ByteDance's design-focused image model. Give it a written brief, a rough sketch or up to 10 reference images and get posters, page layouts and product shots with the text you specify, in more than ten languages. Generate and edit here, with the credit price shown before you start.
Brief to design
Write what the design must contain, draw where it goes, or hand over the pieces, and Seedream 5 Pro plans the layout and fills it in. Each example starts from one kind of brief.
Written brief
List the format, what goes in each area, every headline and figure in quotes, and the colors; Seedream 5 Pro decides the hierarchy and arranges the page. Long body copy comes out as placeholder lines, so set real paragraphs in a layout tool and check every number before printing.
See the full brief
Sketch
Upload the sketch as a reference and say what each box holds, such as the logo name, the exact headline and the button text; handwritten notes on the sketch are read too. The sketch steers the layout but is not a mask, so check sizes and spacing, which can shift a little.
Reference images
Give each reference one job, such as a pattern, an object or a location, and say where each one applies so a pattern lands only where you want it. You can add up to 10 reference images per request; the credit quote updates as you add them.
Strengths
Beyond layouts, ByteDance highlights text in many languages, edits guided by marks on the image, photographic realism and exact color control.
Besides Chinese and English, ByteDance lists French, German, Russian, Japanese, Korean, Spanish, Arabic and more for prompts and in-image text. Put each word in quotes and name its language; short words on signs and packaging are the most reliable.
Draw boxes, circles and short notes on your image before uploading it, then repeat the request in the prompt. Seedream 5 Pro reads the marks as positions: the mirror goes on the boxed wall, the circled box disappears and the rug changes. Oeneo AI has no mask brush yet, so the marks do that job.
Describe the shot the way a photographer would, such as panning, 85mm, wide open or window light, and the model follows the technique, not just the subject. In this panning shot the cyclist stays sharp while the seaside road streaks behind him.
Give a hex code or upload a color swatch, and name the material. The radio keeps its shape, grille, brass dial and strap while the cream plastic becomes glossy cobalt lacquer matching #2450D6. Check the color on your own screen before ordering samples.
Say what the image is for, what goes where and the exact words. ByteDance recommends staying under 600 English words or 300 Chinese characters; longer prompts start dropping details.
Purpose and format first
Open with what it is and where it runs: "two-page magazine spread", "16:9 landing-page hero", "4:5 social poster".
Every word in quotes, with its language
Write the headline, labels and button text exactly as they should appear, and say which language each one is in.
A job for each reference image
"Image 1 is the pattern only, image 2 is the product, image 3 is the location." Without roles, references blend into each other.
The camera, for photographic parts
Name the lens, light and small imperfections. References steer the style strongly: renders as references give render-looking results until you describe a real camera.
Numbers you haven't checked
The model arranges data neatly but doesn't verify it. Supply the exact figures and labels yourself.
Paragraphs of in-image text
Long paragraphs inside the image go wrong more often than headlines. Keep on-image text short and add body copy later.
Ten edits in one request
A few small changes work together; a long list drifts. Make the most important change first, then edit the result again.
Example brief for a spread like the one above: Two-page magazine spread, 16:9. Left page: full-bleed photo of a laughing beekeeper holding up a honeycomb frame in a sunny community garden. Right page: headline "CITY HONEY", subhead "Rooftop hives, one summer", three stat callouts "42 HIVES", "6 ROOFTOPS" and "310 KG", and a small honey-amber bar chart. Cream paper, black serif type, generous margins, body text as placeholder lines.
Seedream 5 Pro has one price per size and adds a little for each extra reference; GPT Image 2's price climbs with its quality setting. To compare the Nano Banana models with each other, see the Nano Banana 2 page.
| Setting | Seedream 5 Pro | Nano Banana Pro | GPT Image 2 |
|---|---|---|---|
| Developer · released | ByteDanceJuly 8, 2026 | GoogleNovember 20, 2025 | OpenAIApril 21, 2026 |
| Credits per image | 1K: 14 · 2K: 27 | 1K: 30 · 2K: 30 · 4K: 60 | 1K: 2–60 · 2K: 2–60 · 4K: 3–90Low to high quality |
| Reference images | Up to 10 | Up to 14 | Up to 16 |
| Each extra reference | +1 after the first | No extra charge | +2 after the first |
| Highest resolution | 2K | 4K | 4K |
| Output formats | PNG, JPEG | PNG, JPEG, WEBP | PNG, JPEG, WEBP |
| Pick it for | Layouts, multilingual text, mark-up edits | Five-person shots, logos on productsContinuing existing Pro work | Low-cost drafts, up to 16 references |
Each model has its own page and runs in the same workbench on one credit balance.
ByteDance8 credits an image with no charge for reference images. Its page compares it with Seedream 5 Pro.
View model
New
GoogleConsistent characters, product placement, exact short text
View model
New
Black Forest LabsComposed scenes from a prompt or references
View model
New
IdeogramRefine image details with source-guided edits and readable text.
View modelRemove the background, upscale, vectorize or denoise any result, all on the same credit balance.
Transparent PNG or WebP with clean hair and product edges, ready for a listing or a new composition.
2× / 4× / 8× with recovered texture instead of smeared pixels, so a 1K draft can become a print.
Logos, icons and flat illustrations become editable SVG that scales to any size.
Remove grain and compression artifacts from low-light photos and scans while keeping real detail.
Seedream 5 Pro (Seedream 5.0 Pro) is ByteDance Seed's image generation and editing model, released on July 8, 2026. ByteDance highlights four strengths: dense layouts and infographics, edits guided by marks or sketches on the image, realistic light and skin texture, and multilingual text. On Oeneo AI you can generate from a prompt or edit with reference images.
No. They are separate models from the same generation. Lite came out first and is built for fast, large outputs; Flash is tuned for speed and cost; Pro focuses on design and precise editing. Oeneo AI offers Seedream 5 Pro and Seedream 4.5, not Lite or Flash.
It isn't free. Each image costs 1K: 14 · 2K: 27 credits, a little more when you edit with several reference images, and the workbench shows the exact total before you generate. New accounts get 10 free credits: not enough for one Seedream 5 Pro image, but enough to try Seedream 4.5 at 8 credits an image.
Choose Seedream 4.5 to spend less, 8 credits an image with no charge for reference images, or to move a person or product into a new scene while keeping its details. Choose Seedream 5 Pro for design work: layouts with the text you specify, multilingual signs and mark-up edits. The Seedream 4.5 page compares the two side by side.
No. ByteDance offers this model up to 2K, and Oeneo AI offers 1K, 2K. For a bigger file, generate at the largest size and enlarge it with the Image Upscaler.
Chinese and English plus more than ten others listed by ByteDance, including French, German, Russian, Japanese, Korean, Spanish and Arabic, with right-to-left layout for Arabic. Short headlines and labels are the most reliable; proofread small type and long lines in every language.
Not on Oeneo AI. To change one area, draw a box or a note on the image before uploading it and describe the change in the prompt. For a transparent background, run the result through the Background Remover.
Upload up to 10 and give each one a job in the prompt, for example "image 1 is the pattern, image 2 is the product". References steer the style strongly: if they are renders, results look rendered until you describe a real camera and light.
Only if you supply them. Seedream 5 Pro arranges charts, labels and callouts well but doesn't check facts, and ByteDance notes that fine text still has room to improve. Put the exact figures and labels in the prompt and verify them before publishing.
A request can be stopped by the provider's safety check or fail when the service is busy. Failed tasks return their credits automatically. Some harmless wording still trips the safety check, so describe the image differently and try again.
Describe the design, upload what you have, and keep editing the result until it's ready.
