This model was trained using photos of weed nugs of different strains. Activation keyword is "weed". Each strain was labeled as indica, sativa, or hybrid weed so those keywords can also be used to add slight variation.
Weed Diffusion generates realistic images of marijuana nugs. It has been trained on a small dataset of real marijuana nug images and is able to generate highly realistic images that closely mimic the appearance of actual marijuana nugs.
Stable Diffusion 1.5 already generated pretty decent weed pictures without training but the training improved the realism immensely and tailored the results to be picked buds/nugs opposed to the plant growing from the ground.
Description
FAQ
Comments (5)
BIG LMAO!
Cool but a whole model? Couldn't this be done with a LoRA or hypernetwork?
this was made before Loras were a thing on here.
Wow this looks tasty.....
Thanks a lot ! This is awesome ! I used your checkpoint to generate training material for my first ever LoRa, which is about... you guessed it ! Weed ! lol
Not quite completely happy with the result (of my LoRa, not your checkpoint), I might need to put some more tasty realistic marijuana in the dataset...
Thank you again !
... You have a model that does what you want. EXTRACT a LoRA from it.
Generating AI images and training a model on already generated images is a terrible idea. AI DOES mistakes. Feeding it its own mistakes will end up in an infinite loop of terrible images reproducing constantly the same mistakes.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.