Qwen-Image-Edit
Alibaba's open-weight 20B image editor, self-hosted by AutorunX as AX-QWN, preserves identity through fine-grained edits at near-zero marginal cost.
What Qwen-Image-Edit does
Qwen-Image-Edit takes an existing photo and a text instruction and produces a modified version that keeps everything the instruction didn't ask to change — a face, a pose, a background — intact. It handles fine-grained object edits, style transfer, and semantic transformations like changing a scene's viewpoint.
On AutorunX it's positioned as the budget-conscious alternative in services where GPT Image 2 or Nano Banana are the polished default: it's selectable in the model picker for Face / Identity persona shots and Banner Ad edits when identity-preserving editing at a lower credit cost is the priority.
Capabilities
Key features
- Identity-preserving edits
- Modifies a specific instructed element of a photo — an object, a pose, a background — while keeping the rest, including faces, structurally consistent with the source.
- Bilingual, accurate text rendering
- Handles both English and Chinese text rendering with high fidelity, a differentiator versus many Western-trained editing models.
- Dual-track access
- Available as AX-QWN, AutorunX's self-hosted open-weight deployment, and separately as Qwen-Image-Edit-Plus, a proprietary hosted tier for reference-locked exact-identity edits.
- Semantic transformations
- Goes beyond pixel edits to handle style transfer and viewpoint changes, driven by the Qwen2.5-VL component's understanding of the source image's content.
- Open-weight cost efficiency
- Because AX-QWN runs on AutorunX's own GPU fleet under an Apache 2.0 license, it's one of the lowest credit-cost image models on the platform.
How Qwen-Image-Edit works
Qwen-Image-Edit pairs a 20-billion-parameter multimodal diffusion transformer (MMDiT) with a Qwen2.5-VL vision-language model for text conditioning. The VL component reads and reasons about the source image before the diffusion transformer edits it, which is what lets the model make fine-grained, semantically aware edits — changing one object, transferring a style, or preserving a face — instead of regenerating the whole scene.
AutorunX runs it on two tracks. The owned track, internally called AX-QWN, is the Apache 2.0-licensed open-weight model self-hosted on AutorunX's own GPU fleet — no per-call fee to a third party, so it's one of the cheapest models on the platform. A separate proprietary hosted tier, Qwen-Image-Edit-Plus, is accessed through a different provider's API rather than self-hosted, and requires a reference image to lock in exact-identity edits.
Both tracks share the same core strength: because the weights are Apache 2.0-licensed and openly published (unlike closed competitors), the underlying model can be inspected, fine-tuned, and deployed independently of any single vendor's API — which is exactly why AutorunX was able to bring it in-house as AX-QWN instead of only ever renting it.
What people use Qwen-Image-Edit for
Face / Identity persona touch-ups
Edit a generated persona photo — outfit, background, pose — while keeping the character's face and identity intact.
Budget banner ad edits
Make targeted edits to banner ad creative at a lower credit cost than the panel's default model.
Product photo corrections
Fix a specific element of a product shot — remove a background object, adjust a color — without regenerating the whole image.
Style transfer experiments
Apply a different visual style to an existing image while keeping its composition and subject recognizable.
How to use Qwen-Image-Edit on AutorunX
Qwen-Image-Edit is available in Character, in the Image module.
- 01
Open Face / Identity or Banner Ad in Image
Both services support Qwen-Image-Edit as a selectable alternative to their default model.
- 02
Set up your input
Upload the photo you want edited and write your edit instruction, or add a reference image if you need the Plus tier's exact-identity lock.
- 03
Pick Qwen-Image-Edit in the model picker
Open the model picker and switch from the panel's default (GPT Image 2 on Face / Identity) to Qwen-Image-Edit.
- 04
Generate
Run the edit and review that the untouched parts of the image stayed consistent with the source.
- Billed per image from your shared AutorunX credit wallet.
- 20 credits / image (AX-QWN owned, and Qwen-Image-Edit-Plus hosted tier)
Tips for better results
- Be specific about what to preserve
- State plainly what should stay unchanged ("keep the face and pose identical, only change the background") — the VL conditioning follows explicit preservation instructions well.
- Use Qwen-Image-Edit-Plus for exact-identity locks
- When a job absolutely needs a reference image to pin identity, reach for the hosted Plus tier rather than the owned AX-QWN model.
- One edit instruction at a time
- The model handles a single, clearly stated edit more reliably than a long compound instruction — chain edits across multiple generations if you need several changes.
- Try it as the cheaper alternative first
- On Face / Identity and Banner Ad, switch to Qwen before defaulting to a pricier model if the edit is simple and cost matters.
Qwen-Image-Edit — frequently asked
Related models
FLUX.2 Klein
AutorunX ownedBlack Forest Labs' compact, Apache 2.0 FLUX model, self-hosted by AutorunX as AX-KLN, delivers sub-second exact-identity generation at the platform's lowest cost.
Z-Image Turbo
Alibaba Tongyi LabAlibaba Tongyi Lab's 6B open-weight model generates photorealistic images in as few as 8 steps, making it AutorunX's cheapest image option.
Nano Banana
Google DeepMindGoogle's Gemini-family image model, nicknamed Nano Banana for its viral photorealistic edits, generates and edits images up to a 4K-capable tier.
Ready to create with Qwen-Image-Edit?
Open Character and generate your first image in a couple of minutes.
Free plan, no card required.