P: Upload a reference image(s) for text to image

[sai chander29048727kz6w]

Adding an existing image for reference on the start might open more realistic images.

[Bulshaawi]

Adding an existing image for reference on the start might open more realistic images.

By @sai chander29048727kz6w

[Sean27446359d9b7]

I did this on another AI which has sketch to image and it was awesome. Not perfect of course as it gives you too many options that are more of a relation to your sketch than a refined sketch, but still gettting closer.

[KatherineHarris]

In addition to written prompts, allow the abilty to submit reference photos for the AI to use.

[wilfreds76]

Please add a prompt option for the whole image that allows the user to transform their image into different art style. Like for example I want my portrait picture to be converted into comic art stlyle.

[simonh5504215]

All other AI's have this feature. Firefly will be awesome with it

[CMass]

Thank you for all your votes and feedback! Style match is now a feature in Adobe Firefly! Check out everything you need to know about this feature here!

[NathanOdellin]

Regarding the new "Style" release....I mean, great and all but "generate images based on the look and feel of an existing image" really IS NOT what we had in mind.......Where Adobe could absolutely lead the field on Ai generated images by a vast margin, is to eneable users to upload their image, and have Firefly adapt the image into an environment (for example) or alter specific aspects of the original image, without chaging any other aspect of the original image. STYLING an uploaded image - i.e merely using the existing image as a styling cue, to then "Interpret" a new concept image, is not what many people had in mind when they requested the upload feature (I would bet). What I have described, is a feature completely lacking in all exisitng Ai platforms, I suspect because of copyright implications. However, Adobe differs in the sense that you have a closed ecosystem and the Ai is trained on an image bank you own already. You can easily get users to sign a legal clause, stating that any images they upload to the server are their own responsibilty and any breaches of copyright will be pointed at the user, not Adobe. As long as Ai platforms inhibit (refuse?) users to "montage/collage" their existing images into new scenes, this will be the forever achilles heel of Ai. Designers and Photographers like myself would throw money at any platform, where I can upload my creations, and simply have the Ai take care of "setting" my creation into a concept redrawn - not styled "based on" my creation. Please listen Adobe! 

[Steve323439161ryd]

Really like the new feature, however styling new images based on the uploaded image, while useful in specific cases, is everything this function should do.

Being able to re-draw an image in a different style (specified in the prompt), for example. This would enable users to see what their image looks like with a different aesthetic. For example, if I have created (or taken) a photograph of a fairground carousel, and wanted to see what it looked like at night, or as a woodblock print, or in a cyber-punk dystopia, that would be a very useful feature.

Yes I know Adobe is concerned about copyright infringment, but we have all already agreed to your T&Cs. Why would you bother doing that if you then don't work on the basis that your users are honest enough actors to abide by that agreement?

[CMass]

@NathanOdellin If I understood you, it sounds like you're looking for a way to upload your own image to Style Match. Here is some more information about Style Match. Let me know if this is what you are looking for. 🙂 

@Steve323439161ryd This sounds like an interesting idea. May I have you create a new post here with what you shared in your comment?

[NathanOdellin]

@CMass Hi there. In my post, I was indeed describing the new "Style Match" feature, but my general point was that this does not do what users have been requesting.

The Style Match function is indeed useful for brand new images and concepts, but users did not (on the whole) request this. What we requested, was the ability to upload an image or multiple images, and not use them as image "styling" prompts, but use our image uploads as "the subject". By then adding additional text prompts, for example....PLACE THIS IMAGE OF A PENGUIN ON A MOUNTAIN EDGE THAT USES THIS PASTA IMAGE AS INSPIRATION, the resulting image would be that exact penguin image uploaded, but placed on a conceptul background using the pasta image as inspiration.

A simple Styling Weight prompt could increase/decrease the level of styling applied to any element of the image where we require styling, but the point is that there should be the option to turn styling off completely and have Firely understand that certain images we upload need to be maintained in their original form, and only altered if we prompt Firefly to alter the uploaded image.

Example: I upload an image of a woman (I've already masked her out so she is on a plain white background) and I add a simple text prompt: PLACE THIS WOMAN IN A BUSTLING PARIS STREET FULL OF PEOPLE, APPLY STRONG BOKEH AND MAKE HER THE CENTRAL FOCUS OF THE IMAGE. The resulting image has not produced a STYLED interpretation of the image upload - it has created an image using my exact image upload, and placed her in a Paris street etc, but it has rendered her perfectly into the scene, having matched her scale, perspective, and applied all correct lighting and shadows. 

I sense that I am stating the obvious with how this SHOULD work, and that Adobe know full well that this is actually the functionality we were all really hoping for so I will leave it there. 

[simonh5504215]

Uploading style reference images is a cool thing, but it currently seems to totally override the text prompt that goes with it. Perhaps it should have its own influence slider on the picture so that you can dial up and down how much the uploaded image overrides the text prompt.

[thekohlervillager]

I really, REALLY would like the ability to upload my own photos and generate art from them by changing the styles (watercolor, etc.)

[EdLee1107]

It would be a great idea if you could add a feature where you import an image and firefly will change the style of the image. Example, You import an image of London, wide angle, and describe what you want it to look like etc Cyberpunk, and firefly will make the image look... futuristic without changing any defining features. Like keeping the overall layout of the image and just touching it up.