Seedream 5.0 Flash is the still-image model on this page. A prompt can become a scene, a product still, or a poster that carries the words you supply. Image to Image takes up to 10 reference files and changes the frame you already have. Output is 1K, 1.5K, or 2K, as JPEG or PNG. This page does not offer 4K, and it does not split a picture into separate layers.
Text or up to 10 references · 1K, 1.5K, 2K · JPEG or PNG
Open the generator
SEEDREAM 5.0 FLASH
Text to Image and Image to Image share the ratio, size, and file-type controls. One starts from a prompt. The other starts from files you upload.
01A prompt can set the subject, the light, and the crop. 1K is the smaller file. 1.5K sits between that and 2K. 2K is the largest size on this page. Use it for a product still, a room, or a poster layout when the frame will be viewed up close. The page has no 4K control.
Set a size
02Switch to Image to Image and upload the frame that should stay. Name the change in the prompt: a different jacket color, a cleared background, or a new object in one area. Say what must remain, such as the chair, the crop, or the face. This page has no click-to-select tool. The region is whatever you describe.
Edit a still
03Image to Image accepts 1 to 10 files, JPEG, PNG, or WebP, 30MB each. Point the prompt at each file: which one is the product, which one is the cloth, which one is only a color. Extra files do not replace a clear instruction. Text to Image does not use uploads.
Add references
04Put the exact wording in quotes and say where it sits: a headline, a label, a short line on a sign. Rendering covers a dozen widely used languages, including Chinese, English, Japanese, Korean, French, German, Spanish, and Arabic, and it follows reading direction, including right to left. Check names, prices, and accents in the finished still. This page does not look up live news or translate a website for you.
Add a line of type
05The ratio row is 1:1, 4:3, 3:4, 16:9, 9:16, 2:3, 3:2, and 21:9. There is no auto ratio and no 4:1 or 8:1. PNG and JPEG are the two file types. Choose PNG when you need a flatter graphic. Choose JPEG when the still is a photograph and file size matters.
Choose a ratio
The controls sit in the generator at the top of this page. The three steps below are the same ones on that form.
Text to Image starts from the prompt alone. Image to Image asks for at least one upload. Switching modes reloads the form, so set the ratio and size again after the switch.
Describe the subject, the light, and any words that must appear. In Image to Image, upload up to 10 JPEG, PNG, or WebP files, 30MB each, and say what each file is for.
Choose one of the eight ratios, then 1K, 1.5K, or 2K, then PNG or JPEG. Those three rows are the output controls on this page.
| Criterion | Seedream 5.0 Flash | Nano Banana Pro | Nano Banana 2 | GPT Image 2.5 |
|---|---|---|---|---|
| Job | Prompt stills and edits | Diagrams and locked faces | Drafts and ultra-wide crops | Multi-round reference edits |
| In-image type | Headlines, many languages | Headlines and longer copy | Short headlines and labels | Posters and mixed type |
| References | Up to 10 stills | 14 stills, 5 people | Up to 14 stills | Up to 16 stills |
| Edits | Prompt on uploaded stills | Light, focus, one region | Follow-up on a draft | Flare speed, Sunburst control |
| Output | 1K, 1.5K, 2K; JPEG, PNG | 1K, 2K, 4K | 1K–4K, plus 4:1 and 8:1 | 1K, 2K, 4K |
Labeled diagrams, longer in-image copy, and up to five people held across 14 references, at 1K to 4K.
Faster drafts and very wide crops, including 4:1 and 8:1, from 1K to 4K.
Reference edits across rounds. Flare is the faster pass. Sunburst holds a tighter edit.
What it makes, how edits work, and which controls this page shows.
Seedream 5.0 Flash is the image model on this page. It turns a prompt into a still, or changes a still you upload. The result is an image, not a video. Sizes on this page are 1K, 1.5K, and 2K.
Prompted scenes and product stills, edits that follow a written instruction, up to 10 reference images in one pass, and short text rendered in the frame. A dozen widely used languages can be set in the prompt, including Chinese, English, Japanese, Korean, French, German, Spanish, and Arabic.
Yes. Choose Image to Image and upload 1 to 10 JPEG, PNG, or WebP files, 30MB each. Describe the change and what should stay. This page does not offer a point, box, or lasso control. If a region matters, name it in the prompt.
Yes. Quote the exact words and name the place: headline, label, or a short sign. It can also follow the reading direction of the language, including Arabic right to left. Read the finished letters yourself before you print a name or a price.
Use the generator at the top. Pick Text to Image or Image to Image, write the prompt, then set one of the eight ratios, a size of 1K, 1.5K, or 2K, and PNG or JPEG. Image to Image needs at least one upload.
This page has two modes: Text to Image and Image to Image. Text to Image uses the prompt only. Image to Image uses the prompt plus up to 10 references. There is no layer-split mode and no 4K size here.
The still tools and the video tools share one account. Set the prompt and the file options in the generator on this page. When a frame needs motion, download it and continue in the video generator.
Open the generator
