AI Lessons: Secret AI Art Feature to Replicate Poses
Welcome to testing AI on this channel. I break down different AI tools and keep you up to date with the AI world. If you like this video, make sure to hit subscribe.
Hello everybody! In today’s video, I’m going to go over an amazing feature that Leonardo AI has that a lot of other tools don’t have, and that’s called ‘image to pose.’ What this means is you can use an image and replicate that pose in your own generations. So today, I’m going to replicate my favorite style but with four different poses. Here are the poses I’m going to use today – this is the first, second, third, and fourth. Like all my other videos, I’ll have the timestamps mentioned in the description below so you can get to any section directly. Alright, let’s get started.
So the first thing is, if you don’t have access to Leonardo AI, they have had a massive influx of users, so it might be a bit longer for you to get an account. And if you don’t know how to get an account, make sure to check my first video where I show you how to get early access. An important thing to mention is if you’re starting from scratch and click an image here to upload, say, this pose, you’re gonna see that there’s no ‘image to pose’ toggle that shows up. That is because currently, their model has been trained only on ‘Stable Diffusion 1.5.’ So, if you change the model to ‘Stable Diffusion 1.5’ here, you’re going to see that this ‘pose to image’ toggle shows up.
Another important thing to mention here is that the image that you use to pose here ideally should be a full-body picture. Because I’ve tried some other pictures which are not full-body, and the images are not as good. For today’s video, I actually want to replicate this beautiful design that I like. So, it’s a beautiful female Alchemist, and there’s a lot of prompt details already mentioned. What I’m going to do is I’m going to click ‘remix,’ and if you don’t know what that is, make sure to check my video here. I used Playground AI for that video, but it applies here too. Basically, what the remix feature does is replicate the entire prompt for me. To use this specific prompt is actually really long, so you see how long that is. What that means is it’s going to take a long time for me to generate, and it’s a lot of load on the system. What I’m going to do instead is I’m going to keep a portrait of a beautiful Alchemist and delete everything else. I’m going to keep all my prompt exclusions in place. If you don’t know what that means, check this video here. Basically, I want to make sure that none of these show up in my pictures using the negative prompt feature before I use the ‘image to pose’ feature.
Let me do a first generation just with this prompt. I’m also going to change the number of images to one. Cool, so I have a portrait of a beautiful female Alchemist right here. Now I’m going to select my first pose from this uploaded image option. You might see that here I did not change the model to ‘Stable Diffusion 1.5.’ That’s because I’m using the remix feature and not starting from scratch. So if you’re starting from scratch, make sure your model is set to ‘Stable Diffusion 1.5.’ Cool, so I’m going to click the toggle ‘pose to image,’ and I’m going to keep everything the same, generating one image of a beautiful, stunning female Alchemist in this pose.
Now, isn’t that amazing? This is the pose I’m trying to use, and this is the female Alchemist in that same pose. She even has her arms on top of each other like here. There’s some issues with the eyes, but that is unrelated to this feature. Now, let me try the second pose. So, I’m going to delete this one. I’m going to use this pose this time. Everything else stays the same, portrait of a beautiful Alchemist using this pose.
Now, isn’t that amazing? This is the pose I’m trying to use, and this is the picture I get. Of course, there’s some issues just like my model right here, my Alchemist also has her right hand on this leg, and her left hand on the waist. There are slight variations on how her arm is, and there might be some issues with the picture that need to be fixed. But the ‘image to pose’ feature is working pretty good. Let me try the next pose.
This time, I’m going to use this one. Make sure my ‘image to pose’ is selected, click generate.
This is so cool! This is another female Alchemist with this exact pose, and I don’t expect Alchemists to be in this pose, so it’s definitely very unique to this picture. I’m just blown away at how accurate this is. Just like my model here, this Alchemist also has her legs crossed, has her left leg on her head, and she’s looking towards her right. I think this feature alone is going to make Leonardo AI so popular.
Alright, let’s try the last pose. So, the pose I’m going to try is this one. Make sure my toggle is selected. Keep everything the same, click generate.
I think this one is not quite there all the way. However, the upper part of the body looks pretty close to this pose. I think due to the lighting of this part of the body, Leonardo might be struggling to identify that it’s her backside that is facing us. But I can feel Leonardo trying really, really hard to replicate this pose. You do need to play around a bit to get the perfect image. But make sure to know that if you put a really long prompt here with a lot of exclusions, create four images, and also put ‘pose to image,’ then your image generation might take quite a few minutes. That is why, for this demo, I made the prompt a bit simpler and the number of images to one because I really wanted to focus on the ‘image to pose’ feature versus the actual design of the image, which also looks really cool, by the way.
So I hope you got some value from this video, and I hope you try out the ‘pose to image’ feature. I think they’re going to add a lot more improvements to this. This is just the first version of it, and it’s already really good. And I’m really excited to see where this goes as this tool gets better. I highly recommend you getting a paid subscription because I think there are so many really cool features in Leonardo AI that are not just basic, they’re really, really unique. If you enjoyed watching this video, make sure to click the like button and hit subscribe. Till the next video, thank you so much!