Frozen elsa -PonyXL attempt to match the original character

This time I bring you a lora for PonyXL checkpoint training, this lora is the same cue word as the SD1.5 version I did earlier.

The "pony" checkpoint is able to generate the character elsa directly, but it's not the same as the image in the film, it's more like a fan made work, personally I like the image in the film more, so that's why I'm making this lora.

The prompts I use are almost all random copies.

To avoid clutter, when I uploaded the lora file, I changed the name of the lora used in the example image from "elsa_xl4" to "Frozen_elsa_textV1-PonyXL", so you can change it back after the download is complete if you care.

Key Word:

elsa's outfit-1: Elsa's Sparkling Blue Dress (braid,braid on right shoulder)

elsa's outfit-2: Elsa's Coronation Dress

elsa's outfit-3: Elsa's Floral Dress

elsa's outfit-4: Elsa's Winter Dress

elsa's outfit-5: Elsa's Light Purple Dress

elsa's outfit-6: Elsa's Purple Nightgown

elsa's outfit-7: Elsa's Travel Outfit

elsa's outfit-8: Elsa's Dark Sea Dress (wet hair)

elsa's outfit-9: Elsa's White Dress (long straight hair)

Guidelines for use：

If you want to get the same result as me:

1. I'm using the webui-forge version.

2. Use the checkpoint "ponyDiffusionV6XL_v6StartWithThisOne".

3. Enable the "FreeU Integrated" default setting.

4. Enable "SelfAttentionGuidance Integrated", "Scale:0.18", "Blur Sigma:1.8", of course you can play around with this, but this parameter will slightly affect the content of the image.

5.Turn on "ADetailer" to repair your character's face.

Notes:

I’m using the online training service on civiai, which only gives me a few options for parameters and two optimizers (the ones I use less often). I can’t train as I would on my home computer. This is a test version, so I can’t fine-tune the training steps for each image. That’s why you might see elsa’s standing pose a lot when you use higher lora weights. Some poses are overtrained and lack flexibility, but don’t worry, You can use promprt adjustment or uses lower lora weights (like 0.6) and still ensures complete character features.