Qwen Image 2.1 is a 7B parameter vision model developed by Alibaba that handles both image generation and complex editing tasks. Developers and designers use it to create assets with built-in transparency or to modify existing photos using multiple reference points. It solves the problem of having to chain separate models together—like a generator plus a background remover—by performing these operations in a single pass.
The model works through the reAPI platform, allowing for pay-per-image usage without managing local hardware. It supports high-resolution outputs at 1K and 2K, making it suitable for professional design workflows. Users can provide up to ten reference images to guide the generation, which is significantly more than standard image-to-image models typically allow.
Key Features
- Native transparency support for generating PNG and WebP files with alpha channels
- Multi-reference composition using up to 10 input images in one request
- Mask-based local editing for targeted changes within a specific region
- High-resolution generation options at 1K and 2K resolution tiers
- 7B parameter architecture for balanced speed and visual quality
- Open weights released under the Qwen Research License for self-hosting
- Automated moderation for inputs and outputs to ensure safety
- Credit-based pricing starting at $0.02 per generated image
How to Use
- Create an account on reAPI and generate a personal API key from the dashboard.
- Select the Qwen Image 2.1 model in the playground to test prompts and settings.
- Choose your desired output format, such as a transparent PNG or a standard JPEG.
- Upload up to ten reference images if you want to perform a multi-image edit.
- Send a POST request to the images endpoint with your JSON request body and API key.
- Poll the task ID provided in the response until the final image URL is returned.
Use Cases
- Generating product stickers with a clean alpha channel for e-commerce sites
- Creating a group photo by combining several individual portraits into one scene
- Performing a virtual try-on by merging a model photo with specific clothing items
- Swapping a specific object in a room while keeping the furniture and lighting the same
- Extracting a subject from a standard photo and converting it into a transparent UI asset
Pricing
Pricing is credit-based at $0.001 per credit. A 1K resolution image costs $0.02 (20 credits), while a 2K image costs $0.04 (40 credits). Check the official website for current pricing updates.
FAQ
What is Qwen Image 2.1?
It is a unified 7B visual model from Alibaba that creates images from text and edits them using reference images or masks.
Is Qwen Image 2.1 free?
No, the reAPI version is a paid service billed per image, though new accounts usually receive trial credits.
Does it support transparent backgrounds?
Yes, the model writes a real alpha channel into the file, producing native transparent PNG or WebP assets.
How many reference images can I use?
You can include up to 10 reference images in a single request to guide the final composition.
What resolution does the model support?
It supports two main tiers: 1K resolution for standard tasks and 2K resolution for higher detail.
Can I edit just one part of an image?
Yes, by providing a black-and-white mask, the model only changes the white area and leaves the rest untouched.
How does the pricing work?
You pay per completed generation with no monthly subscription fees, and failed tasks are refunded automatically.
