Something just like this ft. Adobe Firefly and Van Gogh

I recently came across a video on social media that has used Midjourney to generate artwork pieces in Dali’s style using lyrics from Bohemian Rhapsody as prompts. I thought of doing a similar experiment with prompts on #adobefirefly – if the platform is capable enough to generate such interesting artwork with basic prompts, considering it operates on comparatively limited datasets of licensed images.

I picked one of my favorite songs and my favorite painting of the artist I love to the extent of obsession :D.

Here is the output made by combining various artworks (on Adobe Premiere Rush) created by using lyrics from ‘Something just like this’ as prompts:

Breaking down the video:

• Prompts: Out of the 16 images used, 12 have the exact song lyrics as the prompts, without any edits. The rest four were a little tricky, but the ‘Suggestion’ feature was quite handy.
  
• Copyrighted material: One interesting thing about Adobe Firefly is that it works only on the datasets of its own licensed imagery. Hence, the artwork generated on the platform will not contain any copyrighted material (in this case Spider Man and Batman are copyrighted characters). However, it will pull the closest thing to the copyrighted material. This is a feature that makes it safer for commercial use (only the features out of Beta phase)
• Image style: To have the artwork in ‘Starry Night’ style, I picked an image from google to use as the reference (a ‘Match’ feature that Model 2 of Firefly provides) and I didn’t have to add the style in my prompts. Also, the painting is available in the public domain so no rights issue here. Adobe also checks the metadata of the reference image to ensure no IP infringement and no copyrighted material is added there.
• Metadata: Every artwork will have metadata containing the content credentials, mentioning that the artwork has been created by Adobe Firefly – definitely need of the hour to avoid misinformation and check for authenticity of the artwork.
  
• Understanding the process: The Gen AI models don’t work in a linear way. Hence, its highly unlikely that you are going to get the same output despite putting the same prompt. Every output is going to be unique which makes the Gen AI even more interesting – you just don’t know what you are going to get.

While Adobe was a little late to enter the Gen AI race, they definitely understood the market better – the inhibitions related to the mass adoption of Gen AI due to the legalities and associated risks. With a friendlier UI, integration with Creative Cloud, content authenticity and consideration of copyrights, Adobe Firefly is comparatively more efficient and safer platform to use.