Seedream 5.5 Generator

Make lifelike pictures from a prompt, transparent PNGs ready for your designs, and edits from up to 10 of your images that change only what you ask, with ByteDance’s newest image model. Online in your browser, from 1 credit a picture.

Make a Seedream AI image online

Model

Prompt

0/5,000

Format

Aspect ratio

Size

Background

2K · 1:1 · No watermark

A surreal scene made with ByteDance’s Seedream model: a chrome DREAMSHOT camera beams light over waves, clouds, strawberries and hot-air balloons, with a crescent moon carrying a small houseExample

What is Seedream 5.5?

Seedream is ByteDance Seed’s image model family, and version 5.5, released on November 9, 2026, is its newest. One model does two jobs: text to image from a prompt, and image edit from 1 to 10 pictures you upload. On Melies AI it runs in your browser, with no app, GPU or API key.

ByteDance names three upgrades in the 5.5 release. Multi-turn editing is more stable: parts of a picture you do not ask to change stay the same at the pixel level, with no colour shift, and ByteDance shows one portrait edited 100 times in a row. The model draws transparent and semi-transparent backgrounds itself, so cut-outs need no background removal. Light, materials and skin look more natural, with less gloss and overexposure, and dense small text in English or Chinese reads more clearly.

ByteDance tested the 5.5 release with expert raters on its MagicArena platform in six fields: e-commerce, marketing, film and TV, gaming, office work and fashion design. Against Seedream 5.0 Pro, quality and usability rose in all six, and usability more than doubled in film and TV and in office work. Severe text errors fell to 0.4% of images and minor ones to 1.7%. These are ByteDance’s own tests, not independent ones.

On this site you pick one of 8 shapes (1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9; an edit can also use Auto, where the model decides), and 1K, 1.5K, 2K; a 2K square is 2048 × 2048 pixels. Prompts run from 5 to 5,000 characters. Save as PNG or JPEG, with an automatic, transparent (PNG only) or solid background. There is no visible watermark.

Edit rounds in ByteDance’s demo
100
Reference images
10
Aspect ratios
8
Largest size
2K
Prompt characters
5,000
Credit · one 1.5K image
1

By ByteDance Seed · run through a third-party API · not affiliated with ByteDance · Last updated

What ByteDance’s new image model does well

Five things the 5.5 release is built for, from edits you can repeat to cut-outs with no background.

An animation of one study edited round after round: the furniture and decor change step by step while the walls, window and daylight stay the same
01

Edit the same picture again and again

Each edit changes only what you name: a lamp, a ring, a colour, a line of text. The rest of the picture stays the same, down to the pixel, so you can keep refining one image. Download a result and add it back as @Image1 for the next change.

Gadgets cut out on a transparent background, shown on a checkerboard: headphones, a speaker, a photo printer, a beanie, a radio and a charging pad, each with clean edges
02

Transparent PNGs, no background removal

Pick a transparent background and the model draws the subject on an alpha channel, including semi-transparent areas and fine edges like feathers and claws. Stickers, logos, product cut-outs and character sets drop straight into a layout.

A film-style photo of a father lying on a purple sofa with a sleeping baby on his chest, soft window light and a mustard blanket in the foreground
03

Photos that look like photos

ByteDance tuned the 5.5 release for natural light, real materials and skin texture, with less of the glossy, overexposed “AI look”. Describe the light, the camera and the film stock, and it follows them.

An infographic titled The Blue Morpho with eight small captions around the butterfly, on antennae, wing scales, underside, wingspan, diet, life cycle, habitat and lifespan, every line readable
04

Small text that stays readable

Dense captions, newspaper columns and many-stroke Chinese characters come out cleaner than in earlier versions. In ByteDance’s tests, severe text errors fell to 0.4% of images. Put every word in quotes and check spelling in the result.

A reference sheet of four felt plush toys, a blue bunny, yellow chick, pink bear and green dinosaur, and the same four toys at a picnic on a red checked blanket
05

A series from up to 10 references

Upload characters, products or a style and build a series: the same characters in new poses and places, or a product in every colourway. Name each upload @Image1, @Image2 and so on, and say what must stay the same.

Text to image vs image editing

Both tabs run the same Seedream model. Text to image starts from words only. Image edit starts from your pictures.

Text to imageImage edit
Starts fromA prompt1 to 10 images and a prompt
Name images as—@Image1, @Image2…
Aspect ratio8 fixed shapes8 fixed shapes or Auto
Sizes1K, 1.5K, 2K1K, 1.5K, 2K
Transparent backgroundYes, as PNGYes, as PNG
Price1 credit (1K, 1.5K), 2 (2K)The same, plus 1 per 3 images
Best forPhoto-real scenes, posters, stickers and iconsStep-by-step edits, product sets, character series

How to use the Seedream AI image generator

  1. 01

    Pick a tab

    Text to image to start from words, or Image edit to start from up to 10 of your own pictures.

  2. 02

    Write the prompt

    Describe the subject, light, camera and style, and put any text in quotes. In an edit, name your images @Image1, @Image2, say what to change and what to keep.

  3. 03

    Choose shape, size and background

    8 shapes, 1K to 2K, PNG or JPEG, and a transparent background for cut-outs. The price is on the button: 1 credit up to 1.5K, 2 at 2K.

  4. 04

    Generate and refine

    Press Generate. The picture stays under “My images” for 14 days. To change it again, download it and add it to Image edit.

Seedream image pricing per picture

One price per picture, set by its size: 1 credit at 1K or 1.5K and 2 at 2K. Image edit adds 1 credit for every 3 images you upload. Packs from $9.90, no subscription, failed images refunded.

Model1K1.5K2K
Seedream · text to image1 credit1 credit2 credits
Seedream · edit with 1–3 images2 credits2 credits3 credits
Seedream · edit with 4–6 images3 credits3 credits4 credits
Seedream · edit with 7–9 images4 credits4 credits5 credits
Seedream · edit with 10 images5 credits5 credits6 credits

Seedream AI FAQ

Something else? Email support@meliesai.com.

You can try Seedream for free here. New accounts that sign up with Google get 4 free credits, enough for 4 pictures at 1K or 1.5K. After that a Seedream picture costs 1 credit up to 1.5K and 2 at 2K, from one-time packs starting at $9.90, with no subscription.

A Seedream picture costs 1 credit at 1K or 1.5K and 2 credits at 2K. Image edit adds 1 credit for every 3 uploaded images, so a 2K edit from 10 images costs 6 credits. The $9.90 pack has 100 credits, enough for 50 pictures at 2K. The price shows on the button first.

ByteDance Seed released the 5.5 version of Seedream on November 9, 2026, in a launch post titled “More Realistic Images, More Stable Editing Across 100 Rounds”. It followed Seedream 5.0. ByteDance says a Pro version of the 5.5 series will follow; it has not been released yet.

ByteDance lists three upgrades over Seedream 5.0. Edits keep unchanged areas identical at the pixel level, so one picture can be edited many times. Transparent and semi-transparent backgrounds are drawn natively. Photos look more natural and small text is clearer. In ByteDance’s tests, usability beat Seedream 5.0 Pro in all six fields it measured.

It depends on the job. In ByteDance’s own comparison, the 5.5 release led Nano Banana 2.1 and GPT Image 2.5 Sunburst in text structure and in text-to-image realism. Those are ByteDance’s tests, not independent ones. Pick Seedream for photo-real scenes, dense text and repeated edits. Nano Banana 2.1 is in the same image generator, so you can run one prompt on both and compare.

No. Both come from ByteDance’s Seed team, but Seedream makes still pictures and Seedance makes video. A common workflow is to make a picture with Seedream, then animate it with Seedance by uploading it as the first frame. Both are on this site, each with its own generator.

No. ByteDance has announced no open weights for the 5.5 release, so there is no Seedream checkpoint to download and run locally. Seedream is offered through ByteDance’s own apps and through API providers. Here it runs online through an API, with nothing to install.

Yes. ByteDance filters prompts and images, and our API provider’s safety checker is switched on for every Seedream request. Explicit sexual content, real people without consent and other material against our content standards are refused. A refused picture counts as a failed image, and its credits are refunded.

Yes. Set Background to Transparent and keep the format on PNG, and the picture comes back with an alpha channel and no backdrop. ByteDance built the 5.5 release to draw full and partial transparency and fine edges such as feathers itself. JPEG cannot store transparency, so Transparent is greyed out for JPEG.

Open the Image edit tab and add 1 to 10 photos. Then describe the change, naming each photo @Image1, @Image2 in the order you added them. Name the one thing to change in @Image1 and say that everything else stays the same. Pick a shape (Auto lets the model decide) and a size, then press Generate.

Yes, and that is what the 5.5 release is built for. ByteDance edited one portrait 100 times in a row, and the parts that were not edited stayed the same. Here, download each result and add it to Image edit as @Image1 for the next change. Ask for one change per round and say what to keep.

Seedream on this site makes three sizes: 1K, 1.5K and 2K, where a 2K square is 2048 × 2048 pixels. There is no 4K. The shapes are 1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9. In Image edit you can also pick Auto, where the model decides the shape.

Pick GPT Image 2.6 for 4K pictures, quality levels, GIF or SVG output, or to paint a mask over the exact area to change. Pick MiniMax H3 Image for 16-megapixel pictures or edits from up to 16 images. Pick Seedream for lifelike photos, small text, transparent cut-outs and repeated edits up to 2K. All three are in the same image generator.

No visible watermark or logo is added to Seedream pictures made on Melies AI. Pick PNG for the best quality or transparency, or JPEG for a smaller file. Each picture stays under “My images” for 14 days, so download the ones you want to keep.

Yes. Pictures you generate are yours to use, including in ads, client work and products, as long as your prompt and uploaded images do not infringe anyone else’s rights and follow our content standards. Edits of real people need their consent.

No. Melies AI is an independent platform that gives you access to ByteDance’s Seedream model through a third-party API provider. It is not affiliated with, endorsed by or sponsored by ByteDance. ByteDance’s own Seedream page is on seed.bytedance.com, and ByteDance’s creative app Dreamina also offers the model.

Make your first Seedream picture

Write a prompt or upload your photos, pick a shape and a background, and download the PNG.

Create an image