LoRA Integrations

LoRA integrations let you attach fine-tuned model weights to your designs. Add LoRA URLs from Replicate, HuggingFace, or CivitAI, configure integration parameters, and combine them with your style prompt for precise visual control.

Opening LoRA Integrations

From design edit mode, click the “LoRA Integrations” button at the bottom of the edit form. This opens a modal with a model picker dropdown at the top. Not all models support LoRAs — only compatible models appear in the picker.

Model Picker

The dropdown at the top of the modal lets you choose which image model to configure LoRA integrations for. Only models that support LoRA integrations appear in this list. Switching models loads that model’s LoRA parameters and any previously saved values.

LoRA URL Input

A text input for the LoRA URL. Supported formats:

Replicate: owner/model (e.g., jbilcke/sdxl-cinematic)

HuggingFace: full URL (e.g., https://huggingface.co/user/model)

CivitAI: full URL (e.g., https://civitai.com/models/12345)

Direct: .safetensors URL from any host

Browse links below the input open CivitAI and HuggingFace in new tabs for discovering LoRAs.

Integration Parameters

Other model-specific parameters appear below the URL input. These use the same control types as Advanced Settings: sliders, number inputs, enums (button groups), and booleans (checkboxes). Common parameters include LoRA scale and weight, which control how strongly the fine-tuned weights influence the generated image.

Save & Reset

Save preserves LoRA parameters alongside your general and advanced parameters. Reset clears only the integration parameters for the current model — it does not affect advanced settings. Each design + model combination stores its own LoRA configuration independently.

Combining LoRAs with Style Prompts

Best results come from using LoRAs and style prompts together. The LoRA provides fine-tuned visual weights — character likeness, a specific art style — while the style prompt provides scene-level direction such as lighting, mood, and composition. Together they give you precise, consistent control over your generated images.

For more on LoRAs themselves, see the LoRA Guides section.