P: Maintaining a scene or character across different prompts

[James2808541164de]

Let's say I wanted to storyboard a script and produce different angles of the same thing. It would be nice to have something that holds on to the core look and objects within the frame so we can get a wide, medium , and close shot of the same scene.

[Kida37789344gyz5]

Buenas, consulto si alguien sabe como mantener un personaje al crear diferentes imagenes y por otro lado si es posible elegir de forma externa un objeto para aplicar en una imagen. Gracias

[Monika Gause]

You can try a stylistic reference, but it will still need a lot of luck to get even 2 similar results.

As for the object: do you refer to a structure reference?

[Kida37789344gyz5]

thanks for your answer, I use the structure reference, but sometimes I need to add some objects and I can't find the right one... for example it was impossible to find a mini journalist recorder

[Robert McDaniels]

Every render from Firefly is unique. This makes creating a series of images with consistent characters or items impossible. I would like the ability to select an object or character in a generated image and "lock" it. Then I could create a new generation/render but keep the character. This way a could do a series of renders but keep the same character. For example I could render a person playing baseball, going for a walk, and cooking a meal AND keep the same person through out each render.

[Brasal381599677pcz]

After I create a persona, isn't it possible to make her do other poses while maintaining exactly the same physical appearance?

[Dean_Utian]

At this stage, every new Firefly image creation is relatively unique. You can use a reference image and have similar prompts but you can't guarantee what you're asking for - having the same character across multiple images but in different poses. Maybe something we'll see in the future.

[Pablo Daniel36954445d55t]

Seria muy bueno que una vez generado personajes, en dibujos proximos se pueda hacer que dibuje los mismos personajes. Es muy frustrante cuando se dan las mismas indicaciones e incluso se sube una imagen de referencia que no respete. Por ejemplo para un story board seria muy util. slds.

[ernestol75156626]

adopt the format with seed to be able to generate the same person in different situations, poses or clothing

[natasha246883436eqv]

I purpose that Adobe make it easier to duplicate a character in different poses to be used again and again. 

[Elizabeth30378353ga1g]

This a common theme amongst AI image generators. They just can't seem to visually keep the same character and put them in different situations and image reference doesn't work.
Human beings are so tuned to recognising minute differences in faces that even the slightest deviation, regardless of light, angle and a myriad of other factors, results in an entirely different individual visually.
It's going to take some serious work by AI engineers to develop a cure for this problem.
until then we have to rely on illustrators and artists.

[CSI Productions]

Elizabeth,

It sounds like you have some programming experience. If what you say is true, wouldn't it have been easier for UKW to just say that from the beginning. Don't some of the best ideas come from a collaboration of other ideas presented by multiple minds? Wouldn't the ideas of a half million creative thinkers help programmers solve the issue?

It's better (IMO) than stringing us along thinking they're working on a solution and it should be available soon. For instance. the way "I" think is:

AI image generation has amazing 3D animation abilities as well as 2D. To build a 3D charater you need a front, side and top view. So use this same theory in 2D by establishing your charcter as you would in 3D, by submitting images of the front, top and side view of your character. Once input into the AI software, shouldn't it be able to analyze those frames, establish a three dimensional "model," position the expression, head angle, facial expression the way it would in 3D but output it in the 2D parameters you've set for your image?

It may be a work around but this is just "my" take on it. I'm sure programmers could get ideas from others who really want and need this feature but also want to offer their input to solve this challenge.

Thoughts?

[CSI Productions]

No offense guy but this sounds like an AI step by step on how to draw traditional storyboards not something applicable to keeping consistent characters across a series of storyboard frames. Doesn't really apply to helping others have the same character with different facial expressions, clothing, hair styles or a variety of scenarios using the same character. Just an observation.

[TitoFG]

This would be an amazing feature.
Fully support this idea! 🙌🏼

[Mark Beagle344645835rcz]

I want to show a dad with his two daughters in the kitchen getting ready for school, then show the same dad and same kids at a school entrance when he drops them off.  How do I generate the same people in multiple scenarios? 

[Monika Gause]

You cannot generate the same people.

[Justin 907]

It would be very helpful if I could upload the image of a drawn character and an image of a person or other drawn character in a particular position and could then have Firefly create an image of the first drawn character in the position of the person or character in the second image. (These images are attached. I wanted the state trooper redrawn in the position of the falling guy.) I wouldn't want the result to be kind of like the first character. I would want the result to be that first character in a different position. Here is a prompt I wrote to try to achieve this, but with totally unusable results, "Create the character from the style reference in the position of the character from the composition reference." This would be vey useful, because I can draw a character once and be happy with it. I have a very hard time drawing that same character in different positions and keeping that character consistent. On the bright side, Firefly was way more useful than the last time I used it, which was many months ago. Thank you.

[MRDisperatti]

Save a Favorite Personal Generated object, pet, landscape, etc... to be used in future creations

[Lirui23581211eobf]

I used Firefly to generated a headshot of a person and when trying to use the headshot as a reference to generate a full body image, it seems not very optimized especially in different angles of the face, any idea to optimize this?

[Monika Gause]

That is not possible

[Kevin Stohlmeyer]

@Monika Gause It would be helpful to refer users to this idea for upvoting instead. Very popular request.

[Martin388051654u0d]

This could be a big help to those who are creating storyboard frames. Maybe there could be a prompt wherein we can use thesame look for the person (face, outfit) but we can alter the action/poses and facial expressions. 

[estiim]

i generated a female model in Photoshop beta AI and now i want to create exactly the same model on a deferent format. is it possible?

[CSI Productions]

Tracy, at the risk of sounding rude, your thread entry is from last year.

I don't understand how a statement that you want to track the popularity of this request is even in question. Ever since Firefly was introduced to the world users have been requesting this feature. From children's book illustrations (my personal request) to storyboards to print ads or corporate how to's - this community has been asking for this since it's inception.

This whole time, through many upvotes to anyone who's requested this feature I think it's painfully clear we've been under the impression Adobe has heard us and was working on it.

Your response suggests you need to take a poll to determine if it's something worth Adobe's time suggesting developing this feature hasn't even been started yet. Has it? Is it even possible? I'm not a programmer so I won't even pretend to know what's involved and I respect those who can write code.

I understand with certain feature requests popularity dictates priority but it's been clear to anyone who's followed this thread that we've been asking/waiting for this feature for over a year.

I would submit the jury is in and the verdict is, there's a dire need for this feature. We'll keep waiting, unless another company figures it out first, in which case, problem solved.

Thank you.

[CSI Productions]

Welcome to the world of frustrated AI users. This has been one of the most popular requests since Ai Generated imagery has been released on the creative community. Maybe AIimage generation can't do everything. That being said, it sure is cool for the stuff it can do.

[CSI Productions]

That entry was misleading. You want to use the same character in for example a storyboard, where you can have an established character in various scenes and with various facial expressions, camera angles, clothing etc. What was written as a reference image "suggested" you could establish a character and then use it for a reference image and from there you can generate any other scenario with the character matching the reference - not so.

Reference image won't do that. In the example they are just asking for a cute dog sitting (no different than doing a stock photo search - in this case) and then selecting one of the photos and then changing the look to a different art style.

It won't let you pick one of the photos, use it as a reference and have the same dog chasing a stick, holding a ball in it's mouth or sleeping on a childs bed. This is what we all want but so far it's incapable.

The closest example I've seen is in this thread. It has images of a Latin looking woman with dark hair in various scenarios and that shows promie. Is it Adobe or proper prompt? I submit it's probably a combination of both.

We're all waiting for this feature to get nailed down. If Adobe pulls it off their stock will probably go up because every creative on the planet will be using it.