Qwen Image 2.0 Pro: AI Image Editor with Style Transfer & Compositing

Qwen Image 2.0 Pro — Image Editing

Qwen Image 2.0 Pro Editing lets you modify images using natural language instructions. Describe the change you want — background swap, object removal, style transfer, text overlay — and the model handles spatial reasoning and pixel-level transformations. The Pro tier delivers the highest editing fidelity for production-quality results at $0.075 per image.

What You Can Do

Style transfer — transform a photo into a watercolor, sketch, or anime style
Object insertion and removal — add or remove elements while maintaining visual coherence
Text editing — add, modify, or overlay text directly on images
Background replacement — swap environments while preserving the subject
Multi-image compositing — combine elements from multiple source images
Cross-domain editing — place illustrated characters into photographs, or vice versa

Key Parameters

Parameter	Default	Range	Notes
prompt	—	string	Natural language description of the edit
image_url	—	URL	Public URL of the source image
guidance_scale	4.5	1–20	4–7 for most edits; higher for strict prompt adherence
num_inference_steps	28	1–50	35+ recommended for Pro-quality edits
num_images	1	1–4	Generate multiple edit variations
seed	random	integer	Fix for reproducible results
output_format	png	png / jpeg / webp	Output file format

Quick Start

python
import fal_client

result = fal_client.subscribe(
    "fal-ai/qwen-image-2/pro/edit",
    arguments={
        "prompt": "Change the background to a sunset over the ocean, keep the subject sharp",
        "image_url": "https://example.com/your-photo.jpg",
        "num_inference_steps": 35,
        "guidance_scale": 5,
    }
)

print(result["images"][0]["url"])

javascript
import { fal } from "@fal-ai/client";

const result = await fal.subscribe("fal-ai/qwen-image-2/pro/edit", {
  input: {
    prompt: "Change the background to a sunset over the ocean, keep the subject sharp",
    image_url: "https://example.com/your-photo.jpg",
    num_inference_steps: 35,
    guidance_scale: 5,
  },
});

console.log(result.data.images[0].url);

Writing Effective Edit Prompts

Edit prompts work differently from generation prompts. Be specific about what to change and what to preserve:

Good: "Replace the blue sky with a dramatic thunderstorm, keep the buildings and street unchanged"
Bad: "Make it stormy"

Tips for better results:

Name the change explicitly — "change the background to...", "remove the person on the left", "add text reading 'SALE' in red at the top"
Mention what to keep — "preserve the subject", "keep the lighting consistent", "maintain the original colors of the foreground"
Use guidance_scale 4–7 for natural-looking edits. Higher values enforce the prompt more strictly but can introduce artifacts.
Increase steps for detail work — text overlays and fine object manipulation benefit from 35–50 steps.

Common Editing Tasks

Background Replacement


"Replace the background with a clean white studio backdrop, maintain soft shadows under the product"

Object Removal


"Remove the trash can on the right side of the image, fill the area naturally with the surrounding grass"

Style Transfer


"Convert this photo into a Studio Ghibli watercolor painting style, preserve the composition and character positions"

Text Overlay


"Add the text 'GRAND OPENING' in bold white serif font at the top center of the image with a subtle drop shadow"

Standard vs Pro

Use the standard editing endpoint ($0.035/image) for testing edits and iterating on prompts. Switch to Pro for the final output where precision and detail matter.

Related Endpoints

Image Editing (Standard) — faster, lower cost for iteration
Text to Image (Pro) — generate new images from text
API Reference — full parameter documentation

fal-ai/qwen-image-2/pro/edit

Input

Result

What would you like to do next?

Logs