Qwen Image2

Next-generation Qwen image model for text-to-image generation and guided image editing with strong text rendering.

Model Overview

Qwen Image2 is a general-purpose image model designed for both prompt-based generation and reference-image editing. It supports visual transformation workflows while maintaining reliable prompt adherence and text rendering quality.

Best At

• Generating high-quality images from detailed text prompts.
• Editing existing images with semantic instructions and style guidance.
• Producing cleaner in-image text than many general image models.
• Rapid ideation with optional automatic prompt expansion.

Limitations / Not Good At

• Final quality still depends on prompt clarity and specificity.
• Very aggressive edits can drift from original composition.
• Complex multi-subject layouts may require iterative prompting.

Ideal Use Cases

• Marketing and social visual generation from prompt briefs.
• Fast concept variations from a base reference image.
• Style transfer and visual refinement workflows.
• Bilingual text-in-image creative assets.

Input & Output Format

• Input: required prompt; optional image, negative_prompt, aspect_ratio, enable_prompt_expansion, match_input_image, and seed.
• Output: generated image URI(s) returned on response.

Performance Notes

• Enabling prompt expansion can improve first-pass quality for underspecified prompts.
• match_input_image is useful for edit workflows when preserving source framing matters.
• Use a fixed seed for reproducible iterations.