Qwen Image: lock this still before you buy motion (4K, 2cr)

Qwen Image is the stills row. Animate after approval. Do not generate the pack inside the video model.

Qwen, 2 credits, text-to-image, max output 4K. Description: "High-quality realistic images with excellent detail." requires_image is false. Aspects: 16:9, 3:2, 4:3, 1:1, 3:4, 2:3, 9:16. Styles: Realistic, Artistic, Anime. Qualities: SD, HD, 4K. Max reference: 1. The catalog marks the row discounted (1 credit on the discounted step). audio does not apply. This is a detail still at 4K for 2 credits. It is not a Wan video clip. It is not Qwen Image Edit.

4K realistic detail is the approval test

The description is "excellent detail." Use that. Zoom the still. If the product edge, fabric, or face falls apart at 4K, it will fall apart as a pause frame in a video. Re-roll here. 2 credits (1 on the discounted step). Do not "see how it looks moving." Motion hides detail for one loop and then every scrub shows it.

Seven aspects, including 9:16 and 16:9. Set the shipping ratio on this row. Qwen's stills register is the generic vivid-description checklist: subject, setting, lighting, style, mood. Hit those. Do not write a tracking shot.

Text to image is the door. Qwen's roster is the family, including edit SKUs. Best AI image generator is the ranked shelf. Alibaba's video work is Wan, a different roster.

One reference slot, not a video input

max_reference: 1. You may pin one still. You may also generate from text alone. That slot is not a driving clip, not a last frame, not five style boards. If the job is "make it look like this photo, then we animate," one reference plus a 4K generate is the honest path. If the job is already an edit of a locked still, Qwen Image Edit is the edit SKU. Re-rolling Qwen Image is a new still.

Because this is text-to-image with no required image, it will invent. That is useful when the pack does not exist. It is destructive when a still already exists and you only needed a colour change. Know which conversation you are in.

Do not generate the pack inside the video model

Wan, Kling, H3, Veo will all give you a frame. It will not be this 4K still, at 2 credits, with the aspect you set on purpose. Pulling a frame from a video generate to use as a hero is how packs lose detail and gain random identity. Sequence:

1. Still here. 4K. Ratio from the list. Optional one reference.
2. Human yes.
3. Image-to-video on a motion row.

Qwen Image 3 exists as a later stills SKU on the roster. This page is the qwen-image record: 2 credits, 4K, seven aspects, one reference, realistic detail. Use the row you opened.

FAQ

Is Qwen Image 4K?

Yes. Max output is listed as 4K, with SD, HD, and 4K in qualities. 2 credits is the catalog figure (discounted step 1). Generate at the quality you will actually review.

Does it make video?

No. Image content type. No durations. Animate after the still survives a zoom.

Can I attach reference images?

One. max_reference is 1, and a reference is not required. More than one still, or a driving video, is a different model.

When should I use Qwen Image Edit instead?

When the still is already locked and the job is a change. This row generates. The edit SKU edits. Generating a "new same image" is how identity drifts before anyone buys motion.