Description:
This workflow utilizes the cutting-edge Z-Image Turbo model (based on AuraFlow architecture) to generate high-fidelity images in just 9 sampling steps. It pairs the diffusion model with a massive Qwen 3.4B (Lumina 2) text encoder, allowing for exceptional prompt understanding and adherence to complex descriptions. The setup is optimized for speed without sacrificing quality, producing 1024x1024 images with rich details.
Key Features:
- Turbo Speed: Configured for extremely fast generation (9 steps) using the Euler Simple scheduler.
- Next-Gen Text Encoder: Uses Qwen 3.4B (Lumina 2 type) for superior natural language processing, far exceeding standard CLIP models.
- AuraFlow Architecture: Implements ModelSamplingAuraFlow for optimal flow matching and image convergence.
- Character/Style LoRA: Includes a pre-configured LoRA loader (set to use lenovo_z.safetensors in the example) to demonstrate style injection.
Links to my patreon and youtube:
https://www.youtube.com/@ASF_One-Click-PRO