Create and Edit Pictures with Qwen Image 2.1

Leave the reference area empty and the model draws from your description alone. Drop in a few photos and the same prompt box becomes an editing instruction, so an idea that began as one sentence can later pull in a real product shot.

Combine a product photo, a logo, a background plate and a colour swatch in a single request. Each upload keeps its place in the queue, so you can say which picture supplies the subject and which only lends its light or palette.

The first upload is “the first image”, the next is “the second image”, and so on. A line like “put the bottle from the first image on the wooden table from the third image” is shorter than describing every photo again.

No need to translate an idea into English first. Spanish, Arabic, Korean or any other language is accepted, and the field holds up to 5,000 characters, enough for scene, mood, layout notes and the exact wording you want shown.

When a picture has to carry words, such as a poster headline or a label on a box, both Chinese and English work well. Put the text in quotation marks and say where it goes, so it reads as copy to print rather than scenery.

Use 1K for drafts and 2K when the image will be printed, cropped or shown large. Frames run from square to an extra-wide 21:9 and an extra-tall 9:21, and edit mode adds an auto setting that copies the first reference.

Upload a clean product photo as the first image and a room or landscape as the second, then ask for the product placed naturally in that scene. If the light feels flat, add a third reference and say it is there only for lighting.

A flyer with a Chinese headline over an English subtitle, or the reverse, fits into one prompt. Use 3:4 or 2:3 for print-style posters and 16:9 for screens, and move to 2K when the type must stay sharp at a large size.

Supply several photos of one person, pet or mascot and ask for a new pose or location. Stating which image shows the face and which shows the outfit helps the model keep track of the details you care about.

Gather up to ten pictures that capture a palette, a texture and an atmosphere, then ask for one scene that blends them. It is a quick way to show a client a visual direction before anyone books a shoot.

Before describing the result, say what each upload is for: “the first image is the product, the second is the background, the third is only for colour”. The model then follows a map instead of guessing which photo matters most.

For signs and posters, write the lettering inside quotation marks and add its position and size, for instance a large headline across the top with a smaller line beneath. Short copy gives every character a better chance of landing cleanly.

You can describe the whole scene in French, Turkish or Portuguese and still quote the exact English or Chinese slogan that should appear. Mixing languages inside one prompt is perfectly fine.

Run a few 1K versions while the composition is still moving, then switch to 2K for the take worth keeping. Since 2K takes noticeably longer, the extra wait goes only to the image that deserves it.

The 5,000-character limit is generous, but a fixed order keeps a long brief readable: subject, setting, light, style, then text and layout notes. A structured prompt is also much easier to edit on the next run.