"Lay's Do Us a Flavor" is a promotional campaign launched by Lay's, a popular potato chip brand owned by PepsiCo. The campaign invites consumers to submit their ideas for new chip flavors, with the chance of winning a cash prize if their flavor is chosen. Started in 2012, the campaign has become an annual event, generating significant engagement and excitement among consumers. Finalists' flavors are produced and sold in stores for a limited time, allowing customers to taste and vote for their favorite. The winning flavor typically remains in production for a longer period, while the finalists receive monetary rewards for their contributions. "Lay's Do Us a Flavor" has been successful in fostering consumer engagement and innovation within the snack food industry.
Description
FAQ
Comments (14)
ah that caption feels like sponsored post.. is this?! ;)
No affiliation with PepsiCo or FritoLay.
@patricktoba don't lie. They paid you with gas butt sex lol
@X0l0t0l Could have sworn it felt more like beach sand
You came up with some pretty amazing example images there, thank you for the laughs :D
this had to be a contest just for fun
Example images had me fucking howling, thank you!
Amazing LoRA, I've been creating for two hours now! One problem though, it seems to integrate modifiers like "4k", "UHD", "masterpiece" and so on into the text, is there any way to prevent that?
Also, speaking of text: I found out that whenever there's double letters like in "egg", "tooth" and the like, the model tends to omit one of the letters. it doesn't like "th" for some reason too.
Here is an example prompt for best results of getting the right text on the bag:
lays_potato_chips_do_us_a_flavor, the text on the bag reads: "(((dmt)))", the bag art features a psychedelic man meditating
Thanks for the feedback and have fun!
very cool, perhaps you could tell us a little about the training for this LORA? How many images? what dataset? What sampler? how long? would be fun to see how these ultra-specific LORA's perform with various training strategies.
While I won't give away all of my secrets, I will say that this LORA was trained entirely off of AI generated images. I used an ipad app called Wonder AI to use their formatted access to ChatGPT and Dall-E3 to generate 54 curated images of imaginary Lay's chip flavors. I took all 54 obscure ideas for chip flavors complete with prompts implying abstract artwork for each piece and then I scheduled them for training using many epochs and repeats, ensuring the LAYS logo is all that and a bag o chips...
@patricktoba Cool. Sounds like a decent baseline for training! And well done btw! I didn't know only 54 images could be this effective! Never heard of WonderAI.
I never dare to go below 100 images, but cool! I wonder how that would compare to if you used actual photos of Lays chips bags. Maybe we would get some more coherence? Or would that imply just unnecessary work? Also, i wonder if you used the same chips bags but with photos of various angles, maybe you could generate multiple angles that way too, and thus generate more coherence that way?
Did you do anything special to make sure the Logo stays coherent? (I have a hard time generating coherent text and logos)
I'm so sad this SD Lay's chips no longer exists {"type":"lora","weight":1,"modelVersionId":413122,"modelName":"Lay\u2019s Potato Chips: Do Us A Flavor","modelVersionName":"Lay\u2019s Potato Chips: Do Us A Flavor"}. Cheers on the SDXL & F1
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.