Trained on 16 images of Joan Holloway from the Mad Men TV series. For the ComfyUI workflow, download this image and drag it into ComfyUI: https://civarchive.com/images/48444751.
This isn't a Christina Hendricks LoRA per se since it was exclusively trained on Mad Men content (stills/promo photos, etc.). As such her style will be mid-century unless you specify otherwise.
The training data is posted here for download as well. That will show you how the images and caption files are arranged and what the captioning style was (I used ChatGPT 4o). I also put my Hunyuan training config files (the two .toml files) in the .zip in case those are useful to anyone.
Showcase video prompts:
Joan stands confidently in front of a white background, holding a clipboard close to her chest with one hand while adjusting her glasses with the other. Her red hair is styled in a sleek bun, and she wears a fitted dress. She tilts her head slightly, her eyes meeting the viewer with a playful yet professional smile.Joan, her red hair arranged in a meticulous updo, stands poised in a classroom framed by the green chalkboard behind her. The shot focuses on her from the waist up, drawing attention to her curvaceous figure accentuated by her fitted outfit, which consists of a tight white blouse and a light blue floral skirt. She wears black glasses, exuding an air of mid-century sophistication. She taps her index fingers down on her desk. The camera advances, centering on the sharp intelligence in her eyes. Real professional HD video. High production quality. High resolution. Very detailed. Natural soft lighting.Description
FAQ
Comments (8)
I guess LoRA training for Hunyuan is still in an early state? Would love to train Jennifer Connelly :D
I see you created a Flux LoRA of Jennifer Connelly. If you want to share that training set with me I'm happy to spit out a Hunyuan Video LoRA.
@thebollo here you go https://we.tl/t-8oTbDZKGG4 it's a very specific dataset from "Career Opportunities" 😉
@el_fluppe Received. I'll try to get to this in the next few days.
Thanks for including your dataset. I was having trouble creating my character Lora. Turns out 256 resolution works with training for motion, but does not work with training characters. That's probably obvious to most but it tripped me up.
Good to know! Thanks for sharing.
I guess the fact she is a real actress means the moderators are a bit careful about what they allow. I tried to add a video I created with your lora, where I was trying to replicate Audrey Horne but it was disallowed. It was fun to play with, thanks.
One of the showcase videos I posted for this one was likewise removed for a terms of service violation and it wasn't especially salacious, just a view from above. This is my first experience creating a celebrity LoRA. It does seem like these are much more controlled.