# Qwen (I2I)

> Qwen (I2I) is available on Artifio.ai — a unified AI workspace with a single pay-as-you-go wallet.

- Canonical page: https://artifio.ai/models/qwen-i2i
- Run it: https://artifio.ai/create/qwen-i2i
- Modality: image editing
- Model brand: Qwen
- Pricing: from about $0.005 per generation (exact cost shown in the workspace before you confirm; varies with selected options)
- Typical generation time: ~25s (catalogue estimate)
- Full pricing table: https://artifio.ai/pricing/models

## About Qwen (I2I)

Qwen (I2I) is the image-to-image editing entry for Alibaba's Qwen-Image family, built on the Qwen-Image-Edit line that the Qwen team open-sourced in August 2025. The editing model extends the 20B Qwen-Image foundation and processes every input picture down two paths at once: Qwen2.5-VL reads it for semantic control while a VAE encoder captures its exact appearance. That dual encoding is what lets an edit change the meaning of a scene or leave every untouched pixel exactly where it was, depending on the instruction.

The family inherits Qwen-Image's text-rendering strength, so edits that add, delete, or rewrite words inside an image keep the original font, size, and style. On Artifio it runs pay-per-generation in the same workspace and wallet as the rest of the catalogue.

## Features

- Semantic editing: style transfer, character re-posing, and novel view synthesis including object rotation up to 180 degrees
- Appearance editing: adding, removing, or altering specific elements while all other regions of the image stay unchanged
- Bilingual text editing in Chinese and English that preserves the source image's font, size, and style
- Dual-path conditioning (Qwen2.5-VL for semantics, VAE encoder for appearance) documented in the Qwen-Image technical report
- Built on the 20B Qwen-Image foundation, with weights for the underlying editing models released under Apache 2.0
- Later family releases added multi-image input (Edit-2509) and stronger identity consistency (Edit-2511)

## What people use it for

- Retouching product photos: swapping objects or removing clutter without disturbing the rest of the frame
- Correcting or replacing text on signs, posters, and packaging while matching the original typography
- Turning one character or mascot into new poses, outfits, and scenes for IP work
- Background replacement and scene restyling that keeps the subject intact
- Iterative cleanup passes, fixing one region at a time across several rounds

## Reported strengths

- State-of-the-art results on public editing benchmarks (GEdit, ImgEdit) at the family's launch
- Appearance mode is genuinely conservative: regions outside the edit stay pixel-stable rather than being regenerated
- Text edits are a standout, since few editing models preserve typography while rewriting words
- Open Apache 2.0 weights for the underlying models

## Reported limitations

- Complex fixes can take several rounds; Alibaba's own launch demo corrected a calligraphy artwork through a chain of step-by-step edits rather than one instruction
- Later releases in the line (Edit-2509, Edit-2511) list image-drift mitigation and character consistency among their fixes, so first-generation editing weights can drift on identity across edits
- The original editing release accepts a single input image; combining multiple source images only arrived with the 2509 update

## Sources

1. Qwen-Image-Edit: Image Editing with Higher Quality and Efficiency (Qwen (Alibaba)): https://qwenlm.github.io/blog/qwen-image-edit/
2. Qwen/Qwen-Image-Edit model card (Hugging Face (Qwen)): https://huggingface.co/Qwen/Qwen-Image-Edit
3. Qwen-Image Technical Report (arXiv): https://arxiv.org/abs/2508.02324
4. QwenLM/Qwen-Image repository (GitHub (QwenLM)): https://github.com/QwenLM/Qwen-Image

---

About Artifio.ai: 100+ AI image, video, and audio models in one workspace with a single shared wallet — no per-model subscriptions. This is the markdown rendition of https://artifio.ai/models/qwen-i2i for AI assistants and crawlers; the HTML page is canonical.