Best Consistent Character Tool? - 6 AI Tools Go Head to Head Including the NEW KORA Human

0:00 Hig wheel character, ideoggram character, Kora human, midjourney, flux context. So many consistent character creation AI tools. But which one is the most explosive? Let's find out. Over to Kush Sharma. The most explosive tool will be the one that creates the most real looking consistent characters. And with the introduction of a very new tool which is Kora human, we now have six competitors when it comes to this consistent character creation field which are Hicksfield character, ideoggram character, Kora human, midjourney omni

0:45 reference, flux context multi-reference also called as omni reference and runway gen 4 references. We will be ranking all these tools on the basis of five factors. The most important one being realism. So basically it should not have that AI look. It should be clear. There should be no quality issues. How well does it follow the prompt? Are there any artifacts like extra fingers and other common things that we experience in these AI images? And of course the aesthetics which means the colors, the composition and just the overall look. All these tools will go head-to-head with each other for seven rounds in which these seven head shot will be used where we will be turning these people into different characters and whichever tool performs the best in each round is going to get the points and at the end of it we're going to see which comes out

1:32 on top. As always, I've given you all the images and the links in the description so that you can work along with me also because this is slightly subjective as to which image is the best. I've also given you this document which will have all the images that you're going to see in this tutorial so that you can take your own call later on. For the first time when we do this, I'm going to give you a brief walk through for each of the tools so that you understand how they work. But from second round onwards, I'm directly going to show you the images just to save time. So let's get started. So this will be our first image that we're going to generate. You can see the headsh shot. You can also see the prompt. This prompt was generated inside chat GPT. All the prompts that you're going to see here have been generated inside chat GPT. The part in yellow just gives you the

2:18 summary of the prompt. So even if you read that, that's more than enough. So here we are basically turning this guy into a barista. So let's start off with ideoggram. So Ioggram's character tool was just released a few weeks back and it's a very simple tool to use. Once you go over to Idoggram, you're going to go over to character. Then we simply upload the photograph of this man right here. Let's paste in the prompt. And for this, we'll go for an aspect ratio which is like a mobile uh shot which is 9 by 16. And that's about it. You really don't have to do anything else. We're going to hit generate till the time this is coming up. Let's go on to the next tool which is going to be inside the enhancer app. So the Kora human tool, the new tool which has just been released last

3:03 week. This is by the same company who is behind the very popular enhancer upscaler. You're going to go over to Kora and here first of all make sure they have different versions of this Kora generator. You're going to select Kora human that is the one for the consistent characters. And under here we are simply going to upload the image again of the headshot. I'm going to click on done and that's it. You just get one photograph. After I paste in that same prompt, let's change the aspect ratio. And that's it. So, we're going to hit generate. This is going to come right here. Let's go on to Hickfield. Hickfield works in a slightly different way. By the way, I have a dedicated

3:49 tutorial on Hickfield character, which is this one. In case you haven't seen it, I will leave a link to this in the description. So, you go over to create after logging into your account. Then you go over to image. And here instead of uploading one single shot, you first actually have to create what we call as a character. So here, if I just open up this character window, you just hit create character. And you can even input one image, though more the better. But even if you just have one image, then you create the character. It's going to cost you 30 credits. Keep that in mind. So it's not uh extremely cheap, but the generations from there on are very cheap with just uh one credit. Okay? And to use the character tool, you will have to be on the paid plan. This cannot be done on the free plan. So again, we're going to write in the same prompt here. We're

4:36 going to select 926. This is using the sole generator. Basic quality is fine. And I'm going to choose the new character here. And that's it. Everything is set. I'm going to hit generate. And this is also coming up. Now let's go over to Runway. Good part about Runway is that on a free account, you will be able to do this a couple of times. So once you log into your free account just make sure you go over to the image tab add the references image here just like you add the headshot anywhere else and once this image is reflected here you can just put in the same prompt we can select 9 by6 and we can select 1080p is going to cost us extra three credits but that's fine the image will look much better and this was this will also generate this image right here let's go over to flux flux context

5:23 can be accessed using a lot of different all-in-one AI apps, but the cheapest way to access ites access it is by using the Flux Playground, which I have actually shown in this video where I showed how to create consistent characters, which was dedicated to Flux Playground. But again, the good part here is on a free account, you will be able to do this 50 times because you get 200 credits and each edit will cost you four credits. So once you're here, you're going to go over to generate. We're going to input our headsh shot right here. put in the same prompt and select the aspect ratio. Now, when it comes to the aspect ratio on Flux Playground, you have to remember sometimes it just follows and sometimes it's just going to give you a completely different aspect ratio. We are not in control of that. This is a bug which

6:09 still now hasn't really been rectified. So to solve this, you will just have to use it on some other all-in-one AI app which is using the flux context API like let's say freick or open art AI. Let's go on to our next tool which is going to be midjourney and here we're going to be using the omni reference feature. Again midjourney is only available for the paid users. So first of all I'm going to hit this. I already have the headshot here. So I'm going to select this and not here actually. So just make sure that you drag this over to omni reference and then we're going to paste in our prompt and under settings here we're going to select 9 by6. By the way again I have a

6:55 dedicated tutorial on creating consistent characters using midjourneys omni reference. If you haven't watched the video this is the one. I'll leave the link to this also in the description. And that's about it. We're simply going to hit generate. So this is going to come right here. So now that all these tools are generating these images, let's go back to ideoggram and let's see some of the results here. So you can see IDOG results look really really amazing. So let's get a better look at all these images by starting off with round one. When I had done this before, I picked out a winner for each of the tools. This was for ideoggram. You can see really looks like a proper real photograph. Was very impressed by this. This was by Hicksfield. This also didn't look bad, but the problem was the skin had a lot of artifacts.

7:42 Then this was Kora. Again, looked pretty good, but still had a bit of that AI look. This was runway, and this almost seemed uh it had that animated look. Flux, again, the quality of the image wasn't that great, even though the consistency wasn't bad. And finally, we had midjourney here. And when I evaluated all these on the criteria that I had mentioned before in this video, the winner here was ideoggram. So the score right now is that ideoggram is one and it leads all the other tools which are currently at zero. Let's go over to round two. So for round two, this was our headsh shot and the basic premise of the prompt was that this was a de desert scavenger warrior and he was armored and

8:27 basically like a mercenaryike uh look to it. This was what I produced and you can see this was absolutely stunning. Let's see Hicksfield. Hicksfield just doesn't do very well with these cinematic looking shots or you know anything too aggressive. Basically he's there for everyday regular lifestyle shots and therefore you can see that he's just wearing a jacket. Doesn't look like a warrior at all. This was what Kora gave out. Very similar to the ideagram result. Was very very impressed by this one. This was runway and it just messed up the hair. Also had a very animated look to it. This was flux. I think the same problem. And finally we had this by midjourney which was no surprise because midjourney became popular because of the

9:14 cinematic look. But remember we are judging things on the basis of realism. And therefore the main competition was between Kora human and ideoggram. And the only reason I chose ideoggram here also as the winner is because when you compare it to the headsh shot to the structure of the face of that real person, it just was slightly better. So right now the score is ideog two. Remaining ones are all at nil. Let's go to round three. So this was a slightly challenging prompt. This is the headsh shot of the woman and basically the whole idea was to turn her into a model who's wearing a pink leather suit and there are people around her as she as she walks on the ramp. This was ideog result and for the first time here I saw that ideoggram failed to produce a shot

9:59 with any kind of realism because this had that typical AI look. So this was pretty surprising. I was very very impressed with Hicksfield. But you're going to notice one big issue with Hicksfield when it comes to multiple people which is that if there's more than one person in your shot apart from the character that you have trained soul on it's actually going to give that look to the other characters also. Can you see right behind is the same person even though the consistency was absolutely amazing. But just see all these people are also they have that same face irrespective of the gender. So even the men they have the similar face. This is a huge issue with Hicksfield that you simply can't use it for multiple characters currently and therefore such a scenario becomes of no use even though the quality actually looked much better

10:46 than the other ones that we have seen uh till now. Let's go over to Kora. I was really really impressed with this. The consistency would was good. And the only downside was we couldn't see too many people here. Uh but apart from that, I think this looked really really nice. Runway again, it's just, you know, all over the place. You could see the people here. They have that typical AI face. But the main problem here was there was no consistency at all. This wasn't the same woman. Let's have a look at Flux. Not bad, but this definitely is an AI image. And with Flux these days, I've seen that the quality of the image just isn't very good. So you can see it's already pixelated. Finally, this was midjourney. Again, this just did not look good at all. So here the winner was Kora human. Let's

11:32 move to round number four. And this was a challenging prompt. This was supposed to be a full body portrait of a man standing next to a sports car. and pretty much every tool out there struggled with it when it comes to following the prompt as you're going to see. So, first of all, this is ideoggram's result and you can see this wasn't a full body pro portrait even though it looked very very real. So, from a real realism point of view, this was good, but it just didn't follow the prompt at all. Let's have a look at Hicksfield and you can see this just looked terrible. His head was way too big for his body and overall everything just didn't look good. Uh, Kora, not bad at all, but the pose that he was giving just uh wasn't too great. Then you had Runway, which pretty much looked like he

12:17 was against a background. This didn't look like a real shot at all. Even though uh just going from the consistency point of view, for the first time, Runway actually did a good job because he didn't look uh animated. This was a good result by Flux context. However, if you just look at the proportions here, it's totally off because the car just looks too small. The best result here was by midjourney undoubtedly because here everything was correct. Again, not a full body portrait, but if you compare it with the other shots that we've got, the proportions are right. The consistency was good. And at this point, I can give you a bonus tip which actually slipped my mind when I was creating this image. If you are going for a full body portrait, do mention it in the prompt, the type of shoes. just describe the type of shoes and that forces the AI

13:04 tool to then generate those shoes and therefore it's usually going to be a full body portrait. But like I said here, the winner was midjourney. So midjourney is also now on the table. So for round number five, I wanted to go for something which looks like an amateur shot, something that you know you associate with just taking out your cell phone and clicking. So we wanted this man in a party surrounded by his friends and it should not look professional at all the shot. So this was ideog result and I really loved it because this looked like a natural photograph. Right now if you show this to anyone nobody can say that this is AI and the bit of blur on his body just made it uh you know look really really real. Uh the result by Hicksfield wasn't bad but it still had that AI look but again we have that same problem.

13:50 Multiple people they all have the same face so it serves no purpose. Then we have Kora. Kora also produced a very real looking image. It's just that it didn't follow the prompt because we could not see his friends around him. And I actually tried to generate this multiple times because I was very impressed with the realism here, but it just did not produce those friends in uh front of him or around him. Then we had runway. No point at all. Flux again, it didn't even follow the aspect ratio like was the case before. Just had uh you know, a lot of clarity issues and overall the shot wasn't too impressive. And finally, we had midjourney. Again, from an aesthetics point of view, the colors, this had that cinematic feel to it, but we were not going for that anyway. And even the consistency wasn't great. So, here it was mainly between

14:36 Kora Human and Ideog. I think Idoggram's result not only looked more real, it also followed the prompt. Remember, both of these are factors. So, therefore, here the winner again was Idoggram. So, Idoggram just increased its lead even further. Let's move to round six. So round six was about this head shot and she's just sitting in an outdoor cafe. Again, a very casual looking shot. This is what Ioggram gave us. Not bad at all. Hicksfield from an aesthetics point of view, this was good, but just had that polished AI look to it. Was very impressed with Kora human because this actually looked like a proper photograph and the consistency was also good. Runways again is just gives that animated look to the shots.

15:22 uh which rules it out in most cases. Flux again this time it didn't even follow the consistency. So uh it's not there at all. Midjourney the aesthetics the composition again was brilliant like is the case most of the times but again it does suffer from consistency and it just didn't look too much like that original woman. So here it was mainly between ideoggram and Korahum again and this time Korah human was the winner. So right now it's ideoggram 3 korah human 2 midjourney one. Let's see one more round which is going to be the final one. So this time this was a challenging pose because this woman is doing yoga in a garden. And sometimes these AI tools when you give them anything complex especially involving the human body the

16:08 physics they do mess things up as you're going to see. But I was very impressed with ideoggram. Only issue here was that her uh leg, this one, if it could have been more uh straight, it would have been better because it just feels like she's tilting towards the left. But the consistency was absolutely amazing. And overall, this was a brilliant result. Uh Hicksfield was extremely disappointing. The realism was there for the first time, but it just doesn't look like she's doing yoga, so it's no point. Kora was similar to ideoggram, but the consistency just wasn't there uh this time. If you just look at the original headsh shot and also her feet had uh you know few of those artifacts and runway and this was just not even in the picture again flux not bad but too

16:54 pixelated. So like I said it does suffer from these uh issues a lot and the quality of the image is just not good. Finally we had midjourney which just produced a completely different woman and overall it just didn't look good. So again here the winner was Idoggram and the final score is ideoggram wins this contest with four. Runners up is Kora human with two and in third place is midjourney with a single point. But here are also some important things that you should know about these tools. So ideoggram wiped out all the competition as you saw. But let's just talk about some of the tools here. I think we can safely ignore Runway. The company's more focused towards video. They're doing very well there and I don't think those images are going to improve anytime soon. Flux Context can be a good backup.

17:42 The quality is an issue, but considering that you get 50 edits for free, if none of the other tools are working, you can always use this. You have nothing to lose. Midjourney, I would say, is more important to people who pay more attention to uh the aesthetics rather than the consistency. And especially if you're going for that cinematic look, Pixel surprisingly scored a zero on this. Uh this is a tool I use a lot personally and it's not that bad. That's what I would say. One thing you have to understand about Hicksfield is that you need to really train the character with multiple images. It's not like the other tools where you simply upload one image. So the more the images that you have trained the character on or the Hickfield soul model on the better is your chance of getting a good output. However, even with those good outputs I

18:27 have seen with Hicksfield, they do give out images in which you have that polished AI skin look. And to get rid of that, you will then definitely have to use an upscaler like enhancer. Let me show you an example. So this was an image that I generated inside Hicksfield using the same character that you saw inside round one. And I've just put it through the enhancer upscaler. If I just expand this, you can see this is what we started off with. And you can just see it has that polished look on the skin. But after putting it through enhancer, just see the difference. So with Hicksfield, you'll definitely have to do it. In fact, with most of the tools, you might just have to do it anytime you get that kind of a skin. Kora human can be a good option because you then get enhancer also along with it. But with Kodah human, what I've seen right now is

19:13 that it works great when you're creating images in which your character is filling up the frame. But anytime there's too much of negative space around the character, like let's say in a full body portrait, it really really struggles with that. Ideogram, as you saw, was the winner simply because it produces the most realistic looking images. as of now. But my ideal strategy would be to go for both an the enhancer app or basically Kora human as well as ideoggram. So that if anytime you are not able to get a good image inside Idogram first of all you can try kora human we saw it can produce a good job but secondly anytime ideiogram gives you an image which doesn't look too real remember you can put it then through the enhancer app. Now before I close this video, I just want to talk about one more new tool which actually came during

19:58 the process of creating this video tutorial. Just yesterday, a new tool called Nano Banana was released which is a competitor to Flux Context and I just wanted to try this out because they do allow you to use it two times for free. So I'll leave the link to this in the description uh because this tool is making a lot of noise. It is primarily an editor. So if you have an existing image, you can edit anything by writing a prompt. But if you go over here to try generator, you can use this image to image model where you upload the headsh shot right here and then write in the prawn. So I did this for two of the rounds, the challenging ones that we had seen in this video. And these were the two images that it produced in those rounds. And you can see that I don't think that this was an improvement over either ideoggram or Kora human. These

20:45 look more like the images that we got out of flux context itself. But I definitely plan to make a dedicated tutorial on this new tool. So in case this video helped you out, do give it a like. And for more AI editing tutorials like this one, make sure you subscribe. And I will see you next