Batches, edits and sizes from a single model.
Ask for up to six images per run and compare directions side by side instead of rerunning one prompt again and again. One or two images fit the free tier; larger batches need a Basic plan.
Switch to Image Edit, upload up to ten source images and describe the change. The model rewrites what you ask for and keeps the rest close to the source, which makes it useful for backgrounds, outfits and lighting.
Choose from nine frame presets and 1K, 2K or 4K. 1K and 2K are open on the free tier; 4K needs a Basic plan.
Seedream 4.0 is ByteDance's image model that generates and edits in one architecture. On this page the Seedream workspace opens on version 4.5, so start by selecting 4.0 in the version bar. Two modes appear: Text to Image builds a picture from your prompt, and Image Edit changes one to ten source images according to your prompt, which is required in both modes. Choose a frame preset, a resolution of 1K, 2K or 4K, and how many images you want, from one to six. Results are saved to My Creations. Newer versions 4.5, 5.0 Lite and 5.0 Pro sit in the same bar if you want to compare. A practical rhythm is to explore at 1K with a batch of four, pick the strongest direction, then rerun that prompt alone at 2K or 4K for the final file.
Each tab is a separate mode or control.
Describe the scene and the model renders it, from product shots and room designs to illustrated characters. Put the subject first, then setting, light and style, so the most important details lead the prompt.
Long Prompts
Room for detailed briefs.
Photo or Illustration
Set the style in words.
Quick Reruns
Adjust and generate again.
Six capabilities of the model.
Built by ByteDance's Seed research team.
Two modes in one model.
Compare directions at once.
Merge several photos in an edit.
Three resolution steps.
4.5 and 5.0 in the same version bar.
Three groups, two workflows each.
Describe the room once and ask for four images in one run. Clients pick a direction from the grid instead of waiting days for revisions, and you rerun only the chosen one at a higher resolution.
Upload the client's room photo in Image Edit and describe new furniture or paint. Check proportions before presenting.
Upload a garment photo and describe models, poses and settings. A single batch returns several looks for the same item, ready for a lookbook draft or a product page carousel.
Use Image Edit to replace a cluttered background with a plain one, keeping the garment as close to the source as possible.
Describe a character and a setting, ask for six images, and use the set as a rough storyboard for a video or comic. Reorder the frames, then refine the ones that carry the story.
Upload your character image as a source and change only the pose or place, so the series stays recognisable.
Three actions on this page, from version choice to download.
Pick 4.0 in the version bar, then Text to Image or Image Edit.
Write the prompt, add sources if editing, choose size and image count.
Review the set and download favourites from My Creations.
Cost, versions and limits.
Each image costs 5 credits at 1K or 2K and 15 at 4K, multiplied by the number of images in the run. New accounts get 10 credits once, valid for 30 days.
Yes. Seedream comes from ByteDance's Seed team. This site is an independent online studio that runs the model.
Click 4.0 in the version bar above the prompt box. The modes and settings below update to the 4.0 options, including the image-count control.
One to six per run. One or two are open on the free tier; three to six need a Basic plan or above.
Text to Image starts from your prompt alone. Image Edit needs at least one source image, up to ten, and changes it according to your prompt.
On this site it runs online in your browser, so there is nothing to download or install, and results are saved to My Creations.
ByteDance's Seedream 4.0 creates and edits images in one model, with up to six results per run and 1K, 2K or 4K output. The page opens on Seedream 4.5, so pick 4.0 in the version bar first.
0 / 3000
