02/18/25: Works great with DeepSeek captioning!
Try asking something like "Can you help me come up with a good FLUX prompt to generate an action figure in a package for (SUBJECT)?"
Built up from my Action Figurator Pony model, this was an experiment in trying out a concept FLUX already knows fairly well to see how much it can improve results - in this case making realistic looking, consistent, action figure packages.
"But FLUX can already do this with the right prompting!" I hear you say.
I put on a sly smile and reply "Yeah, you're right. But this makes it BETTER. Just look at this comparison sheet!"

Then everyone applauds.
Recommended strength: 0.8-1.0
Good sample prompts are supplied in the image gallery, and here's some mad libs as well.
Top:
The image is a photograph of a boxed action figure from the "____" line, featuring a character named _____
This image is a photograph of a packaging box for a figure toy being held by a hand.
The image is a photograph of a detailed, highly realistic action figure display featuring a _____-themed character.
Middle:
The box features a detailed display of a highly realistic _____ figurine dressed as a ____.
The main figure is an ____, standing upright on the front of the box. The figure has a _____ build, wearing ______.
The figure is a muscular, anthropomorphic _____, depicted in a pose reminiscent of a ____. It has a humanoid body with a strong build, detailed _____ skin texture, and a prominent jawline. The _____ wears ____.
Bottom:
____ accessories include a stethoscope and syringe.
The box the figure is in has the title "______"
Extra text: "____!"
Description
FAQ
Comments (16)
can't agree more. ppl keep saying flux is powerful and no need for loras. but good loras are like a scope to help you aim better with your rifle. great lora btw. love it.
Thanks! Yeah, really I just wanted to see if the LORA would improve quality over base prompting, and figured I'd post it when I had enough proof that it does!
Super cool!! 😱✨
Took a bit of creative prompting and turning cfg up a bit more than usual but Awesome results.
I now have my own action figure thanks!
Fartman? BOOGERMAN!
ok these look like bootleg toys they sold in every russian kiosk in 2006
what a time to be alive
I'm getting this error "Given groups=1, weight of size [512, 16, 3, 3], expected input[1, 4, 152, 104] to have 16 channels, but got 4 channels instead"
I'm not sure what would cause that, is that from the CivitAI image generator?
This is seriously one of my favorite loras for flux. so much fun. been making all my friends into action figures
Interesting, how? Can you elaborate?
@herpderpin7 sorry for the late reply but i basically just use my own custom loras of my friends and use fal.ai.com playground to add the action figure lora with my custom loras
This is amazing. Its so much fun
Great work
This is very well done!
Details
Files
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.
















