I just stumbled upon your channel and I gotta say, you're killing it with your teaching style! You have a knack for explaining things in a way that's easy to understand, without any unnecessary fluff. It's crazy to me that you don't have more subscribers. You definitely deserve way more recognition for the great content you're putting out there. Maybe try tweaking a few things to work better with the UA-cam algorithm, because more people need to see your videos. Looking forward to more : )
Well, you got me over here blushing 😊. Thank you so much. I'm a student of UA-cam and the tricks you're supposed to use to attract views and subs. There are some things I just can't bring myself to do, like putting "secret revealed" or "mind blowing" in every title, or starting with "make a million bucks" as the lead to tell you about a software tool. But there are other things that I'm working on to improve. If you have any suggestions, I'm all ears.
I agree, too. Just found this channel and subscribed. I have a subscription with Midjourney and have always kept up with Leonardo, but this tutorial has convinced me to try the paid subscription, too. Thank you for your video!
You're right, I should have spent some time on prompt writing. I'm sorry to disappoint, but I sincerely appreciate the feedback. Maybe I'll do one that focuses on prompts, or a redo of this one.
Thanks so much for sharing your knowledge and experience. Well done! The tool is absolutely amazing! Would you have any tips on how to get the same character dressed in the same clothes, in different poses. I am working on a project where I have a great story about a woman who lost her husband a son and four grandchildren in an earthquake, but the interviewer only got a very low res image of the woman with a cellphone. Leonardo did an amazing job of taking that one low res shot and is producing wonderful images for me, but I would like to have four or five different poses of her in the interview setting, looking at the interviewer, looking down with grief, etc, but with the same clothes. WHen I try and add content or image to image, it basically just duplicates the Character reference image. Is there something I should be doing with the prompt. Like "keep the same clothes as in the reference photo"
I'm not sure that giving it direction like "keep the same clothes as in the reference photo" will help. Personally, I've never had any luck with that. I suspect the trick is going to be using a combination of the image to image reference options, with adjustments to the amount of influence applied to each. For instance, a combo of character reference, content reference, style reference, etc. At the moment you can't do multiple image references in the new V2 image generation (they're adding that back soon) so you might switch it back to legacy mode (where you CAN use multiple reference options) and play with those options. It *might* even be a matter of going through one reference option to get one aspect of the image generated (same clothes or different pose) and then going through another reference to get another aspect dialed in. In other words, get there one step at a time rather than asking the AI to do 2-3 things at once. The images you're trying to create, and the overall story, are very intriguing. I love hearing about what others are creating and how you're using the tools to get there.
@@excelerator thanks for taking the time. I hope this is something that they work on next. I would like to illustrate stories with Leonardo for my English language channel. Having consistent characters is a great first step, but I also need the characters to keep the same outfit on from scene to scene or the story doesn't make sense or gets confusing. Love the interface of Leonardo over MJ, and it keeps getting better!
It's quite difficult to keep the clothing consistent when I tried it. You could try using the style reference and content reference as well to keep it as close to the original as possible. Also try and use the same prompts for the character description and set the two settings I mentioned to low. That sometimes work for me as that means Leonardo will weigh more heavily on your prompts if the other two settings are set to low.
Thanks @excelerator for fantastic videos, clear and to the point. I agree, having a consistent character is one thing, but the whole point is to able to have that consistent character doing different things in various poses. Without that, it's not much use. Thanks again, great info well presented.
midjourney now lets you have multiple crefs, and the output though different from the crefs are consistant amoungst themselves. i have used neither leonardo or midjourney so cant say which is better although i know midjourney has image extention which allows you to create entire scenes by panning out or to the side expanding the image so that is handy if you dont want just a portraitm and the multi crefs might also help u get consistant costumes and clothing in every scene but leonardo may have these too
It isn't available yet in the new interface. In the upper right-hand corner, click the "Legacy Mode" toggle. Then, just below the prompt box and the model, click the "image Guidance" tab. Upload your source image and select "Character reference form the dropdown." Hope that helps
For photorealistic images I generate, I find switching to Lightning XL for the characters works really well. I am curious if anybody has the same experience
Oh No! 1k tokens and not getting anywhere is not good. I'm assuming you've tried different models, different generation mode settings (fast vs quality) and different prompts (between simple and really detailed) to see if you can isolate the issue. I don't mean to sound obvious, but the reference image has a clear well-lit face from the front and not a profile or anything, I assume. If you've experimented 1k tokens worth, you might want to ask Leonardo (there's an icon in the lower right hand corner of the generation screen where you can get help/chat with Leonardo. Or, maybe jump in the Leonardo Discord - the folks in there seem to be really friendly and helpful.
My experience thus far is that Leonardo almost ignored the uploaded reference image and then, to add insult to injury, kept using its wrong version no matter what info I put in the prompt even after selecting MAX Strength (adherence?), no matter what negative prompt I added. It's truly frustrating. I WISH it worked the way it seemed to work for you @Excelerator. (Ex. black jeans yields gray slacks; ankle boots yields knee boots, even barefooted!; no open shirt yields a bare chested guy; athletic build yields a guy skinny enough to slip through the slats in a snow fence.) 😠 P.S. Free version. Does a subscription yield better results and control?
I just tried it with another headshot -- a professionally lit photo of a woman. I might as well have started with text! Went from being the uploaded tan Italian woman to being a pale English brunette. WTH? Complexion and facial structure, as well as hair, are completely different. Additionally, I started with a headshot and Leonardo decided a full-body image was indicated. I only described the subject from the shoulders up! Holy crap, it's frustrating.
That sounds bizarre. Not to be insulting, but are you sure you selected "character reference" under image guidance and not one of the other options like style reference? Subscription should only give you more features, not make the features work any better. I have to laugh at the open shirts and bare chested guys ... Every platform and model I use seems to be obsessed with putting women in a shirt that only has the two bottom buttons buttoned, and frequently produces shirtless men. 🤷🏽♂️
A tan Italian woman to a pale one is a real disappointment (as a guy who's married to a tan Italian woman!) Using character reference, you can definitely get from a portrait to full-body, and it's great that it can do it... When that's what you want. Have you tried switching models? Phoenix is the default model and it's pretty good in my experience, but maybe seeing what portrait perfect or cinematic comes up with... Maybe worth a shot. Also, I would double check all the settings to make sure something didn't get turned off/on or slid too far. What you're describing does not sound normal. Sure, sometimes it won't give you exactly what you wanted right away (or ever) but it sounds like the results you're getting are so far off.
You can use this in the lowest-tier paid plan, Apprentice, which is $12 (USD)/no or $120/yr. I'm not sure if it's 12 euro since I'm not in that region it only shows me USD. Look for the upgrade option and confirm the price in euro for "Apprentice" excelerator.tech/getleonardo
Get Leonardo AI (FREE) 👉 excelerator.tech/GetLeonardo
I just stumbled upon your channel and I gotta say, you're killing it with your teaching style! You have a knack for explaining things in a way that's easy to understand, without any unnecessary fluff. It's crazy to me that you don't have more subscribers. You definitely deserve way more recognition for the great content you're putting out there. Maybe try tweaking a few things to work better with the UA-cam algorithm, because more people need to see your videos. Looking forward to more : )
I agree
Well, you got me over here blushing 😊. Thank you so much. I'm a student of UA-cam and the tricks you're supposed to use to attract views and subs. There are some things I just can't bring myself to do, like putting "secret revealed" or "mind blowing" in every title, or starting with "make a million bucks" as the lead to tell you about a software tool. But there are other things that I'm working on to improve. If you have any suggestions, I'm all ears.
Thank you. 😊
I agree, too. Just found this channel and subscribed. I have a subscription with Midjourney and have always kept up with Leonardo, but this tutorial has convinced me to try the paid subscription, too. Thank you for your video!
@@EmilyNilsen OK, I'm blushing again. 😊😊Thanks for the kind words. I hope you enjoy Leonardo!!
thank you so much for this update. It really helps. I was getting very frustrated trying to find everything in the new version.
These platforms keep changing! I think this will be a good change once they get everything in the new format.
Your instructions are very in-depth and easy to understand. Leonardo is a great tool, I like it better than Midjourney
Thank you so much! Glad it was helpful. 😀🙂😄
Yaayy! Character consistency can be frustrating. So glad Leonardo ai launched this feature!
For sure! This is so much easier than assembling a dataset and training a fine-tuned model for a character.
Leonardo AI's ability to create consistent character references is amazing. As someone who's always looking to improve creative projects.
I agree! 😊
Great video
Glad you enjoyed it!
great .expecting more on beginner level prompt writing for Leonardo AI and basics of Leonardo AI .your style clear explanation is greatly appreciated.
You're right, I should have spent some time on prompt writing. I'm sorry to disappoint, but I sincerely appreciate the feedback. Maybe I'll do one that focuses on prompts, or a redo of this one.
Wow, better than MidJourney!
Wow!
Subscribed.
Awe, thank you so much!!!
Thanks so much for sharing your knowledge and experience. Well done! The tool is absolutely amazing! Would you have any tips on how to get the same character dressed in the same clothes, in different poses. I am working on a project where I have a great story about a woman who lost her husband a son and four grandchildren in an earthquake, but the interviewer only got a very low res image of the woman with a cellphone. Leonardo did an amazing job of taking that one low res shot and is producing wonderful images for me, but I would like to have four or five different poses of her in the interview setting, looking at the interviewer, looking down with grief, etc, but with the same clothes. WHen I try and add content or image to image, it basically just duplicates the Character reference image. Is there something I should be doing with the prompt. Like "keep the same clothes as in the reference photo"
I'm not sure that giving it direction like "keep the same clothes as in the reference photo" will help. Personally, I've never had any luck with that. I suspect the trick is going to be using a combination of the image to image reference options, with adjustments to the amount of influence applied to each. For instance, a combo of character reference, content reference, style reference, etc. At the moment you can't do multiple image references in the new V2 image generation (they're adding that back soon) so you might switch it back to legacy mode (where you CAN use multiple reference options) and play with those options.
It *might* even be a matter of going through one reference option to get one aspect of the image generated (same clothes or different pose) and then going through another reference to get another aspect dialed in. In other words, get there one step at a time rather than asking the AI to do 2-3 things at once.
The images you're trying to create, and the overall story, are very intriguing. I love hearing about what others are creating and how you're using the tools to get there.
@@excelerator thanks for taking the time. I hope this is something that they work on next. I would like to illustrate stories with Leonardo for my English language channel. Having consistent characters is a great first step, but I also need the characters to keep the same outfit on from scene to scene or the story doesn't make sense or gets confusing. Love the interface of Leonardo over MJ, and it keeps getting better!
It's quite difficult to keep the clothing consistent when I tried it. You could try using the style reference and content reference as well to keep it as close to the original as possible. Also try and use the same prompts for the character description and set the two settings I mentioned to low. That sometimes work for me as that means Leonardo will weigh more heavily on your prompts if the other two settings are set to low.
@@SciFiStoriesUniverse Thanks for adding your experience & expertise!!!
Thanks @excelerator for fantastic videos, clear and to the point. I agree, having a consistent character is one thing, but the whole point is to able to have that consistent character doing different things in various poses. Without that, it's not much use. Thanks again, great info well presented.
midjourney now lets you have multiple crefs, and the output though different from the crefs are consistant amoungst themselves. i have used neither leonardo or midjourney so cant say which is better although i know midjourney has image extention which allows you to create entire scenes by panning out or to the side expanding the image so that is handy if you dont want just a portraitm and the multi crefs might also help u get consistant costumes and clothing in every scene but leonardo may have these too
Thanks for adding this helpful info to the conversation. I appreciate it, and I'm sure others will too.
Character Reference grayed out as unavailable "coming soon" in my PAID suscription...
It isn't available yet in the new interface. In the upper right-hand corner, click the "Legacy Mode" toggle. Then, just below the prompt box and the model, click the "image Guidance" tab. Upload your source image and select "Character reference form the dropdown." Hope that helps
For photorealistic images I generate, I find switching to Lightning XL for the characters works really well. I am curious if anybody has the same experience
Thanks for the tip.
Excellent as usual .
Appreciate you watching and your kind words.
I am having at present using( I D M) and all is making me crazy. Nothing is working to download my choices.
Have you tried downloading without using IDM? Or a different browser? I've not heard of, or experienced, any issues with downloading from Leonardo.
I'm not getting even remotely similar results to you, it's like the character setting isn't turned on at all. 1k tokens deep I'm a little confused.
Oh No! 1k tokens and not getting anywhere is not good. I'm assuming you've tried different models, different generation mode settings (fast vs quality) and different prompts (between simple and really detailed) to see if you can isolate the issue. I don't mean to sound obvious, but the reference image has a clear well-lit face from the front and not a profile or anything, I assume. If you've experimented 1k tokens worth, you might want to ask Leonardo (there's an icon in the lower right hand corner of the generation screen where you can get help/chat with Leonardo. Or, maybe jump in the Leonardo Discord - the folks in there seem to be really friendly and helpful.
I find youtube vids have great results, but in reality it's very hit and miss
@@Tyrell-d1y bait and switch
My experience thus far is that Leonardo almost ignored the uploaded reference image and then, to add insult to injury, kept using its wrong version no matter what info I put in the prompt even after selecting MAX Strength (adherence?), no matter what negative prompt I added. It's truly frustrating. I WISH it worked the way it seemed to work for you @Excelerator. (Ex. black jeans yields gray slacks; ankle boots yields knee boots, even barefooted!; no open shirt yields a bare chested guy; athletic build yields a guy skinny enough to slip through the slats in a snow fence.) 😠 P.S. Free version. Does a subscription yield better results and control?
I just tried it with another headshot -- a professionally lit photo of a woman. I might as well have started with text! Went from being the uploaded tan Italian woman to being a pale English brunette. WTH? Complexion and facial structure, as well as hair, are completely different. Additionally, I started with a headshot and Leonardo decided a full-body image was indicated. I only described the subject from the shoulders up! Holy crap, it's frustrating.
That sounds bizarre. Not to be insulting, but are you sure you selected "character reference" under image guidance and not one of the other options like style reference?
Subscription should only give you more features, not make the features work any better.
I have to laugh at the open shirts and bare chested guys ... Every platform and model I use seems to be obsessed with putting women in a shirt that only has the two bottom buttons buttoned, and frequently produces shirtless men. 🤷🏽♂️
A tan Italian woman to a pale one is a real disappointment (as a guy who's married to a tan Italian woman!)
Using character reference, you can definitely get from a portrait to full-body, and it's great that it can do it... When that's what you want. Have you tried switching models? Phoenix is the default model and it's pretty good in my experience, but maybe seeing what portrait perfect or cinematic comes up with... Maybe worth a shot. Also, I would double check all the settings to make sure something didn't get turned off/on or slid too far. What you're describing does not sound normal. Sure, sometimes it won't give you exactly what you wanted right away (or ever) but it sounds like the results you're getting are so far off.
Same here, does not look like pic in low, mid and high guess I am going back to head swapping.
does this work with 2 subjects in the source file?
I don't believe so, but I can try it out to confirm.
At 8:08 should have gone with "Portait of James Bond"
Bond, James Bond. 😂😁😆
hi can i use this feature for 12 euro a month
You can use this in the lowest-tier paid plan, Apprentice, which is $12 (USD)/no or $120/yr. I'm not sure if it's 12 euro since I'm not in that region it only shows me USD. Look for the upgrade option and confirm the price in euro for "Apprentice" excelerator.tech/getleonardo
you're getting used to all this praise, but in addition to the best tuts on here, you look extremely distinguished in the white tx.
Awe, thank you. 😊 I think anybody would look distinguished in a white tux though. 😉
Bro not open
I'm not sure what you mean?
Hey I'm gonna pay for a month of Leonardo thru your affiliate link! MAKE DAT MONEY haha
Well, thank you. I really appreciate it. 😊
Paid version only
Zehahahahaha you think for free how naive person😂😂😂