Authentic Dirndl LoRA for ZImage Turbo models.
No cosplay, no carnival – just real Bavarian style.

🧠 Description

Time flies – and spring is finally here 🌸
Which means: Volksfest season has officially started.

What better time for an update of the Dirndl LoRA?

This LoRA is designed for ZImage Turbo models in general, with a particular focus on Big Love Zia1 / Big Love ZT3, where it delivers the most consistent and high-quality results.

I previously experimented with an SDXL version, but was never fully satisfied with the lighting and overall flexibility. This version focuses on delivering more reliable and controllable outputs.

The LoRA was trained using the excellent AI Toolkit by Ostris on RunPod (my local machine simply wasn’t powerful enough 💻).

Trained on 130+ real Dirndl references for maximum authenticity.

Important:
The training dataset contains no faces, meaning this LoRA will not interfere with your characters or base model identities.

✍️ Prompting

For best results, follow this structure:

wearing dirndl  
[COLOR] bodice  
[COLOR] blouse with [long|short puffy] sleeves  
[COLOR] apron  
[COLOR] skirt

💡 Example:

wearing dirndl, black bodice, white blouse with short puffy sleeves, red apron,  black skirt

🎀 Additional Details (Optional)

You can enhance the look with accessories and fine details:

choker, ruffled collar, lace trim, silver chains, buttons, zipper, lace details, detailed textures, floral belt buckle

These help to create more realistic and high-quality Dirndl variations.

⚙️ Recommended Settings

Weight: 0.8 – 1.0
Works best with Big Love Zia1 / ZT3
Compatible with ZImage Turbo models in general
No trigger word required

⚠️ Limitations

Two things are still a bit tricky with current models:

The way the apron sits and interacts with the Dirndl skirt
And the position of the apron bow (left, right, etc.), which traditionally indicates the wearer’s relationship status

These details are part of what makes a Dirndl authentic – but they’re not always perfectly consistent yet.
And yes… the bow placement still has a mind of its own sometimes 😉

📌 Notes

Designed for authentic Dirndl looks (not costume-style)
Works with multiple poses and compositions
Can be used for both SFW and NSFW generations

Search AI Models & Addons

very few results