Image Generation ModelDeprecated

SDXL LoRA

SDXL LoRA is an image generation model from Stability AI that supports custom LoRA adapters for fine-tuned visual styles.

Start Building with SDXL LoRA View All Models

Publisher

Stability

TypeImage

Context Window10,000 tokens

ReleasedJuly 2023

Replaced byFLUX.1 [schnell] LoRA

Provider

WaveSpeed

LoRA

Try SDXL LoRA →

About SDXL LoRA

Customizable image generation with LoRA adapters

SDXL LoRA is an image generation model built on Stability AI's Stable Diffusion XL architecture, released in July 2023. It extends the base SDXL model by natively supporting Low-Rank Adaptation (LoRA) weights, which are lightweight fine-tuning modules that modify the model's output toward a specific style, subject, or aesthetic without retraining the full model. This makes it distinct from the base SDXL model in that users can load one or more LoRA adapters at inference time to steer generation toward a desired look.

SDXL LoRA is well suited for workflows that require consistent visual styles, character likenesses, or domain-specific imagery that a general-purpose model would not produce by default. Inputs include image dimensions (width and height), guidance scale, number of inference steps, a negative prompt, a seed for reproducibility, and one or more LoRA adapter references. The model is available through the Wavespeed provider on MindStudio at a cost of $0.001 per image.

Capabilities

What SDXL LoRA supports

LoRA Adapter Support

Accepts one or more LoRA adapter weights at inference time to steer image output toward a specific style or subject without full model retraining.

Custom Image Dimensions

Accepts explicit width and height inputs so generated images can be sized to match a specific layout or aspect ratio requirement.

Guidance Scale Control

Exposes a guidance scale parameter that controls how closely the output adheres to the text prompt versus allowing more generative variation.

Inference Step Configuration

Allows the number of diffusion inference steps to be set explicitly, giving control over the trade-off between generation speed and image detail.

Negative Prompt

Accepts a text-based negative prompt to suppress unwanted elements, styles, or artifacts from the generated image.

Reproducible Seeding

Supports a seed input so that identical prompts and settings produce the same image output, enabling reproducible generation runs.

Ready to build with SDXL LoRA?

Get Started Free

FAQ

Common questions about SDXL LoRA

What is a LoRA and how does it work with this model?

LoRA (Low-Rank Adaptation) is a lightweight fine-tuning technique that adds a small set of trained weights on top of a base model. With SDXL LoRA, you can supply one or more LoRA adapter files at inference time to guide the model toward a specific style, character, or visual domain without modifying the underlying SDXL weights.

How much does it cost to generate an image with SDXL LoRA on MindStudio?

The published price is $0.001 per image, served through the Wavespeed provider.

What context window does SDXL LoRA support?

The model has a context window of 10,000 tokens, which applies to the text prompt inputs used to guide image generation.

Can I control the resolution of generated images?

Yes. The model accepts explicit width and height number inputs, so you can specify the pixel dimensions of the output image.

Does SDXL LoRA have a knowledge cutoff date?

SDXL LoRA is an image generation model, not a language model, so a knowledge cutoff date is not applicable. The model was released in July 2023 and generates images based on prompts and LoRA adapter weights rather than a training knowledge base.

Resources