadata and how you want to specifically train your model so first you'll be asked to give the model a name and description I'm keeping it the same as the data set so I'll name it female Instagram model and then the description will be the same as before then below that is where you can tweak and fine-tune the model to your liking and I should point out that we can always come back here and edit this metadata if you need to again giving you creative control over the output with some trial and error so first the training resolution if you click that drop down you'll see that there are only two sizes to choose from at least as of March 2024 there are only two to choose from and as I mentioned earlier in the video I probably should have gone with an image resolution closer to one of these when I was generating my images however the size of all my images within the data set are 512 by 768 so I'll keep it here as 512x 512 next you'll set the base model and this is what the AI will use as the base model one training and Leonardo recommends using the stable diffusion version 1.5 512 by 512 so I'm keeping that as is as well next you'll set the category and simply click the drop down and select the category that best fits your model I'm going to keep it set to General but I'll show you how to change it in just a bit you can easily configure this category within the image generator to F tune as you go all right next we have the instance prompt and this is basically a simple way to direct the model to properly utilize its training data framework simply put just enter the words that you want to use in your prompt to help tell the AI what to generate so again I'm using the words female Instagram model and these are the words that I'll use within my actual prompt to help guide the AI model when generating the image then finally you could set your model to be not safe for work this is for any adult themed content but ours isn't so I won't need to flip that switch on then when you're ready go ahead and click the start training button and keep in mind of the number of models that you have left to create which will be displayed right below the button just FYI all right then the AI gets to work and starts training off of that data set now this usually takes about 30 minutes and you can always view the job status by clicking this button here and you'll be able to monitor all of your current jobs that are processing and your completed jobs Leonardo will also email you when the job is done so you don't need to constantly monitor this screen just keep an eye out for an email that looks something like this it's pretty basic but once you get this email your model is ready to use and when you revisit the job status page you'll see your model data set has now a status of done and you can now use the model within Leonardo's image generator so real quick if we head back to your data sets by clicking that tab you'll see the new data set is listed here where if you hover your mouse over it you can edit the data set and even retrain it this is just another way to tweak and fine-tune your model if needed which we'll do in just a bit but for now let's put this to work and see what it can do so let's head back to the image generator by clicking on image generation within the leth hand sidebar menu then if you recall for our original method of creating consistent characters I was using the photo real pipeline which by default uses the Leonardo photo model so in order to use our new custom model I'll need to turn that pipeline off by flipping that little switch then next we'll need to select the custom model that we just made So within that little model icon click that Arrow to open the drop down menu and this will present you with a handful of current internal Leonardo models however if you click the select other model option all the way at the bottom of that list you'll have access to all of the other Leonardo models including your trained models and whenever you have some extra time I recommend experimenting with different models to see what works best for you and your projects but again for this example I want to use one of my trained models so if you click on the your models tab it'll give you access to all of your current trained models within your account now I currently only have two custom models but if you want to use one of them simply hover your mouse over it and click on on The View button then from here click the generate with this model button and you can now use this model to guide your image Generations also I do have the Alchemy pipeline turned on and these are my current settings again if you have a paid plan you'll have access to this Pipeline and we'll be able to configure the settings here if you like next I'm going to use the 5 12x 768 image Dimensions so I'll select that really quick and finally I want to switch the style being used it's currently set to Dynamic but if you click that little arrow there you'll be presented with various styles to choose from and for this example I think I'm going to go with photography so I'm going to select that really quick and I think we're ready to test this new custom model now one thing I want to point out before we get started is that we're not using a fixed seed with this model as you can see when we open the advanced settings the fixed seed feature is empty and turned off however I will show you how to add the fixed seed to the custom model in just a bit but first let's create some new images with our new custom model so once again in the prompt field I'm keeping a lot of the same prompt intact I'm also using the prompt instance text which is female Instagram model to help guide the AI when using this model however I will change the setting so I'll have our AI influencer be jogging in the woods this time okay let's send this through so click the generate button and when we fast forward through the processing check that out our female Instagram influencer is now jogging through the woods now I will say that I can tell that there are some slight differences in quality due to the fact that this custom model is using a completely different base model than what the data set images were created with and that's because the Leonardo Kino XEL model isn't available to use in the model training yet but either way I think we got some really good results okay next let me show you some quick tips to help tweak and fine-tune your model's results first let's change the model category if you recall when we first trained it we used the general category well let's see if changing that category has any effect on the image output so if we open the model dropdown and click on the select other models option then make sure you're in the your model section and find the custom model and hover your mouse over it and click on settings then this time within the category section click that drop- down and select photography and let's see if this changes the quality and style of our models images then don't forget to click the save button button to save these changes then let's use this updated model so hover your mouse over the custom model again but this time click the view button and then click the generate with this model button then I'll just keep the same prompt model and style in place and I'll regenerate the images by clicking the generate button then we'll fast forward through the processing to see the results and there we go now again the quality of these images does seem different from the Leonardo Kino XL model and her eyes seem to be a lot Bluer here and there also seems to be something weird going on with her legs in that one image but for the most part I'm really happy with these results the consistency is there and being able to do this all Under One Roof within the Leonardo platform is a huge time and money saver trust me okay let me show you another fine-tuning tip that I experiment with with and that was to use the fixed seed from our original image data set within the train model so if we scroll down to those particular images and click the three dot icon and then copy the seed then let's open the advanced settings and enable the used fixed seed feature by flipping that switch on and then paste the seed in that field again I'm on a Max on pressing command p on my keyboard to paste it then keeping all the settings in place I'll click the generate button again to regenerate a new set of images using the custom model and fixed seed and in a few moments we'll get the results so let me fast forward through it really quick and looks pretty good here as well again her eyes seem to be really blue when using this model but that's something I could tweak and work on whenever I have some extra time but either way I hope you learned something new in this video and you're starting to get some ideas as to how you can use these methods to create consistent characters with Leonardo for your upcoming projects again I highly recommend revisiting Leonardo's help docs on this topic they have even more tips and walkthroughs to guide you through different ways to approach this and remember this AI technology is moving at a light speed pace so by the time I publish this video something may have changed or are even easier to implement but whatever the case may be I hope you found this video helpful and as always if you have any questions suggestions or need some extra help feel free to leave a comment below and I'd be more than happy to help you out so that's going to do it for this video if you found it helpful I'd greatly appreciate it if you would like comment share and subscribe to the blog with bin YouTube channel also if you're thinking about starting a blog you got to check out my step-by-step tutorials they'll show you everything you need to know in order to build grow and monetize a professional WordPress blog and as always your support means a great deal to me and my family and for that I thank you so with that being said I'll see you in the next video and thanks for watching

Channel: Blog With Ben

Views: 30,982

Keywords:

Id: vRg1PdJQKRA

Length: 31min 43sec (1903 seconds)

Published: Sat Mar 16 2024