Z Image Turbo LoRA: Fast Text-to-Image with Bilingual Precision

tongyi-mai/z-image/turbo/text-to-image/lora

The advanced LoRA-enabled version of Z Image Turbo. Load up to 3 custom LoRAs via URL/Path to mix styles and characters with high-speed generation.

Prompt *

LoRAs

LoRAs 1

LoRA Path

URL, HuggingFace repo ID (owner/repo), or local path to LoRA weights.

LoRA Scale

Scale of the LoRA model.

List of LoRAs to apply (maximum 3).

Width

The width of the generated image in pixels. Each dimension should be between 256 and 1536.

Height

The height of the generated image in pixels. Each dimension should be between 256 and 1536.

Seed

The random seed to use for the generation. -1 means a random seed will be used.

Idle

The rate is $0.0085 per image.

Introduction to Z Image Turbo LoRA

Z Image Turbo LoRA is a specialized playground built for applying custom styles and character adaptations to the ultra-fast Z Image Turbo model. Unlike the standard version, this tool unlocks the ability to inject up to three external LoRA models simultaneously via URL or HuggingFace ID. It allows creators to blend specific art styles, fix character identities, or apply concept adjustments on top of the high-speed Turbo generation engine, delivering customized visuals in seconds.

Examples Created with Z Image Turbo LoRA

Z Image Turbo LoRA on X Platform

Key Capabilities

External LoRA Injection: Supports loading custom LoRA weights directly via URL, HuggingFace path (owner/repo), or local path, giving you infinite expandability beyond the base model.
Multi-Style Blending: Load up to 3 distinct LoRAs in a single run. Mix a "Character LoRA" with a "Style LoRA" to create unique combinations.
Precision Weight Control: Fine-tune the influence of each LoRA using the scale slider (0.0 to 4.0). Dial in subtle enhancements or force heavy stylization.
Turbo-Speed Customization: Retains the sub-second inference speed of the Turbo architecture while applying complex model adaptations.

How to use Z Image Turbo LoRA

Define Your Prompt: Write a standard text description of your subject.
Load LoRAs:

- In the LoRAs list, add an item.

- Path: Paste a download link or HuggingFace ID (e.g., renderartist/Technically-Color-Z-Image-Turbo).

- Scale: Set the intensity (Default is 1.0). Increase for stronger effect, decrease for subtlety.

- You can add up to 3 LoRAs to mix effects.

Set Resolution: Adjust Width/Height (256-1536px) to match your desired aspect ratio.

Example Configurations

Single Style Transfer:

- LoRA: renderartist/Technically-Color-Z-Image-Turbo (Scale: 1.0)

- Prompt: "A cyberpunk street scene..."

Character + Art Style:

- LoRA 1: user/anime-character-lora (Scale: 1.0)

- LoRA 2: user/watercolor-style-lora (Scale: 0.7)

- Result: The specific character rendered in a watercolor style.

Pro Tips

Finding LoRA Paths: You can use direct URLs to .safetensors files or use the username/repo-name format from HuggingFace.
Balancing Scales: If using multiple LoRAs, try keeping total weight around 1.0-1.5 to avoid burning the image (artifacts). If the image looks messy, lower the scales.
Turbo Compatibility: Ensure the LoRAs you use are trained for or compatible with the Z Image Turbo / SDXL Turbo architecture for best results.

Related Tools

For standard high-speed generation without custom LoRAs, use the base model: Z Image Turbo Text to Image.
For advanced control using ControlNets, try: Z Image Turbo ControlNet.

Related Models

wan-2-2/text-to-image

Generate high quality images from text prompts with Wan 2.2 Plus.

stepx-edit2

AI image editing from text with region control and brand consistency.

wan-2-5/text-to-image

Generate images from text prompts with Wan 2.5 Preview.

flux-1-1-pro/ultra/text-to-image

Dive into 2K worlds of photorealism.

z-image/turbo/inpainting

Precision-driven tool for photo retouching and visual reconstruction

qwen-image/qwen-image-edit-2511/lora

LoRA-based visual editing model offering structure-aware asset transformation for creative pros

Frequently Asked Questions

How fast is the Z Image Turbo Lora API compared to local hosting?

It is significantly faster and more scalable. Because Z Image Turbo Lora is engineered for low-latency generation, RunComfy’s auto-scaling GPUs allow it to handle high concurrency effortlessly. This eliminates the queue times and significant hardware overhead associated with running high-speed diffusion models locally on consumer-grade equipment.

What are the input limits for resolution and LoRAs?

You can generate images with dimensions ranging from 256px up to 1536px. The model accepts standard text prompts and allows you to stack up to three simultaneous LoRA models. This capability enables Z Image Turbo Lora to handle complex style blending tasks that typically require much heavier workflow configurations.

How do I deploy Z Image Turbo Lora to production?

Transitioning is seamless. Once you refine your parameters in the playground, simply switch to the "API" view to copy the JSON payload. You can then trigger the Z Image Turbo Lora model via standard HTTP requests (Python or cURL), ensuring your production app matches the exact quality and settings defined during testing.

RunComfy

RunComfy is the premier ComfyUI platform, offering ComfyUI online environment and services, along with ComfyUI workflows featuring stunning visuals. RunComfy also provides AI Models, enabling artists to harness the latest AI tools to create incredible art.