Available now

Wan 2.2 A14B Image with LoRAs: text to image

By Alibaba

Wan 2.2 A14B Image with LoRAs makes images from a written prompt and lets you add your own LoRA weights to set a style or character.

Model ID
Maker
Alibaba

Input

Fields marked * are required. Review the price before generating.

The text prompt to guide image generation.

Example prompt. Edit it or write your own.

Additional settings

What to keep out of the image.

Random seed for reproducibility. If None, a random seed is chosen.

Number of inference steps for sampling. Higher values give better quality but take longer.

Acceleration level to use. The more acceleration, the faster the generation, but with lower quality. The recommended value is 'regular'.

Classifier-free guidance scale. Higher values give better adherence to the prompt but may decrease quality.

Guidance scale for the second stage of the model. This is used to control the adherence to the prompt in the second stage of the model.

Shift value for the image. Must be between 1.0 and 10.0.

LoRA weights to be used in the inference.

Price$0.08 / run

You sign up first. Your settings are kept.

About this model

Overview

This card runs the 14B Wan 2.2 model for still images in six square, portrait or landscape sizes. Add LoRA files by link, each with its own scale, to bring in a trained style, character or product.

The model works in two stages, each with its own guidance scale. Steps run from 2 to 40, and a shift setting and optional prompt expansion give further control.

What you can build

  • Images in a style you trained as a LoRA
  • A recurring character kept consistent across scenes
  • Product shots with a LoRA trained on your product
  • Tests of how strongly a LoRA should apply

Prompting tips

  • Include the LoRA's trigger word in the prompt if it was trained with one.
  • Start with a LoRA scale of 1 and lower it if the style overpowers the prompt.
  • Leave acceleration on regular unless fine detail looks soft.

What to check

Compare the result with your source, especially object boundaries, fine text, faces and shadows.

Inputs

  • Prompt (required): The text prompt to guide image generation.
  • Negative Prompt (optional): What to keep out of the image.
  • Seed (optional): Random seed for reproducibility. If None, a random seed is chosen.
  • Number of Inference Steps (optional): Number of inference steps for sampling. Higher values give better quality but take longer. From 2 to 40.
  • Enable Prompt Expansion (optional): Whether to enable prompt expansion. This will use a large language model to expand the prompt with additional details while maintaining the original meaning.
  • Acceleration (optional): Acceleration level to use. The more acceleration, the faster the generation, but with lower quality. The recommended value is 'regular'. Options: none, regular.
  • Guidance Scale (1st Stage) (optional): Classifier-free guidance scale. Higher values give better adherence to the prompt but may decrease quality. From 1 to 10.
  • Guidance Scale (2nd Stage) (optional): Guidance scale for the second stage of the model. This is used to control the adherence to the prompt in the second stage of the model. From 1 to 10.
  • Shift (optional): Shift value for the image. Must be between 1.0 and 10.0.
  • Loras (optional): LoRA weights to be used in the inference.
  • Image Size (optional): The size of the generated image. Options: Square HD, Square, Portrait 4:3, Portrait 16:9, Landscape 4:3, Landscape 16:9.
  • Image Format (optional): The format of the output image. Options: png, jpeg.

Output and limits

  • Image sizes: Square HD (1024×1024), Square (512×512), Portrait 4:3 (768×1024), Portrait 16:9 (576×1024), Landscape 4:3 (1024×768), Landscape 16:9 (1024×576).
  • Small or dense text in the image can come out misspelled.
  • Hands, crowds and complex poses can show artifacts; generate a few variations and pick the best.
  • Some prompts or outputs can be refused by safety checks.

Price guide

  • up to, default options: $0.08 / run