Turn a written scene into one image with Seedream 5.5. Describe what the viewer should see, pick the shape of the frame, and generate for 3 credits. Try the product shot below if you want a starting point you can change.
Seedream 5.5 is ByteDance's text-to-image model, sometimes written Seedream5.5. Start with a sentence you could ask someone to photograph. “A bookbinder pulls red thread through an open spine” gives the picture a subject, a gesture, and a detail worth looking at.
Give the subject something specific to do. A cyclist waiting at a crossing makes a different picture from one leaning into a turn. Include the detail that lets a viewer tell the two moments apart, such as a foot resting on the curb.
Decide what fits around the subject. In a close view of the bookbinder's hands, you might see the thread passing through the paper. A wider view can show the entire bench, with a stack of finished books at one end.
Describe the surface and the direction of the light together. A narrow reflection down the side of a glass bottle suggests a tall window. For an unglazed pitcher, ask for a soft side shadow that brings out the grain of the clay.
Keep a copy of your first prompt. If the important detail is too small, change the viewing distance before adjusting the palette. Each new generation can vary, but keeping the other instructions gives you a clearer basis for choosing the next image.
Use these four prompts as starting points in Seedream 5.5. Each one solves a different framing problem. Replace the object or setting with your own, then choose the aspect ratio in the generator so the canvas and the description agree.

Seedream 5.5 lets you choose the frame and file format before generating. If you already have a destination in mind, check the shape of its image slot. A cover, a slide, and a thumbnail may all need different space around the same subject.
Choose 1:1 for a square, 4:3 or 3:4 for a moderately wide or tall frame, and 16:9 or 9:16 for a slide or a vertical story. Consider how much room the subject needs on either side before picking the tallest or widest option.
JPEG is a practical choice for sharing a photographic scene. Choose PNG if your next step involves placing the image in a layout and you want lossless compression. A PNG can contain an opaque background, so check the actual result before placing it over a colored layer.
Each run returns a single image. Look at it at the size your audience will see. A reflection that works in a large product shot may look like a distracting white spot in a small thumbnail. Adjust that detail for the next run.
One Seedream 5.5 generation costs 3 credits across the five ratios and both file formats. If you plan to try a square composition and then generate a separate vertical composition, allow 6 credits for those two runs.
These answers cover the choices you make between entering your first prompt and saving an image for a layout.
Sign in to Img2Img AI and use the generator on this page. Enter a short description of the scene, choose the aspect ratio and file format, then generate the image. The preview lets you inspect the result before you download it.
Each image costs 3 credits. Switching from JPEG to PNG before generating keeps the same price. If you decide to try a second prompt after seeing the first result, that second generation is a separate run and costs another 3 credits.
Tell Seedream 5.5 what the object is made of, how it sits, and where the light comes from. For example, describe a brushed steel kettle seen from handle height, lit by a kitchen window. Add one highlight or shadow you want the viewer to notice.
Yes. Choose Image to Image and upload up to 10 reference photos, then describe what should change. One reference costs 5 credits. Text to Image, with no photo, still costs 3 credits.
Specify both the subject's position and what belongs behind the future headline. For example, put a cyclist in the lower third and ask for plain pale sky across the top. Add the lettering to the downloaded image so you can adjust line breaks at the final size.
Find more examples in our image generation guides.