Diana from Pragmata — Z-Image Turbo LoRA

Updated the LoRA to volume 2! Matches more closes Diana from Pragmata! Enjoy.

A high-fidelity character LoRA for Z-Image Turbo capturing Diana from Capcom's upcoming title Pragmata. Trained on a 300+ image dataset of CGI-quality renders, this LoRA reproduces Diana's distinctive android aesthetic with exceptional detail — the circuit seam lines, her iconic blue tactical puffer jacket, and the cinematic lighting quality that defines the Pragmata visual style.

Model Details

Base Model Z-Image Turbo (Tongyi-MAI) Character Diana — Pragmata (Capcom) LoRA Rank 32 Alpha 4 Training Resolution 1024px Dataset Size 300+ images Optimizer AdamW Precision bf16 Trigger Word dprag

Usage

Trigger Word

dprag

Example Prompts

Portrait:

masterpiece, CG, extreme macro close up portrait, focus entirely on the face, dprag, a young android girl with striking, hyper-detailed grey-hazel eyes showing subtle mechanical iris rings, long straight blonde hair cleanly framing her face, subtle dark facial marking lines tracing her cheekbones. The high collar of her oversized blue tactical puffer jacket is zipped up to her chin, showing intricate fabric weave. Neutral expression, ultra-shallow depth of field with the ears and background completely blurred, soft volumetric lighting, dramatic shadow details, 8k resolution texture.

Action shot:

masterpiece, CG, dynamic wide-angle shot through a shattered glass window, dprag mid-action. Long straight blonde hair catching fractured glass reflections, intense grey-hazel eyes, subtle dark facial marking lines. She is firing a massive, over-engineered plasma assault rifle one-handed, the weapon venting green exhaust plasma gases. Wearing an oversized blue tactical puffer jacket with intricate harness straps, dark knee-high armoured boots kicking off a concrete pillar. Hundreds of tiny, razor-sharp glass shards floating in mid-air reflecting vivid magenta and cyan cyberpunk neon signs, high-speed shutter effect.

Close-up:

masterpiece, CG, medium portrait shot, eye-level angle, dprag, a young android girl with long straight blonde hair tucked behind her ears, piercing grey-hazel eyes, subtle dark facial marking lines. Wearing an oversized blue tactical puffer jacket with hard-surface integrated shoulder armour panels, intricate black harness straps forming a stark geometric pattern across her chest. Standing against a dark, weathered minimalist concrete backdrop, dramatic high-contrast Rembrandt lighting casting deep, artistic shadows across one side of her face.

Negative Prompt

(leave empty — Z-Image Turbo does not use negative prompts)

Recommended Settings

Setting Value Sampler res_multistep Scheduler simple Steps 8-9 CFG Scale 1.0 LoRA Strength 0.8 – 1.2 Resolution 768x1280 / 832x1216 / 1024x1024/ 1024x1280

⚠️ Z-Image Turbo requires CFG 1.0 and no negative prompt. Always use res_multistep sampler for best results.

Tips

• Portrait ratio brings out the best in Diana's face detail — try 1024x1280

• The circuit seam lines are more visible at higher resolutions

• LoRA strength 1.0 is the sweet spot — drop to 0.8 for a more subtle influence

• Diana's android features respond well to prompts that specify lighting direction — try rim light, golden hour, or cool blue studio lighting

• The blue puffer jacket detail is strongest at 1024px and above

ComfyUI Setup

1. Place the .safetensors file in ComfyUI/models/loras/

2. Add a Load LoRA node to your workflow

3. Connect between your Z-Image Turbo checkpoint loader and your sampler

4. Wire both Model and CLIP outputs through the LoRA node

5. Set strength to 1.0 and use dprag in your prompt

About

Part of an ongoing Z-Image Turbo character LoRA series. Diana was trained on one of the most carefully curated CGI character datasets built for this model — 300+ high quality images capturing consistent lighting, varied poses, and the full range of her distinctive android design details.

Full workflow tutorials, ComfyUI node setups, and early access to new LoRA releases available on Patreon.

Tested in ComfyUI with the standard Z-Image Turbo workflow. Drop a comment if you need help getting it working.

💙 Enjoyed this LoRA?

Every release is trained on curated datasets built from scratch. If you want to support more character releases, get early access to future drops, and access exclusive models that never go public — come join us on Ko-fi:

👉 https://ko-fi.com/winnougan

I'm also saving toward a hardware upgrade so I can expand into Klein9b and LTX-2.3 video LoKRs and deliver releases faster. Every coffee helps make it happen. ☕

💙 Discord:

Join my Discord, where I talk about all of my releases and entertain requests for future LoRAs!

👉 https://discord.gg/CJv5wceJaN