Shannon - Wan

Introducing Shannon. She is a red-haired woman with pale skin, green eyes, and a long slim arms and legs. The Shannon LoRA has been tested with both Wan 2.1 and 2.2. The results are a little different between the two, more on that below. A few topless and nude images were included in the training, so the LoRA won't need very much help with nudity.

Important: This model is 100% fictitious and is not based on images of any real person. She originated with images generated from still image checkpoints and has evolved from there using only AI rendering. No images (neither actual nor AI-generated) of any real person was ever involved in the making of this model.

UPDATE: For the best results, please give my workflow a try. This workflow is what I use to make all my videos using my models and it produces great results in Wan 2.2 using LoRAs that were created with the Wan 2.1 model.

All of the sample videos in the main showcase above were rendered with Wan 2.2 using a combination of Lightx2v version 2 for the high noise and Lightx2v version 1 (32 bit) for the low noise. The BoobPhysics LoRA was also used in all the sample videos. No other LoRAs were used, but the Shannon LoRA was applied at different strengths in different shots. I will try and call out the strength numbers in the comments for each video since that does affect the output.

Prompting hints:

The keyword is "shannon". She should generally appear with long, wavy/curly red hair and green eyes. If she doesn't, then you can add that to the prompt for a little extra emphasis. Or don't - they're your videos, you choose.

If you render with Wan 2.1, the LoRA is pretty simple. Any workflow you are comfortable with will work and the standard strength of 1.0 with your favorite other settings should work fine. Shannon is quite a bit more busty by default in Wan 2.1. She needs help with strength settings and prompt guidance to get the same body shape in Wan 2.2.

For reference, I published a few of my training images to show what I'm trying to achieve in the Wan LoRAs. The images are here: https://civitai.com/posts/20752644

Warning, complicated Wan 2.2 rendering stuff below.

Some things I've learned about applying LoRAs in Wan 2.2... It has been published and discussed that the high noise first pass covers overall shapes, not fine details, as well as motion in the frame (not fine detail motion). With that in mind, you can apply LoRAs and strengths in the first pass LoRA stack that influence the overall shapes and motions, leaving detail LoRAs for the 2nd pass stack. What does that mean for Shannon?

In short, her figure is more exaggerated the more strength you apply in the first pass. She will be more busty, with a thinner waist, and longer limbs. The second pass is more important for things like her face details, eyes, lips, skin textures and freckles, and so on. A higher strength there will enhance those features. However, a higher strength often changes the motion in the final scene. Since this model was only trained with still images, boosting the strength will reduce the overall motion in the scene. So it's a compromise. I tested a variety of combinations and found several that look good, so you will just have to experiment to decide what you like. Also remember: sometimes prompting can accomplish some of the same things. You can specify "very thin waist" for example without using a high strength on the LoRA.

Also remember that using speed enhancers like lightx2v and causvid will usually reduce, or at least change, the motion in the scene. If you need more motion, you can render without those enhancers, or as a compromise, you can render with plain Wan for the first pass, then apply a speed enhancer for the second pass. Just remember to adjust your CFG and steps for the two different pipelines.

All of that said, my samples tended to use one of just a few different combinations:

• Shannon at 1.0/1.0 - great motion, very pretty model

• Shannon at 2.0/1.0 - usually less interesting motion, more busty similar to Wan 2.1 output

• Shannon at 1.25/1.25 or 2.0/1.25 - slightly enhanced facial features approaching the "ideal" from my training set.

Please feel free to mix and match your own values. The numbers you use will also likely be different if you don't use lightx2v or causvid.

I hope you enjoy making movies with Shannon. If you do, please post the videos here or tag them with this LoRA when you post so I can see what people are doing with the model. Enjoy!