Use clear references
A blurry or partly hidden face gives the model little to hold on to. Choose evenly lit images in which the subject fills a good share of the frame.
4CREDITS
Remaining credits：
Public
BestAIVideo
Generate and Edit Images with Qwen Image 2.1
Qwen Image 2.1 is Alibaba's unified model for generating images and editing the ones you already have. Describe a picture in words, or upload up to 10 reference images and tell it what to change, choose 1K or 2K, and see the credit cost before you submit.
What Qwen Image 2.1 Brings to Your Workflow
Qwen describes the model as compact but capable, with improved typography and stronger identity retention in edits. This is what the generator exposes.
One model for making and editing
Use text mode for a new image, or edit mode with one to ten references to change a picture you already have.
Identity that survives edits
Qwen says people and products remain recognizable when you edit them. Choose sharp, front-facing references for the best chance of a match.
Nine ratios, including 21:9 and 9:21
Choose 1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9 or 9:21. Edit mode adds an auto option that follows the shape of your reference.
Two sizes at low prices
1K costs 4 credits and 2K costs 7, for text and edit alike. The reference count never changes the price.
Prompting Qwen Image 2.1 Step by Step
Qwen's examples use short, plain instructions, and quoted text for anything that must appear in the picture. Build from that pattern and add detail only where it helps.
Write the picture in one sentence
Name the subject, the setting and the mood, for example a neon shop sign on a rainy night with wet pavement. Add detail after you see the first result.
Quote the words and describe the type
Qwen highlights typography that considers type style and layout. Put the exact text in quotes and describe the style of lettering, not just the words.
Issue edits as instructions
For editing, write what you want changed, such as change the background to a sunset beach, and say what must remain unchanged.
Tell it where each reference belongs
With several images, say which one supplies the character, which the background and which the style, so the model does not have to guess.
Three Qwen Image 2.1 Examples Worth Reading Closely
A research-paper style architecture figure
Diagram blocks, bar charts, a table and a caption share a single dense page. It is a stress test for small labels and layout, and the right place to check every character after generation.
A sticker of a dragon riding a lion dance costume
A cartoon dragon in a red outfit rides a festive lion with lanterns and confetti on a clean white background. Prompts for sticker art work best when they ask for bold outlines and a plain backdrop.
A travel app home screen
A phone screen with a greeting, a hero card for Santorini, four planning icons and trip lists is composed in a single portrait image. It shows how interface mockups can be drafted in one prompt.
Qwen Image 2.1 Production Notes
Match the ratio to the job
Use 9:21 for tall story formats and 21:9 for cinematic headers. In edit mode, auto keeps the shape of the first reference.
Read every character
Dense figures and small captions are where generated text most often slips. Zoom to full size before you place an image in a document.
Projects That Suit Qwen Image 2.1
Product and e-commerce edits
Upload several shots of an item and place it in new scenes while its shape and colour stay recognizable.
Posters and signage
Generate a poster with a quoted headline, then switch to 9:21 or 21:9 for a vertical or panoramic variant.
Sticker and mascot sets
Create clean-backdrop character art for chat stickers, packaging and event materials.
Interface mockups
Rough out app screens and landing pages before a design team builds them.
Portrait retouching
Edit a face photo with a light touch, using the original as the reference to keep identity.
Scene compositing
Combine pieces from up to ten uploads into one composition and say which element each file supplies.
Qwen Image 2.1 Questions, Answered
Qwen Image 2.1 is a model from Alibaba's Qwen team released in September 2026. It combines text-to-image generation and image editing in one model.
Up to 10 in edit mode, with at least one required. Text mode takes none.
Sizes are 1K and 2K. Text mode offers 1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9 and 9:21, and edit mode adds auto as its default.
A 1K image is 4 credits and a 2K image is 7. Text and edit modes are priced the same, and the generator shows the amount before you submit.
The prompt can run to 5,000 characters, enough for a detailed scene plus several quoted lines of text.
Use a plain instruction, like change the background to a sunset beach, and add what should stay the same. Keep one change per request when you can.
Qwen highlights improved typography that accounts for type style and layout. Put exact words in quotes, describe the lettering and proofread small text.
Qwen describes native transparent output for the model, but this generator does not include a transparency control, so ask for a plain backdrop and cut it out yourself.
Qwen Image 2.1 is a separate, newer model with its own prices and the 1K and 2K sizes. The earlier Qwen entries keep their own settings and pricing.
At 4 and 7 credits, Qwen Image 2.1 sits between Seedream 5.0 Flash at 3 and GPT Image 2.5 at 5, 8 and 13.
Create or Edit with Qwen Image 2.1
Write the prompt, add references if you are editing and see the credit cost before you submit.