Release · Models
Seedream 5.5 is on FlyMy.AI, and it is now the default image model for agents
ByteDance's newest Seedream makes images from a prompt and edits or blends up to 10 reference images, keeping the subject and the parts you did not ask to change. It has been on FlyMy.AI since 16 November 2026, and since 17 November it is the image model an agent gets when it names none. Our five test runs, what each one cost, and how to call it from an agent, an assistant or your own code.
What is Seedream 5.5?
Seedream is ByteDance's family of AI image models, and Seedream 5.5 is its newest release. On BytePlus
ModelArk its model id is dola-seedream-5-5-261028, and it has been generally available there
since 12 November 2026.
On FlyMy.AI it has been live since 16 November 2026 as the endpoint
flymyai/bytedance-seedream-5_5, and it runs through BytePlus ModelArk. You need no BytePlus
account or key: you sign in to FlyMy.AI, and each image is billed to your FlyMy.AI balance. Its page in
the app is Seedream 5.5.
One endpoint does both jobs. A request with a prompt alone makes a new image; add reference images and the same endpoint edits them or blends them into one picture.
What's new: the default image model for agents
Agents and assistants on FlyMy.AI choose a model with recommend_model: they describe the job,
such as "a poster for a jazz night", and get the candidates ranked best first. Since 17 November 2026
(platform log 2.2.5), Seedream 5.5 comes first when the request is for an
image and names no model, and Seedream 5.5 Edit comes first for an edit. Both are the same endpoint: an
edit is a request with reference images.
- Ask for something else and you get something else. A request for speed, for a lower price or for a model by name still picks another model.
- Only the image picks changed. Video, audio and chat picks are the same as before.
- Nothing to switch on. The change is in the ranking
recommend_modelreturns, so any agent or assistant that asks it for an image without naming a model sees Seedream 5.5 first.
Make a poster for a jazz night called BLUE NOTE NIGHTS.
No model named, so your agent picks Seedream 5.5.
What Seedream 5.5 does on FlyMy.AI
It takes a prompt and up to 10 reference images and returns one image. These are its inputs, as its live schema lists them:
| Input | What it does | Default |
|---|---|---|
prompt | The image to make, or the change to make to the references | Required |
image, image1 to image9 | Up to 10 reference images to edit or blend, as public https URLs or uploaded files | None |
size | 2k, or one of 11 fixed sizes | 2k |
watermark | A watermark on the image | Off |
optimize_ | Prompt optimization | standard |
With size at 2k, the default, the model picks the aspect ratio from the prompt: our
poster prompt came back at 1664x2496, a tall 2:3 sheet. Or set a size yourself:
| Shape | Fixed sizes |
|---|---|
| Square | 1536x1536, 2048x2048 |
| Landscape | 1280x720 and 2560x1440 (16:9), 2304x1728 (4:3), 2496x1664 (3:2), 3024x1296 (7:3) |
| Portrait | 720x1280 and 1440x2560 (9:16), 1728x2304 (3:4), 1664x2496 (2:3) |
- One image per request. For three versions, run it three times.
- No watermark unless you turn it on.
- About 20 to 30 seconds per image in our runs.
Our five test runs
We tested Seedream 5.5 with five runs through FlyMy.AI. Every image in this section and the one at the top is one of those runs as the model returned it: we only scaled the images down for the page, trimmed the one at the top to its frame and cut out two label details at full size. Each price is what FlyMy.AI billing charged for the run. The clips and stills marked as official examples further down come from BytePlus.
1. A packaging shot at 2560x1440: $0.035
The photo at the top of this post.
A wide studio photo of a frosted glass perfume bottle standing next to its box on a pale sand-colored background, the box reads LUMEN No. 5 in fine gold serif letters, a sprig of dried lavender, soft morning light from the right, long gentle shadows, minimal premium packaging shot.
The name on the box is spelled as asked, in thin gold serif letters, and the frosted glass, the paper of the box and the long shadows read as real materials in real light. The model also added two things we did not ask for: the name on the bottle itself and a ceramic dish at the edge of the frame.
2. A poster at the default size: $0.035
A minimalist poster for a jazz night: a golden saxophone on a deep navy background, the headline BLUE NOTE NIGHTS in bold serif letters, the date NOV 28 below, soft film grain.
With size left at 2k, the model chose a tall poster, 1664x2496. The headline and
the date are spelled right. It also wrote a line we did not ask for above the headline, RIVER CITY JAZZ
CLUB / MUSIC LIVES HERE, so read every word of a generated image before you print it. The poster is on the
left below.
3. The same poster, one edit later: $0.043
We sent the poster from run 2 back as the one reference image, with this prompt:
Replace the golden saxophone with a silver trumpet lying on the same table. Keep the headline BLUE NOTE NIGHTS, the date NOV 28, the navy background, the lighting and the layout exactly the same.
The trumpet is in and the rest held: the headline, the date, the club line, the navy background, the light, the case on the left and the microphone on the right are where they were. Laid over each other, the two images differ only around the instruments: the brass itself, its shadow on the table and two music notes beside it. The run cost $0.035 for the 2k image plus $0.008 for the reference.
4. A small product shot at 720x1280: $0.020
A tall glass bottle of cold-brew coffee on a wet black stone, condensation droplets, the label reads NORTH ROAST in small clean sans-serif type with the line 12 OZ under it, soft studio light from the left, vertical product photo.
720x1280 is one of the three sizes at the lower price, and the small type on the label still came out clean: NORTH ROAST, with 12 OZ under it, and condensation on the glass around it. The bottle is on the left below.
5. Two references blended at 1536x1536: $0.036
We sent the poster from run 2 and the bottle from run 4 as two reference images:
Place the NORTH ROAST coffee bottle from the second image on the wooden table next to the saxophone from the first image. Keep the poster typography BLUE NOTE NIGHTS and NOV 28 and the navy background of the first image.
The bottle crossed over with its label, and the poster kept its type and its navy background, recomposed into a square. Two references at $0.008 each on top of $0.02 for the size made $0.036.
Where Seedream 5.5 is strongest
ByteDance positions Seedream 5.5 for four things, and our runs bear out each one.
Edits keep the picture steady
Over rounds of edits, the subject, the composition and the areas you did not mention stay put. In run 3 only the instrument and what it touched changed.
Small text stays legible
Labels and fine print come out clean: NORTH ROAST and 12 OZ on a 720x1280 bottle, LUMEN No. 5 in thin gold serif on a perfume box.
Realistic materials and light
Frosted glass, condensation on a cold bottle, wet stone, paper, brass and silver look like what they are.
Product, packaging, poster and campaign imagery
The work it is made for, and the work of all five of our test prompts.
Use cases with prompts to paste
Six kinds of work Seedream 5.5 suits. Each comes with an official example clip from BytePlus's Seedream 5.5 showcase and a request to paste into a FlyMy.AI agent or an assistant connected to FlyMy.AI. The requests are examples to try, not runs we recorded: change the details to yours.
1. E-commerce product imagery
Packshots on a clean background, and the same product in new scenes without the product changing.
Make a 2048x2048 product photo of a white ceramic mug on a light oak table, soft daylight, empty space on the right for text. Then put the same mug on a picnic blanket in a park and keep the mug exactly the same.
2. Design and advertising
Ad visuals and banners with the headline drawn into the picture.
Make a 2560x1440 banner for a running shoe: the shoe in mid-air over a wet city street at dawn, with the headline RUN THE CITY in bold sans-serif letters at the top left.
3. Portrait and retouching
Light, color and background fixes that keep the person as they are.
Retouch the portrait at https://example.com/portrait.jpg: soften the harsh light on the face and warm the colors a little. Keep the person, the expression, the hair and the background the same.
4. Game and film concepts
Environments, props and key art, in the wide formats of a pitch deck or a storyboard.
Make a 3024x1296 concept art of a floating market at dusk for a fantasy game: wooden boats with paper lanterns, mist over the water, a stone bridge in the distance.
5. Content creation: stickers and style transfer
Sticker sheets, and photos redrawn in another style. There is no transparent background on FlyMy.AI, so ask for a plain one.
Turn the photo of my dog at https://example.com/dog.jpg into a sheet of six cartoon stickers with thick white outlines on a plain light gray background.
Redraw the photo at https://example.com/street.jpg as a watercolor illustration and keep the composition.
6. Presentations and infographics
Posters, banners and cover art for decks, reports and events.
Make a 1280x720 cover image for a quarterly report titled Q4 GROWTH REVIEW: a rising line of light over a dark blue grid, the title in clean white sans-serif type.
How to use Seedream 5.5
Four ways in, from a sentence in the app to a request from your own code.
1. Ask an agent in the FlyMy.AI app
Describe the image to a FlyMy.AI agent and name no model. It asks recommend_model, gets Seedream 5.5 first and runs it.
Make a poster for a jazz night called BLUE NOTE NIGHTS.
2. From Claude Code, Claude, ChatGPT or Cursor
Add the FlyMy.AI MCP at https://mcp-agents.flymy.ai/mcp and sign in:
- Claude Code: run
claude mcp add --transport http flymyai https://mcp-agents.flymy.ai/mcp, then/mcpto sign in. - Claude and ChatGPT: Settings, Connectors, Add custom connector, then paste the URL and sign in.
- Cursor: add the same URL as a remote MCP server.
Then ask for an image. Your assistant calls recommend_model, then run_model with:
{
"endpoint_id": "flymyai/bytedance-seedream-5_5",
"input": {
"prompt": "A jazz night poster titled BLUE NOTE NIGHTS, NOV 28",
"size": "2k"
}
}
An edit passes the picture as image, and up to nine more as image1 to image9, each a public https URL:
{
"endpoint_id": "flymyai/bytedance-seedream-5_5",
"input": {
"prompt": "Swap the saxophone for a silver trumpet, keep the rest",
"image": "https://example.com/blue-note-nights.png",
"size": "2k"
}
}
The answer carries a link to the image that expires after 15 minutes, so save the files you keep. If
the run is still going, your assistant reads it with get_model_result.
3. From your code: the Seedream 5.5 API
Send form fields, not JSON, with your FlyMy.AI API key in the X-API-KEY header. Your keys are
on the Developers page; keep yours in an environment
variable, never in code you share.
curl -fsS --no-buffer --max-time 600 \
-H "X-API-KEY: $FLYMYAI_API_KEY" \
-F 'prompt=A jazz night poster titled BLUE NOTE NIGHTS, NOV 28' \
-F 'size=2k' \
https://api.flymy.ai/api/v1/flymyai/bytedance-seedream-5_5/predict
For an edit, add -F 'image=https://example.com/blue-note-nights.png', or upload a file from
your disk with -F 'image=@blue-note-nights.png'. /predict waits for the picture
and answers with it in the image output. To queue a run instead, send the same fields to
/predict/async/ and read the result at
/predict/async/result/?request_id=<prediction_id>. The model's
openapi.json lists
every input, and Model endpoints in the docs
covers the routes.
4. In the playground on the model page
Open Seedream 5.5 in the app, write a prompt, pick a size, add reference images for an edit and run it.
What Seedream 5.5 costs
Seedream 5.5 is billed per image, by the size you ask for and the number of reference images you send.
These are the list prices in FlyMy.AI billing, and an assistant can quote them before a run with
get_pricing.
| What you ask for | Price |
|---|---|
2k, or a fixed size of 2K or larger: 2048x2048, 2304x1728, 1728x2304, 2560x1440, 1440x2560, 2496x1664, 1664x2496 or 3024x1296 | $0.035 per image |
| 1536x1536, 1280x720 or 720x1280 | $0.02 per image |
| Each reference image, for an edit or a blend | + $0.008 |
A 2K image from Seedream 5.0 Pro costs $0.09, so the same size from Seedream 5.5 costs less than half as much. Our five runs, as billed:
| Our run | Size and references | Billed |
|---|---|---|
| 1. Perfume packaging shot | 2560x1440 | $0.035 |
| 2. Jazz poster | 2k, came back at 1664x2496 | $0.035 |
| 3. Poster edit, a trumpet for the saxophone | 2k, 1 reference | $0.043 |
| 4. Cold-brew bottle | 720x1280 | $0.020 |
| 5. Poster and bottle blended | 1536x1536, 2 references | $0.036 |
| All five runs | $0.169 |
Charges as FlyMy.AI billing recorded them for our runs: $0.043 is $0.035 plus one reference at $0.008, and $0.036 is $0.02 plus two at $0.008.
What Seedream 5.5 does not do on FlyMy.AI
- No transparent background, layers or image sequences. The transparent-background switch, layer decomposition and multi-image sequences are not exposed through FlyMy.AI.
- One image per request. For several versions, run it several times; each run is billed.
- Up to 10 reference images per request:
imageandimage1toimage9. - No video. For a video, ask
recommend_modelfor a video model. - Read the text before you publish. The model can add words you did not ask for, like the club line on our poster.
FAQ
What is Seedream 5.5?
Seedream 5.5 is ByteDance's newest Seedream image model. It makes images from a prompt and edits or blends up to 10 reference images, keeping the subject and the untouched areas steady across rounds of edits. On BytePlus ModelArk its id is dola-seedream-5-5-261028, and on FlyMy.AI it is the endpoint flymyai/bytedance-seedream-5_5.
When was Seedream 5.5 released?
It has been generally available on BytePlus ModelArk since 12 November 2026 and on FlyMy.AI since 16 November 2026. Since 17 November 2026 it is the default image model on FlyMy.AI.
How much does Seedream 5.5 cost?
On FlyMy.AI, $0.035 per image at 2K or larger sizes and $0.02 at 1536x1536, 1280x720 or 720x1280, plus $0.008 per reference image. A 2K edit with one reference costs $0.043. Seedream 5.0 Pro costs $0.09 per 2K image.
Is there a Seedream 5.5 API?
Yes. Send form fields with your FlyMy.AI API key in the X-API-KEY header to https://api.flymy.ai/api/v1/flymyai/bytedance-seedream-5_5/predict. From Claude Code, Claude, ChatGPT or Cursor, the FlyMy.AI MCP runs it with run_model.
Do I need a BytePlus or ByteDance account?
No. Seedream 5.5 runs on FlyMy.AI through BytePlus ModelArk: you sign in to FlyMy.AI only, and each image is billed to your FlyMy.AI balance.
Can Seedream 5.5 edit images?
Yes. Send the picture as image, and up to nine more as image1 to image9, with a prompt that says what to change and what to keep. In our test it swapped a saxophone for a trumpet and kept the rest of the poster. Each reference image adds $0.008.
What sizes can Seedream 5.5 make?
2k, the default, where the model picks the aspect ratio from the prompt (1664x2496 for our poster), or one of 11 fixed sizes: 1536x1536, 1280x720, 720x1280, 2048x2048, 2304x1728, 1728x2304, 2560x1440, 1440x2560, 2496x1664, 1664x2496 and 3024x1296.
Does Seedream 5.5 add a watermark?
Not on FlyMy.AI unless you ask for one: watermark is off by default.
Can Seedream 5.5 make transparent backgrounds or video?
No. Through FlyMy.AI the transparent-background switch, layer decomposition and multi-image sequences are not exposed, and Seedream 5.5 makes no video. Each request returns one image.
Why did my agent pick Seedream 5.5?
Since 17 November 2026 it is the default image model on FlyMy.AI. When an agent or assistant asks recommend_model for an image and names no model, Seedream 5.5 comes first, and Seedream 5.5 Edit for an edit. Ask for speed, a lower price or another model by name to get a different one.
How long does Seedream 5.5 take?
About 20 to 30 seconds per image in our runs through FlyMy.AI.
Make your first image with Seedream 5.5
Ask a FlyMy.AI agent for an image and name no model, or open Seedream 5.5 and run it in the playground. From $0.02 per image.