If you're interested in how to create scenes with multiple consistent characters, check out this video guide I made here: th-cam.com/video/tdAkcKXG99Y/w-d-xo.html
@@Warmer_Bros I think Midjourney will get you better looking images with more prompt control compared to Leonardo. But either would probably be okay. If you mean using the Leonardo motion Ai, it's pretty restrictive and I don't think it has lip sync.
Dear Tao, I am quite an Old-Schooler, but your video is outstanding. You name everything super acurate, and you show simple but perfectly how it works. Thank you. This is really helpful
You've explained how to create consistent characters so well and clearly, thank you for the valuable tips! I’m curious about your thoughts on making the locations just as consistent. For example, ensuring that the character's background and the environment stay the same, but are shown from different angles throughout the story. How can this be achieved, so that the setting also aligns with the narrative and characters, while being presented from various perspectives? Would you consider making a video about this?
There's not really a good way to view the same background from different angles, you'd have to construct a real 3D environment for that. But you can keep the visual style of the background similar with style reference
Both work best with images generated by their own Ai models, and both work better with realistic photos. Leonardo gives you more flexibility and control over the pose. I think Midjourney generates a larger variety of different looking people that you can try generating consistent characters for, and the images are more photorealistic. Leonardo generates a smaller variety, but it seems like the characters generated are a bit more consistent overall. I will have to do more testing to confirm that though.
Your tutorials are so in depth and absolutely amazing. I have a simple question... at 2:38 mark, as there are 4 images from left to right... how do i prompt Leonardo to make an image front facing (with both ears are visible) like the very left image, not 3/4 view where only one side is showing towards the camera? Thank you!
Thanks for taking out time to fulfill my video request for consistent characters in Leonardo Ai. Great tutorial as usual. I have watched it thrice it already. I do like the images generated by Leonardo but clearly some work still has to be done in the area of prompt adherence. I will try playing around with the settings some more in case it's user error. Midjourney I love but find it prohibitive having to pay $60 just for private generation privileges. Thanks once again for the tutorial. By the way, your Midjourney tutorials are one of the best I have come across. Keep Up the good work!
Thank you! Yes, Leonardo models still need some work but I like that they are always adding/testing new features, even if they're not perfect. Midjourney seems to have slowed down a bit on feature releases, maybe they are just focused on version 7.
Great video, subscribed. If you could make a video where you talk about the use cases for all these features Leonardo has, I'm having a hard time wrapping my mind around the ways to use them lol
I’ve always used Midjourney, but watching your videos has me really interested in switching to Leonardo. My primary interest is to upload my own artwork and then “train” Leonardo on my style and the unique characters - but the characters are drawings, obviously. Will Leonardo be able to do this with drawn figures as well?
I'm not sure, they do have a way to train a model for your own characters, although I think this probably works better for Realistic photos. You'll probably also need 5-10 images of your character for best results. I'll test this out at some point though.
Leonardo has better consistency but with a much smaller variety of images. If you want very generic looking characters, it can give you decent consistency. But for more variety, Midjourney is better.
I have a question: if I use a posable mannequin to take a photo from the pose I want and upload that as the pose-reference to the portrait: would it work, too? Or does Leonardo only recognize pictures of "real humans" (be it AI generated, drawn or natural)?
Leonardo has a hard time keeping the clothing consistent. here's a tutorial on finetuning a Flux model to get consistent clothing and different scenes: th-cam.com/video/5Z8fwEeWfRg/w-d-xo.html
Hi friend, I generated a model using Leonardo and then used the character reference function to create more consistent images of the character. However, the images I created afterward don’t resemble the original model (even though I selected the levels low, med, and high). Can you share some tips on how to make them look consistent, like in your video
The more "generic & basic" the model looks, the better character reference will work. Make sure you are using the same image model & settings. You can also include more details about exactly what the character looks like in the prompt
The character reference feature will work better if you use it on an image generated using the website. You can try it on your own image, it just won't work as well.
You can use reference images from a website like Pexels. I would take advantage of pose reference and content reference to try to get the exact character photos
Hi friend, ex: I have a traditional black Vietnamese attire, I would like the leonardo create character but wear the same my clothing, we can do that?, thank for your great help
You could try using a photo of the clothing as an image reference and describe it in detail in the prompt. However, it won't be able to create anything that exactly matches the clothing you have.
But if you copy poses from different pictures, then you end up with different pictures, right? The body isn't always the same. Sometimes the shoulders are broader or the arms are longer. What do you think?
Keeping the same outfit is difficult. You could try being more specific in your prompts and describe very closely what the person is wearing. But it will be hard to it to look exactly the same.
Yeah, you can use the Leonardo motion Ai for very basic animations or try a more advanced ai video generator like Luma Dream Machine. Just use the image-to-video features
as you mention the problem with this face swaps and consistent characters are that it doesn't perfectly matches the textures and certain aspects of the image, i've generated a image from realismengine sdxl model and when i put in leonardo it always give me dark spots and unrealistic skin tones..
This feature will work the best if the reference image uses the same model as the consistent character images. You could try using a style reference with skin tones more similar to what you want, but Leonardo does tend to have a smooth/airbrushed look.
Leonardo Ai isn't the best way for consistent characters right now, something like finetuning a flux model can work better. I have a guide for that here: th-cam.com/video/5Z8fwEeWfRg/w-d-xo.html
@@taoprompts Thank you, but I bought a subscription, and now it's a waste. Can I create a good manga using Novel AI or Open Ai with consistent characters?
One key reason why it doesn't work well on our own images is that it's a technical challenge. Basically you would need to make minor updates to the Ai model itself in order for the "character reference" feature to work well with your own images. This can be done in Leonardo with the "train model" feature, but right now it's still not super consistent. I may do a video on that in the future.
Download the image generated using your own photo as a character reference. Then go to mid journey and try using the face swap plug-in on that image I’ve had pretty good results.
@@Author_SoftwareDesigner I tried this method in Stylar, but I think it looked bad because the picture I used was a selfie :D not that good resolution/quality. So it came out a bit grainy. I guess we should have good pictures with different expressions for that?
after trying on multiple characters, leanardo failed to generated character with consistency, it just following one girl which is similar to girl shown on video even i tried with mutiple faces but no use getting similar girl image not what i want,
The issue I have with Leonardo is its rigidity. You can't really create complex images. Just one character doing one thing... It's not really good for expressive multi character art
If you use a reference image with two characters with different poses as input to "pose reference", it can do an okay job of using both poses. But it does get mixed up between the two characters quite a bit. I don't think that works with multiple consistent characters though.
@@taoprompts actually i mean flexibility of its creation in general... "E.g create a cinematic close up side shot of a dark skinned native African man with scars on his body, wearing glowing gold bracelets, in an archer's stance/pose. the bow is made of gold light, and there is no arrow. His left arm has a unique, mystical, mutated feature that stands out. The background is a motion view. The scene should have muted colors, a fairy tale style, comic book illustration, with a vertical aspect ratio." will never come out as you want in Leo
@@MedievalFantasyTV I don't have one. My content uses ai images and i would really love my characters to have same face while still doing complex actions. It's why I jumped on this video. But Leonardo only gives same face, but can't really create some complex actions.... so we wait for more upgrades from other ai generators.
My results are literally nowhere close to that good. My character never looks real, like some weird exaggerated version of them. Tried different times with different base models, ugh
No one. Even your self do not talk about camera angles for interiors on Mid Journey. Its very hard to what you want. Why not do a video on that Tao. Honestly I only see video about camera shots and people.
I don't think that's really possible in Midjourney or any other Ai art generators right now. You would need real 3D understanding to accomplish that. You're best bet would be to use a 2D to 3D tool, take snapshots from different positions, and then use those as backgrounds.
If you're interested in how to create scenes with multiple consistent characters, check out this video guide I made here: th-cam.com/video/tdAkcKXG99Y/w-d-xo.html
But this method is not working in my cell. is it not for free Ai Leonardo users ? Please reply 😢
@@QisaeDil The advanced reference settings are restricted to paid users.
You teach in seven minutes what others take half an hour. And I appreciate that.
Thanks, I try to keep things concise and simple. Great to hear you found this helpful.
@@taoprompts
For anime style short movie, would you go Midjeurmey + Pika or Leonardo?
Considering it's cartoon, lip sync etc.
@@Warmer_Bros I think Midjourney will get you better looking images with more prompt control compared to Leonardo. But either would probably be okay.
If you mean using the Leonardo motion Ai, it's pretty restrictive and I don't think it has lip sync.
Dear Tao, I am quite an Old-Schooler, but your video is outstanding. You name everything super acurate, and you show simple but perfectly how it works. Thank you. This is really helpful
Hey, I'm happy to hear you like these guides! I try to cover topics that I think people will want to see, so it's good to know this helped you out 👍
Thank you! Your channel keeps being one of my favourites ones: straight to the topic, interesting, useful and very well edited!
Thanks I appreciate that, I love the way you make your thumbnails btw.
You've explained how to create consistent characters so well and clearly, thank you for the valuable tips! I’m curious about your thoughts on making the locations just as consistent. For example, ensuring that the character's background and the environment stay the same, but are shown from different angles throughout the story. How can this be achieved, so that the setting also aligns with the narrative and characters, while being presented from various perspectives? Would you consider making a video about this?
There's not really a good way to view the same background from different angles, you'd have to construct a real 3D environment for that. But you can keep the visual style of the background similar with style reference
@@taoprompts Thank you!
Finally there is a video on this topic, that is easy to understand, with real examples ! 🧡🧡🧡
Thank you very much for this detailed and very well explained tutorial. Your video has actually made me want to try it myself.
It's always worth trying out new tools 👍
Super grateful for this video! Thank youuuu 🤩
Glad this helped Irene!
Ideogram doesn't have this feature at the moment. Well done to leonardo. The competition is always great to consumers. 👍
Clear and well-done tutorial! Thank you!
AMAZING RESULTS !!!
Thank you!
I was eagerly waiting for this tutorial. Thanks very much 🙏
💯, this was a great feature release, leonardo has really stepped their game up.
A well-needed update!
So how would you rank Leonardo vs Midjourney for consistent character creation?
Both work best with images generated by their own Ai models, and both work better with realistic photos.
Leonardo gives you more flexibility and control over the pose.
I think Midjourney generates a larger variety of different looking people that you can try generating consistent characters for, and the images are more photorealistic.
Leonardo generates a smaller variety, but it seems like the characters generated are a bit more consistent overall. I will have to do more testing to confirm that though.
@@taoprompts Thanks for sharing your insight.
Thank you for sharing your knowledge.
Your tutorials are so in depth and absolutely amazing. I have a simple question... at 2:38 mark, as there are 4 images from left to right... how do i prompt Leonardo to make an image front facing (with both ears are visible) like the very left image, not 3/4 view where only one side is showing towards the camera? Thank you!
Thank you, Try using "centered view" or "front shot" to get the front facing images
Super thorough and amazing video thank u love your content def stands out learning new tricks tips every time thank u
Thanks Agness, I appreciate that. And it's great to hear you find these guides helpful 👍
Man that helped me alot , thank you
Glad to hear this helped 👍
Thanks for taking out time to fulfill my video request for consistent characters in Leonardo Ai. Great tutorial as
usual. I have watched it thrice it already. I do like the images generated by Leonardo but clearly some work still has to be
done in the area of prompt adherence. I will try playing around with the settings some more in case it's user error. Midjourney I love but find it prohibitive having to pay $60 just for private generation privileges.
Thanks once again for the tutorial. By the way, your Midjourney tutorials are one of the best I have come across. Keep Up the good work!
Thank you!
Yes, Leonardo models still need some work but I like that they are always adding/testing new features, even if they're not perfect.
Midjourney seems to have slowed down a bit on feature releases, maybe they are just focused on version 7.
Thank you, It was great
Glad you liked it Sean, Leonardo Ai is really starting to pick up steam.
You fixed my biggggggg problem ! thanks bro , so appreciate it
For sure man, glad this helped you out 👍
Thanks For The Information": Great Video! Subbed!
Thanks for subbing! Great to have you here.
Great video, subscribed. If you could make a video where you talk about the use cases for all these features Leonardo has, I'm having a hard time wrapping my mind around the ways to use them lol
I will try to make some more guides covering what Leonardo can do, it's a really cool tool.
Can you do multiple pose ?
Sort of, I made a guide for that here, it won't be able to do complicated interactions though
th-cam.com/video/tdAkcKXG99Y/w-d-xo.html
Thanks for the tips
💯 This was a huge update for Leonardo
I’ve always used Midjourney, but watching your videos has me really interested in switching to Leonardo.
My primary interest is to upload my own artwork and then “train” Leonardo on my style and the unique characters - but the characters are drawings, obviously. Will Leonardo be able to do this with drawn figures as well?
I'm not sure, they do have a way to train a model for your own characters, although I think this probably works better for Realistic photos. You'll probably also need 5-10 images of your character for best results. I'll test this out at some point though.
@@taoprompts I’m getting started too. So if I have any results I’ll hit you up. :)
Thank you
Would you say it’s easier in this app over midjourney? I’ve tried for days on mid journey and can’t keep my character or style consistent
Leonardo has better consistency but with a much smaller variety of images. If you want very generic looking characters, it can give you decent consistency. But for more variety, Midjourney is better.
I have a question:
if I use a posable mannequin to take a photo from the pose I want and upload that as the pose-reference to the portrait: would it work, too? Or does Leonardo only recognize pictures of "real humans" (be it AI generated, drawn or natural)?
It does work with mannequins, however like I mentioned it won't be able to do very complex poses well
Good Tutorial keep it up
Can i made consistent face and clothes in different scene?
Leonardo has a hard time keeping the clothing consistent. here's a tutorial on finetuning a Flux model to get consistent clothing and different scenes: th-cam.com/video/5Z8fwEeWfRg/w-d-xo.html
Hi friend, I generated a model using Leonardo and then used the character reference function to create more consistent images of the character. However, the images I created afterward don’t resemble the original model (even though I selected the levels low, med, and high). Can you share some tips on how to make them look consistent, like in your video
The more "generic & basic" the model looks, the better character reference will work. Make sure you are using the same image model & settings. You can also include more details about exactly what the character looks like in the prompt
Can I apply these explanations to an image from my device, or does it have to create an image from the website?
The character reference feature will work better if you use it on an image generated using the website. You can try it on your own image, it just won't work as well.
Is there a way to have a close resemblance to the original photo from my album, for example?
I hope we will also get the ability to pose multiple consistent characters together in Leonardo.
I think that should be possible in the future, right now that is still very difficult to do.
how we can i get, all kind of character clothes, poses, camera shots, and facial and hand expressions
You can use reference images from a website like Pexels. I would take advantage of pose reference and content reference to try to get the exact character photos
amazing video
Thanks, glad you enjoyed 👍
Obrigado! Desde São Paulo, Brasil.
De nada! Quero visitar São Paulo um dia.
@@taoprompts A cidade de São Paulo é grande e o Estado de São Paulo é um país... 44 milhões de habitantes, será bem vindo.
thank you
15k 🔥🔥🔥🔥🔥🎉
It's hard to believe how fast we're growing 🙏. And always great to see you in the comments 👊
Hi friend, ex: I have a traditional black Vietnamese attire, I would like the leonardo create character but wear the same my clothing, we can do that?, thank for your great help
You could try using a photo of the clothing as an image reference and describe it in detail in the prompt. However, it won't be able to create anything that exactly matches the clothing you have.
Hi I put in my prompt in Leo photo of a 18 year old blonde At Miami wearing a swimming suit and a red filter block came up how do I get by that ?
You could remove the phrase "18 year old".
Is this feature available for free users?
Yes, this feature is free to try.
But if you copy poses from different pictures, then you end up with different pictures, right? The body isn't always the same. Sometimes the shoulders are broader or the arms are longer. What do you think?
There's no perfect way to map characters -> poses with Ai image generators unfortunately. There will be some slight to noticeable differences.
Do you take requests via direct message?
You can dm me on instagram, if you want to connect @taoprompts
Character reference is not available, it says coming soon
I think they made it a paid feature now
please reply, can leonardo ai create consistent character with same outfit?
Keeping the same outfit is difficult. You could try being more specific in your prompts and describe very closely what the person is wearing. But it will be hard to it to look exactly the same.
Is it possible to make a video using this character?
Yeah, you can use the Leonardo motion Ai for very basic animations or try a more advanced ai video generator like Luma Dream Machine. Just use the image-to-video features
how to generate multiple images with the same clothing?
It's not really possible in Leonardo, you have to finetune a flux model
as you mention the problem with this face swaps and consistent characters are that it doesn't perfectly matches the textures and certain aspects of the image, i've generated a image from realismengine sdxl model and when i put in leonardo it always give me dark spots and unrealistic skin tones..
This feature will work the best if the reference image uses the same model as the consistent character images. You could try using a style reference with skin tones more similar to what you want, but Leonardo does tend to have a smooth/airbrushed look.
But this method is not working in my cell is it not for free Ai Leonardo users
They have moved this feature for paid users only now.
Character Reference Not Working in Anime Images
Leonardo Ai isn't the best way for consistent characters right now, something like finetuning a flux model can work better. I have a guide for that here: th-cam.com/video/5Z8fwEeWfRg/w-d-xo.html
@@taoprompts Thank you, but I bought a subscription, and now it's a waste. Can I create a good manga using Novel AI or Open Ai with consistent characters?
@@taoprompts and Which Best Anime Model for Open Ai
I think they will never let us use our own pictures and will always distort the effect to avoid trouble.
One key reason why it doesn't work well on our own images is that it's a technical challenge.
Basically you would need to make minor updates to the Ai model itself in order for the "character reference" feature to work well with your own images.
This can be done in Leonardo with the "train model" feature, but right now it's still not super consistent. I may do a video on that in the future.
@@taopromptsoh ty. I was wondering why my creations didn't look anything like me lol
Download the image generated using your own photo as a character reference. Then go to mid journey and try using the face swap plug-in on that image I’ve had pretty good results.
@@Author_SoftwareDesigner I tried this method in Stylar, but I think it looked bad because the picture I used was a selfie :D not that good resolution/quality. So it came out a bit grainy. I guess we should have good pictures with different expressions for that?
I have used a photo of myself perfectly successfully.
next video on generating multiple consistent character in Leonardo ai with characters mapping
That's tricky to do in Leonardo AI right now, I will try to see if that can work though,
thx sir
💯
I can use for anime art characters
Yeah, I made a video for how to make comic style illustrations with consistent characters here: th-cam.com/video/rIsus4VLVTw/w-d-xo.html
after trying on multiple characters, leanardo failed to generated character with consistency, it just following one girl which is similar to girl shown on video even i tried with mutiple faces but no use getting similar girl image not what i want,
The more generic the character looks (average), the easier it is for the AI to replicate it.
The issue I have with Leonardo is its rigidity. You can't really create complex images. Just one character doing one thing... It's not really good for expressive multi character art
If you use a reference image with two characters with different poses as input to "pose reference", it can do an okay job of using both poses. But it does get mixed up between the two characters quite a bit.
I don't think that works with multiple consistent characters though.
@@taoprompts actually i mean flexibility of its creation in general... "E.g create a cinematic close up side shot of a dark skinned native African man with scars on his body, wearing glowing gold bracelets, in an archer's stance/pose. the bow is made of gold light, and there is no arrow. His left arm has a unique, mystical, mutated feature that stands out. The background is a motion view. The scene should have muted colors, a fairy tale style, comic book illustration, with a vertical aspect ratio." will never come out as you want in Leo
@@Mr_noah0What is the alternative, then?
@@MedievalFantasyTV I don't have one. My content uses ai images and i would really love my characters to have same face while still doing complex actions. It's why I jumped on this video. But Leonardo only gives same face, but can't really create some complex actions.... so we wait for more upgrades from other ai generators.
Have you tried using a guy with glasses as your character reference, then the final product result, then the guy doesn't have glasses anymore?
Yes, it can sometimes remove glasses from a subject. Inside the prompt you can specifically ask for glasses to be generated in the image
i try with my personal photo he not give me good results
This feature works best with images generated from Leonardo Ai, it's not really meant to work with your own personal photos.
My results are literally nowhere close to that good. My character never looks real, like some weird exaggerated version of them. Tried different times with different base models, ugh
Try generating a character reference with a very simple prompt. Character Reference works better on basic generic looking characters
how u got so much quality generation, coz when Im trying my generation come out too much sharp and bright =( it is immediately obvious that it's AI
You could try using specific prompt keywords like "muted colors", "polaroid camera", "instant camera", "taken with an iphone in 2008".
midjourney or leonardo?
Midjourney gives high quality images overall, leonardo gives the user more control on the elements of the image
No one. Even your self do not talk about camera angles for interiors on Mid Journey. Its very hard to what you want. Why not do a video on that Tao. Honestly I only see video about camera shots and people.
I don't think that's really possible in Midjourney or any other Ai art generators right now. You would need real 3D understanding to accomplish that.
You're best bet would be to use a 2D to 3D tool, take snapshots from different positions, and then use those as backgrounds.
sorry😂🤣🤣🤣 its not free anymore need pro subscription
Oh dang really? It was free to try a few days ago.
That’s nice, but I don’t want a photorealistic consistent character. I want to create stylized consistent characters.
I think this is mostly designed for photorealistic type of images, but I will test to see if it also works for stylized characters.