This is a collection of LoRAs that capture the actress at different points in her life/career. This allows for more granular control of the look and feel of the pictures you want to create.

Check out the sample pictures for each LoRA for ideas and examples.

Celebrity LoRAs

The LoRAs of the actress herself are divided into three arbitrary eras.

Harry Potter Era (Ages: 18-21): When she was actively filming and promoting the Harry Potter movies (not trained with any images of the character Hermione). Also includes when she famously cut her hair to a pixie cut just after filming the last movie. This model trends toward her more youthful look, but not always All photos used to train were 18+.

Movie Era (Ages 22-30): When she was acting in various movies from Perks of Being a Wallflower to Little Women and was the most active in social causes. This model gives the widest range of looks. This is probably seen as her "default" look as most other LoRAs are based on this age range.

Prada Era (30+): When she became the new face of Prada (Technically starting at age 32). This definitely gives a very different look than her previous "eras" and includes another iteration of the pixie cut as well.

There is a lot of variety in each set of training data so when different hair style/color is specified it will match to how she actually looks with those hairstyles, so use modifiers like: short hair, long hair, medium hair, brown hair, blonde hair, wavy hair, messy hair, etc. I also include 'pixie cut' if you want to match when she's cut her hair really short (the only samples are in the HP Era and Prada Era LoRAs).

The training data also includes a lot of different outfits, which can be a double edge sword because if you don't specify a look / clothing, there can be times when she'll generate in a very fancy dress from one of the many award/fashion shows she's been a part of.

In addition pose and POV modifiers should work a lot easier and maintain her look: "from side," "from behind," "looking back", etc.

---

Character LoRAs

There are LoRAs that represent different iconic characters the actress has played. Both LoRAs include movie stills and promotional images. Use "movie still" in the prompt to better control an image to look more like the style / lighting of the movie itself.

Hermione: You can prompt for "Hogwarts School Uniform". you may need to add "emblem" if you want one to appear. You can also prompt "holding wand". All of the movie still photos used in training are from the two Deathly Hallows movies, so clothing style will trend to what she wore in those movies.

Belle: You can prompt for "peasant dress" to get the blue and white dress she wears at the beginning of the movie; "winter outfit" with "red cloak" to get the outfit she wore in winter (song "Something There"); and the iconic "yellow ball gown."

----

Yet Another Emma Watson Lora; Why?

There are a few dozen Emma Watson LoRAs across every base model from SD 1.5 to Flux including several really good and flexible ones in Flux alone. So what's special about this one?

I sorted through hundreds of pictures, picked and categorized the best ones, used both automated and manual tagging, normalized each set, and went through multiple training cycles to make sure I got the right one each time. Why?

The honest answer is: I have a ton of credits and just wanted to see what I could make if I was extremely meticulous. I do this for fun, so I wanted to see if I could make my own LoRA of my favorite actress since I use other versions all the time. At first I was going to try and make one LoRA hold all of it combined with different triggers for each "era" or "role", but then I decided to go for this method instead as it provides more granular control. I'm happy with the results, but feedback is more than welcome. I'll take any constructive criticism and use it to make these LoRAs better in the future.