Thank you for your video but I have a request at the end below.... Maybe it's just the examples in this video and in other consistent character generations but I have yet to be impressed. Why? Because the characters are always looking directly straight into the camera. It's very easy to pick out AI generated images of people because they all have that direct eye contact with the camera. What about making your consistent characters in action poses? Like riding a bike from the side or behind the character? What about making two consistent characters looking at each other like they're having a conversation? Are these possible yet? I would really like to see if you can show that in one of your future videos if possible. Thank you.
To avoid characters looking at the camera, you just need to prompt specifically. You want to give Leonardo/Midjourney or image tools keywords that let it know you're going for something cinematic (actors don't look at a camera) instead of portraits. So with that in mind, "side profile", "looking to her left", "high angle shot" - these are examples of things that'll tell Leonardo, or Midjourney and other image generators not to have the subject look at the camera, because it gives specific direction. As for two consistent characters looking at each other in the same shot, it is VERY hard. I just did this for a TV show I'm doing and had to do some photoshopping. What I did using Midjourney was prompt for a "two shot of a man and woman sitting together on a couch. On the left is a woman with blonde hair blah blah. On the right is a man with blah blah". Then using the CREF (character reference) parameter, I pasted links to an image of each. It comes out a little wonky, so I used "Vary Region" and photoshopped a bit from there. This is pretty advanced, as they don't naturally do this two shot style very well.
Really nice video with great comparison, I also tried leonardo ai ,but I recently use it because it is free and have more style than leonardo, hope you can make a tutorial about Stylar
Leonardo absolutely does not have the most consistent characters. The only reason they even have it is because they are using ipadapter and/or pulid both are in comfyui for everyone and for free.
Great comparison video! 🙂
Thank you for your video but I have a request at the end below....
Maybe it's just the examples in this video and in other consistent character generations but I have yet to be impressed.
Why?
Because the characters are always looking directly straight into the camera.
It's very easy to pick out AI generated images of people because they all have that direct eye contact with the camera.
What about making your consistent characters in action poses? Like riding a bike from the side or behind the character?
What about making two consistent characters looking at each other like they're having a conversation?
Are these possible yet?
I would really like to see if you can show that in one of your future videos if possible.
Thank you.
To avoid characters looking at the camera, you just need to prompt specifically. You want to give Leonardo/Midjourney or image tools keywords that let it know you're going for something cinematic (actors don't look at a camera) instead of portraits.
So with that in mind, "side profile", "looking to her left", "high angle shot" - these are examples of things that'll tell Leonardo, or Midjourney and other image generators not to have the subject look at the camera, because it gives specific direction.
As for two consistent characters looking at each other in the same shot, it is VERY hard. I just did this for a TV show I'm doing and had to do some photoshopping. What I did using Midjourney was prompt for a "two shot of a man and woman sitting together on a couch. On the left is a woman with blonde hair blah blah. On the right is a man with blah blah". Then using the CREF (character reference) parameter, I pasted links to an image of each.
It comes out a little wonky, so I used "Vary Region" and photoshopped a bit from there. This is pretty advanced, as they don't naturally do this two shot style very well.
Have you tried it in Leonardo ai? For two consistent character conversation?
QQ, how to make multiple consistent characters in one image, in Leonardo AI? Many Thanks! @The Reel Robot
When is this character consistency tool going to be released on Leonardo AI?
Should be by mid next-week based on timeline I was told
muitoooo bom, obrigado pelo vídeo
Question: can we combine character reference with image to image in leonardo ai?
Yes. Just tried it out to be sure and it combines quite well
Hi! I don't see the Character Reference option in my Leonardo subscription! Which subscription are you using? Thanks for the informative video!
It’s early access. I’m in a creative partner program. Should be there by mid next week for sure
@@TheReelRobotDamn, I honestly can’t wait 😂
Was the reference photo generated in Leonardo and then used as the character reference or was it an uploaded photo?
It was actually a Midjourney photo I uploaded into Leonardo
With this feature, can you generate multiple characters with the same face in a single image? (Clones, basically 😂)
I haven’t tried it in Leonardo, but I’m pretty sure you can because I’ve done it in Midjourney with their character reference
@@TheReelRobotThanks for the feedback 👍
Really nice video with great comparison, I also tried leonardo ai ,but I recently use it because it is free and have more style than leonardo, hope you can make a tutorial about Stylar
Bro this is only paid
Leonardo absolutely does not have the most consistent characters. The only reason they even have it is because they are using ipadapter and/or pulid both are in comfyui for everyone and for free.