Master the Art of Creating Consistent And Diverse Faces In Playground
Вставка
- Опубліковано 24 лип 2024
- Learn how to create consistent faces on Playground AI! In this video, we'll show you the step-by-step process to develop consistent and diverse faces. Certain models tend to have default looks but we can use that to our advantage prompting a certain way to create a consistent look. It's not a perfect method but it works very well and it's quite easy to do!
Time Stamps
0:00 Samples of random faces
0:49 Models have default "looks"
2:14 Use names to develop your character
3:22 Combining names
4:23 Using celebrity last names
5:00 Change it up with filters
6:00 Define ethnicity
7:20 Some filters are limited in variation
Find us on
Playground AI - ai image generator: playground.com
X (Formerly Twitter) https: // playground_ai
can you teach us how to create consistent character using a base character? obtaining all details in the photo and create many samples
This video was quite helpful. One of the things I eventually hope to do is create "storyboards" or narratives for the same character. I've had some success keeping the character's appearance consistent but I think this will work better. I've only been at this for two weeks and I have made a lot of progress.
I finally upgraded to Pro and I'm having a blast experimenting with Playground at full speed. I've been using the same model since SD 1.5 that is a combination of two actresses which gives me a very distinct and consistent appearance. It also works well for drawings, paintings, and other media as well.
I have nothing but kind words for the Playground team. Keep up the fantastic work guys.
Glad it's working out for you! I've also been using this technique since SD1.5, it's a great way to create new faces. Appreciate the kind words and support!
Same here, just joined pro last Sunday and I’m blown away 😮❤
Quite a few things coming soon! Sit tight!
Thanks a lot for this video, I had already tried to mix 2 celebrities names, it's very amazing 😄 Congratulation for the PGAI Team and your great work. 👍
Glad it was helpful!
YES!... havent even seen it and Im exited already!
Love to hear that!
*Спасибо вам за проделанную работу и титанический труд)*
You're welcome, we appreciate the support! 🙌
I'm having a hard time creating the clothes or accessories of some historical characters. Can I create a photograph by integrating at least the dress and accessories into the image I will create, using "image to image" from real photographs?
Hello,
Thanks for the information informative yet again and helpful.
Just one question; awhile back created an image of an muscular woman. Firstly, the image shown was of a European woman, but then added the name Kim Yuen (learned that from one of your prior videos, mixing two names together) for a more far eastern looking woman. Got several good images. Then changed the surname to Khan for a more middle east/Asian look and received a message stopping the generation due to bad prompt. But the same prompt had been used, only the word Yuen had been change for Khan.
Is there an actual list at Playground AI of words that cannot be used?
Can you create a video on how to write a prompt correctly so that the eyes of the shown character are directed exactly at the viewer or in another expected direction?
Thanks once again for a clear and concise video. I'm on a pro plan. I'm more than happy with the process and have been getting decent results with this technique. The biggest bug bear that I have is hands. So often they are misformed, too many fingers, look like they're broken etc and even trying to fix them on canvas is very difficult. Do you have plans to improve this aspect?
You're welcome! Hand have improved greatly compared to last year. Although it still can use some improvements. Comes down to how the model was trained initially. Of course we are always looking at ways to improve them for our users. That being said, starting with a good model helps. I'll be posting a video on this very soon!
I love it!
good ai I love it don't stop free version or don't Lowering the standard of free version like midjourney keep it up I love this ai
Thank you! We'll try our best to keep the free version packed with value.
this is very helpful. But to make this process easier, would suggest if there is a "characters" feature instead so you dont have to type prompt each time.
Thanks for the feedback. 👍
Thank you for this very interesting video. I would like to ask one thing: how is it possible to create a very detailed drawing with a character and a particular outfit, not necessarily a photo, and then create multiple images of the same character with the same outfit in several different surroundings?: for example a dog with its blue dress walking, then the same dog with the same dress traveling on a motorcycle, Then the same dog climbing a mountain and then baking a pizza
Are the images I generate and edit private to me or can the others see what I generated or edited if I am on the free plan?
What if you have already created a character you want to keep?
"I uploaded a photo and I don't want to alter just the face; I only want to add prompts like a mustache or glasses. A year ago, I used this AI playground for this purpose, but why can't I do it now?"
Awesome!
Unfortunately I have no idea what Refinement is doing; I always put it at max becasue I have the beliefe the pictures will be better looking.
Maybe you could explain me what this does on my pictures or are there any videos about this topic that I may have missed?
Great work by you guys, I love PlaygroundAI!
We appreciate the support! I go over the refiner here ua-cam.com/video/jml4bdtbOhw/v-deo.htmlsi=Mq5n072W6tRPVOyP. Basically use it sparingly, max will do the opposite and make it look horrible! 😬
@@playgroundai Thanks! Love the support and enthusiasm you give us! :)
Can you make it the consistent faces with consistent cloth?
Could you please add an option for extension the image in Canvas in all directions at once? Now need 4 times at least or many more if the image is not square. It could be like a mask for at least 3000 pixels in each direction which to be applied above the selected image. It will be better if the mask is with more pixels.
Also PLEASE open all the filters for the extensions. Now this is not possible ( only the old are available) , and the extension often get very different than the original piece.
Also a library with the prompts will be nice to be added.
Or option to create our own styles.
Thank you very much. Playground is amazing :)
Unfortunately outpainting doesn't work that way at the moment however as technology progresses, I can see something like this down the road.
Originally, when I created the image, the text section of the prompt that should be shown disappeared. And when I click on the remix, the prompt doesn't create the same. I'm thinking of paying for a pro when I have some financial leeway, but now I'm a free user, and I've been using this program since it's been a while since I logged on and created images, so many things have changed and, of course, the images have turned into high quality, but I hope free users can give me some permission to use the prompts. I want to make an image, but I can't do that now, so I'm leaving a comment because I want to know how to use it. Please comment!
Hello,
Thank you, always enjoying your videos.
Just this still looks like a "trick". No guarantee of real consistency.
I quite understand your trick.
Question : wouldn't it be interesting to give an "id" to our character, for the AI to remember it and assign it to 1 single person ?
For example : "Cinematic portrait photo of Alicia Thyme Sylvia witherspoon (remembered as or call : id: 18f947d531x ) , film, bokeh, professionnal, 8k, Highly detailed" ?
Thanks
Not really a trick but definitely a work around. Moving forward we're looking at ways for this to be easier. Consistency in general is just difficult for current AI image generation but there are solutions that can get you close which we are looking into.
Great video. Thank you.
If I copy a seed from an image and then make changes, I sometimes end up with a completely different image as a result. Why is this happening?
You're welcome! Generally any change you make to the prompt will alter the image to some degree, seeds are just a random number and have no control over consistency. It can keep certain things similar but the more changes you make the more it will sway away from the original. Consistency across is still a big challenge for stable diffusion and we're looking at ways to address this.
@@playgroundai Ok, thanks for the answer and information. 👍
What has Midjourney done that you haven’t? Images from Midjourney are lifelike but playground are AI (illustrations). Can you fix this???
Hello, thanks very much. I'm addicted to Playground AI and am considering upgrading to the paid version. I have a question. According to the information on the web page, in the free plan, I can generate 500 images per day. However, after generating 60-70 images, I receive an error message that says “Unknown error. Please try again later or contact support if it continues.” When I try again, I receive a message that says “Your queue position is based on the number of images created and demand is high right now. Please try again later or Upgrade.” Except today and yesterday, I haven’t generated any images in the last three weeks because I’ve been travelling. The reason I haven’t yet enrolled in the paid version is also because of my travels. My question is, if I apply for the paid version, will I still be subject to the limitation of long queue positions? 🤔
Hi there, that's mostly due to the server load, during high traffic times free users get put on a queue system and paid members get priority. Those times can vary depending on overall usage. We are looking at ways of making it a but better for the free tier but we do have to prioritize paying members to ensure they get the service that's promised. Inference can be very costly and in order for us to offer a good "freemium" there will be some trade offs.
@@playgroundaiThanks a lot for your reply. Very much appreciated.
gracias, me siento como Frankestein 😁🤔
You're welcome!
Hello. I do not wish to use my CC and I am from India. How can I get a premium subscription?
please contact support@playground.com. 👍
Folks folks folks. I love Playground but you really need to work on an upscaler. Look what Leo****o does with theirs. This is really missing in Playground.
We agree and we are, sit tight, It will be worth the wait. 👍
@@playgroundaiI’m still sitting tight here on my iPad waiting to be able to use canvas…… 😢 🇦🇺❤️
sorry I have to say this but I think the old version was better but now for some reason it produces pictures very slowly :C
If you are on the free account there is a queue system now giving priority to paid members. Typically during peak times it can generate a bit slower for free tier.
中国来的。能加您vx么。我想请教您。
👋
👋 Hope you are well my friend!
Nothing about diverse faces. And whats the point of talking about consinstent faces? There is only one face in Playground AI.