Text to Image

Text to image turns a written description into a still picture. Here it is the natural first step of a video job: generate the frame you want, adjust it until it is right — which costs far less than re-rendering a clip — and only then animate it with the image-to-video route.

Prompt for a frame, not for a scene

A prompt that works for a single still is more specific about appearance and less about time. Name the subject, the setting, the lighting and the framing:

• Subject and setting — "a ceramic coffee cup on a wooden windowsill".
• Light — "low morning light from the left" changes far more than any style word.
• Framing — "close-up, shallow depth of field" is a direction the model can follow.

Because there is no motion to describe, avoid verbs that imply one. Anything that should not appear at all belongs in the negative prompt.

Working between stills and clips

Generating stills first is the cheapest way to iterate, and it is the workflow this workspace is built around:

1. Generate the frame here and refine it until the composition is right.
2. Take that image into image to video and describe only the motion you want.
3. Come back and generate a new frame when you need a different camera angle or a new scene.

Settings that change the result

• Model route — the routes listed in the form are the ones reachable right now; live controls are the source of truth, not this page.
• Aspect ratio and resolution — choose these to match where the image will be published. Both affect the credit cost shown beside the form.
• Seed — fix it to re-run the same prompt with one setting changed; leave it empty for a new result.
• Negative prompt — for the artefacts you want suppressed, such as text, watermarks or extra limbs.

What it is good at, and what it is not

It is good at subjects, textures, lighting and composition — product shots, portraits, backgrounds, concept frames. It is unreliable for legible text inside the image, for exact real people, and for replicating a specific copyrighted character. Read the licence of the route you used before publishing anything.

Treat output as a starting point you may need to regenerate: quality varies by route and by prompt, and the first frame is rarely the best one.

Questions people ask

Can I use the image in a video? Yes — that is the intended path. Generate the still, then take it into image to video.

How much does an image cost? The credit cost is shown beside the form before you submit and depends on the model, resolution and aspect ratio. Credit prices are on pricing.

Where do I start? Open the Wan workspace and choose the text-to-image tab.