Blake Belladonna (RWBY)

Hi gents! This is a Wan 2.1 lora meant to generate realistic pics and videos of Blake Belladonna from RWBY. To see the Yang Xiao Long counterpart, see here: /model/1835966/yang-xiao-long-rwby

The next step in this project is to make a model that will be able to generate pics and videos of Yang and Blake together and separately. I know I've attempted this before and failed (multi-character loras are typically where I run out of steam), but that was before Flux Kontext became a thing. Now, I can generate pics of Yang and Blake separately and then stitch them together with Kontext, before putting that into my dataset.

I think that an addition of 15 photos + 5 videos of Yang and Blake together, on top of 30 pictures of each of them separately + 10 videos, should be enough for Wan 2.1 to learn how to generate them together. I'll test it out and then share with you guys.

This lora is meant to be used with the Wan 2.1 FusionX T2V model. I use the default recommended lora for the video generations. For the image generations, I use a modified version of this workflow: https://drive.google.com/file/d/1WeH7XEp2ogIxhrGGmE-bxoQ7buSnsbkE/view

The modifications I made were that I changed the model from default Wan 2.1 to Wan 2.1 FusionX. switched the text encoder to full FP16, replaced the lora model loader with the Power Lora loader and have two loras loaded (this one set at 1.0 strength and the smartphone lora set at .75 strength).

For the video generations, I use the default workflow found here with the only change being made is the addition of the lora: /model/1663553/wan2114b-fusionxworkflows

This is the most realistic and high-quality lora I have ever made; better than Flux by several miles. I hope you guys enjoy!