Mary-Ann "Huckleberry" (green)

A fan-made Character LoRA for Mary-Ann "Huckleberry" from the comics by Jay Naylor.

This version was trained specifically with PG / PG-13-friendly and SFW use in mind, so the goal was to create a version of the character that works well for Civit Green generation while still keeping her recognizable look and personality.

If you're looking for a LoRA that can generate Mary-Ann reliably with just a simple prompt, this one should give you a very solid starting point.

✨ Quick Info

• Character: Mary-Ann "Huckleberry"

• Type: Character LoRA

• Trigger word: mary-ann-greenV1

• Base architecture: Pony Diffusion

• Training checkpoint: AutismMix SDXL (AutismMix_pony)

• Recommended LoRA weight: 1.0

✅ Green / PG-13 Focus

This LoRA was made with Green compatibility as a priority and is intended for PG / PG-13-style use only.

That means this version focuses on:

• recognizable character likeness

• flexible prompting

• clean and usable outputs

• compatibility with SFW generation workflows

🚀 Recommended Settings

These are the settings that worked best for me so far:

• LoRA weight: 1.0

• Steps: 30

• CFG: 7

• Sampler: DPM++ 2M Karras

Since this is the first version, you can still experiment a bit, but 1.0 is currently the recommended default.

🏷️ Trigger / Prompting

Required trigger:

mary-ann-greenV1

In most cases, the trigger word alone is already enough to generate the character quite well.

You can improve consistency further by adding appearance tags when needed.

Helpful optional tags:

• tanned skin

• blonde long hair

• twintails

• low twintails

• freckles

• young face

• wide hips

• wide thighs

• round breast

• shiny skin

Default outfit:

• blue overalls

• white t-shirt

👀 Default Appearance

This LoRA is mainly trained around Mary-Ann's recognizable core look:

• tanned skin

• long blonde hair

• low twintails

• freckles

• young face

• curvy build

• shiny skin

Her default outfit is:

• blue overalls

• white t-shirt

That said, the LoRA is not locked to one outfit only and can handle other clothing as well.

🧪 What Works Well

This LoRA performs especially well with:

• portraits

• full body shots

• multiple viewing angles

• different poses

• different outfits

It should work nicely for both simple and more detailed character prompts, and in my testing it handles a good range of angles better than expected.

⚠️ Known Weaknesses

Current weak points:

• hands

• extreme perspectives

• multiple characters in one image

So if a generation gets messy, those are the first things I would simplify.

🧩 Checkpoint Compatibility

This LoRA was trained on:

• Pony Diffusion

• AutismMix SDXL (AutismMix_pony)

Other Pony-based checkpoints may also work, but I have not tested additional checkpoints yet, so feedback is very welcome if you try it on other models.

📝 Example Prompts

1. Simple portrait

mary-ann-greenV1, (portrait), looking at viewer, tanned skin, blonde long hair, low twintails, freckles, young face, shiny skin, smiling, looking at viewer, brown background,

2. More detailed character prompt

mary-ann-greenV1, wearing green oversized sweater, white shorts, ((thigh high socks)), sitting on a wooden bench, background public park,

💬 Final Notes

This is v1.0, and overall I'm very happy with how well it captures the character already.

If you use it, feel free to share your generations and let me know:

• which checkpoint you used

• what LoRA weight worked best for you

• whether it stays consistent for your prompts

• where it still struggles

Feedback is always appreciated and will help improve future versions.

Have fun with it!