Qwen Image Max Edit for Editing Images

Qwen Image Max Edit is the high‑end editing version of Qwen Image Max: it takes one or more images plus a text instruction and performs precise, context‑aware edits while preserving composition, lighting, and style.

Overview

• Built on the same 20B backbone as Qwen Image Max, with a dual‑path architecture: a multimodal encoder for semantics and a VAE‑style path for appearance (color, texture, lighting).

• Extends Qwen’s strong text rendering into editing, so it can add, remove, or modify text in images while keeping fonts, size, and typographic style consistent.

What it does well

• Precise text editing: Add, change, or delete text in signage, UI, labels, and documents in both English and Chinese, keeping layout and font appearance.

• Semantic edits: High‑level changes like changing outfits, rotating or re‑posing objects, re‑styling characters, or altering scene layout while keeping identities intact.

• Appearance edits: Low‑level operations such as color correction, style transfer, adding/removing objects, or adjusting background elements.

• Multi‑image edits: Up to about 6 input images can guide the result (for example, keep character from image 1, outfit from image 2, background from image 3).

Key capabilities / parameters

From a typical API like Qwen Image Max Edit:

• Inputs:

  • prompt (required): text describing the edit, up to ~800 characters.

  • images (required): 1–6 images, usually between 384–5000 px on the long side.

• Size control:

  • size for preset ratios (1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3), or explicit width/height (256–1536 px) if you want custom resolution.

• Output format: jpeg, png, or webp, defaulting to JPEG.

• seed: for reproducibility across runs.

Why use it

• Combines layout‑aware text editing with high‑quality visual editing, so you can update both imagery and on‑image copy in one pass instead of juggling Photoshop plus a T2I model.

• Bilingual support makes it ideal for localization workflows where you must swap signage, labels, or UI text between languages without re‑designing assets.

• Multi‑image semantics let you enforce character or product consistency across edits, which is crucial for brand assets and multi‑panel designs.

Typical use cases

• Marketing & product: Change packaging text, swap colors or variants, add/remove props, or localize copy on the same hero image.

• UI and docs: Fix or translate UI labels, charts, and document screenshots while preserving layout and typography.

• Creative and style transfer: Re‑skin characters, apply new art styles, or adjust environments (season, time of day) while keeping core composition and identities.