Seedream 4.5 on Twelver

Seedream 4.5: photoreal 4K
with readable text

ByteDance’s image model for studio-grade results: dense photorealism, legible text inside the image, native 4K — all on one Twelver account.

Advertising-grade photorealism
Text inside the image stays legible — banners, covers, product cards
4K and wide 21:9 — drop it straight into the layout

Seedream 4.5 right in the Twelver chat

The agent is already wired to this model: write a prompt or drop a photo and the result lands in the thread. Then just say “darker”, “drop the caption”, “same but vertical”.

A ready-made agent — the model is already selected
Refine with words instead of rewriting the prompt
History, inputs and outputs all stay in the thread

Every model is included in one Twelver subscription — cheaper than paying each service separately. Start on free tokens, no card needed.

Seedream 4.5 inside Claude, Cursor, LM Studio and other AI clients

A remote MCP server over HTTP: tick the tools you want, paste the connector URL, and the model shows up in your assistant. No npm, no local servers, OAuth sign-in.

Open Claude → Settings → Connectors

Add custom connector → paste the URL

Build your server

Check the tools — URLs and config update themselves.

Генерация изображений (Seedream)

Connector URLs

Генерация изображений (Seedream)

https://twelver.ru/api/mcp/v1/seedream-image-generation

Seedream 4.5 from your own app

Three calls: upload the source (if there is one), create the task, fetch the result. The request schema and the tool list are served by the API itself.

Key and config

Create a config and a key right here — the samples below fill in your real config id.

1. Настройте конфигурацию

2. Получите API ключ

Upload a source fileoptional

Only needed when editing an existing photo or animating a frame. Skip it when generating from scratch.

curl -X POST https://twelver.ru/api/ai/<config-id>/v1/media/files \
  -H "Authorization: Bearer $TWELVER_API_KEY" \
  -F "[email protected]"

# → { "url": "https://twelver.ru/api/files/u/<id>/url?token=…",
#     "content_type": "image/jpeg" }

curl -X POST https://twelver.ru/api/ai/<config-id>/v1/media/files \
  -H "Authorization: Bearer $TWELVER_API_KEY" \
  -F "[email protected]"

# → { "url": "https://twelver.ru/api/files/u/<id>/url?token=…",
#     "content_type": "image/jpeg" }

Create the generation task

A prompt, the toolsets, and optionally the source file URLs. You get a task id back immediately.

curl -X POST https://twelver.ru/api/ai/<config-id>/v1/media/generations \
  -H "Authorization: Bearer $TWELVER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "Editorial product photo of a perfume bottle on wet stone, moody lighting, 4K",
    "toolsets": ["seedream-image-generation"],
    "files": [{ "url": "<url from step 1>" }]
  }'

# → 202 { "id": "<task-id>", "status": "running" }

curl -X POST https://twelver.ru/api/ai/<config-id>/v1/media/generations \
  -H "Authorization: Bearer $TWELVER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "Editorial product photo of a perfume bottle on wet stone, moody lighting, 4K",
    "toolsets": ["seedream-image-generation"],
    "files": [{ "url": "<url from step 1>" }]
  }'

# → 202 { "id": "<task-id>", "status": "running" }

Retrieve the result

Poll the task until its status is completed — the output carries the text and the URLs of the finished files.

curl https://twelver.ru/api/ai/<config-id>/v1/media/generations/<task-id> \
  -H "Authorization: Bearer $TWELVER_API_KEY"

# → { "status": "completed",
#     "output": { "text": "…",
#                 "files": [{ "url": "…", "content_type": "image/png" }] } }

curl https://twelver.ru/api/ai/<config-id>/v1/media/generations/<task-id> \
  -H "Authorization: Bearer $TWELVER_API_KEY"

# → { "status": "completed",
#     "output": { "text": "…",
#                 "files": [{ "url": "…", "content_type": "image/png" }] } }

API schema

A separate endpoint returns the available toolsets with their tools and the exact request/response shape — the same contract the backend runs on.

curl https://twelver.ru/api/ai/<config-id>/v1/media/generations/schema \
  -H "Authorization: Bearer $TWELVER_API_KEY"

# → { "toolsets": [{ "id": "seedream-image-generation", "tools": [...] }, ...],
#     "request": {"prompt":"string","toolsets":["string"],"files":[{"url":"string","content_type":"string","description":"string"}]},
#     "response": { ... } }

curl https://twelver.ru/api/ai/<config-id>/v1/media/generations/schema \
  -H "Authorization: Bearer $TWELVER_API_KEY"

# → { "toolsets": [{ "id": "seedream-image-generation", "tools": [...] }, ...],
#     "request": {"prompt":"string","toolsets":["string"],"files":[{"url":"string","content_type":"string","description":"string"}]},
#     "response": { ... } }

API generations spend the same tokens as the chat — there is no separate plan to buy.

More about the API →

📸

Photorealism

Materials, skin, reflections and light read like a studio shoot — noticeably less plastic than most models.

🔤

Text inside the image

Signage, packaging and captions come out legible, which makes it the model for banners, covers and listings.

🖼️

4K and 21:9

Generate straight to 4K and wide frames — no upscaling step before layout.

🔀

Reference-driven edits

Feed your own photos: style transfer, swapping an object in the frame, composing one scene from several sources.

Model: Seedream 4.5 (ByteDance / BytePlus)
Resolution: 2K and 4K
Aspect ratios: 1:1, 3:4, 4:3, 16:9, 9:16, 3:2, 2:3, 21:9
Inputs: Text, one photo, or several references
Tools: Text-to-image, image-to-image, reference blending

seedream

You

Make an ad frame of a thermos flask: wet dark stone, water splashes frozen in the air, cold rim light. Studio, tack sharp, 4K.

Twelver

Done — the file is in the chat

You

Now a key visual for a coffee brand: a latte on a marble counter, morning light, with a large MORNING BLEND headline and a roasted daily subtitle.

Twelver

Done — the file is in the chat

1Photorealism: materials, splashes and rim light — like a real shoot

2The text in the image is legible — the banner is ready to ship

Создать аккаунт
+50,000
Зарегистрируйтесь, чтобы сохранять чаты, агентов и настройки на всех устройствах.
Первый веб-поиск
+20,000
Задайте вопрос, требующий свежих данных — AI выполнит поиск в интернете и укажет источники.
Создать первый документ
+20,000
Создайте структурированный документ — статью, отчёт, план — прямо из чата и редактируйте его совместно.
Загрузить первое изображение
+20,000
Перетащите изображение в чат — AI увидит его и ответит на вопросы о содержимом или отредактирует.
Создать первого агента
+50,000
Создайте собственного AI-агента со своим характером, инструментами и инструкциями. Требуется пополненный баланс.
Создать аудио
+50,000
Сгенерируйте речь, озвучку или звуковой эффект по текстовому промпту.
Создать видео
+50,000
Выберите агента с поддержкой видео (или создайте специального) и опишите нужную сцену.
Создать изображение
+40,000
Попросите чат сгенерировать изображение — пробуйте разные стили, соотношения сторон и подробные промпты.
Общаться 2 дня подряд
+20,000
Заходите два дня подряд, чтобы начать серию. Любой короткий диалог засчитывается.
Общаться 4 дня подряд
+40,000
Поддерживайте серию — четыре дня подряд активности в чате.
Общаться 7 дней подряд
+60,000
Целая неделя серии подряд. Вы в огне!
Общаться 30 дней подряд
+100,000
Целый месяц ежедневных чатов — высшая награда за серию.
Пополнить баланс
+100,000
Пополните баланс, чтобы открыть премиум-модели и инструменты. Токены доступны, пока активна подписка — они сгорают, когда подписка заканчивается.
Пополнить баланс дважды
+100,000
Второе платное пополнение даёт бонусные токены сверх купленных.

Nano Banana Ideogram Grok Imagine

Everything Twelver does with images →

We compared both on our own prompts and kept 4.5 — it gave us the better output. That is a deliberate choice, not us lagging behind: the moment a newer version wins the comparison we switch the agent over, with nothing to change on your side.

Yes — free tokens land every month, enough to try the model. After that you pay per generation.

Yes: upload an image and describe the change — Seedream supports image-to-image and blending several references.

The REST generation endpoint with toolsets: ["seedream-image-generation"] — see the request example in “Three ways to use”. The same engine is also an MCP server for Claude, Cursor and LM Studio.

Yes — the images you generate are yours, commercial use included.

Make your first shot with Seedream

One subscription covers every model on Twelver — cheaper than paying each service separately. Start on free tokens, no card needed.

Seedream 4.5: photoreal 4K
with readable text

Three ways to use Seedream 4.5

Seedream 4.5 right in the Twelver chat

Seedream 4.5 inside Claude, Cursor, LM Studio and other AI clients

Build your server

Connector URLs

Seedream 4.5 from your own app

Key and config

Upload a source fileoptional

Create the generation task

Retrieve the result

API schema

What Seedream is best at

Photorealism

Text inside the image

4K and 21:9

Reference-driven edits

Specs

An example, start to finish

Получайте бесплатные токены

Other image models

FAQ

Make your first shot with Seedream

Seedream 4.5: photoreal 4Kwith readable text

Three ways to use Seedream 4.5

Seedream 4.5 right in the Twelver chat

Seedream 4.5 inside Claude, Cursor, LM Studio and other AI clients

Build your server

Connector URLs

Seedream 4.5 from your own app

Key and config

Upload a source fileoptional

Create the generation task

Retrieve the result

API schema

What Seedream is best at

Photorealism

Text inside the image

4K and 21:9

Reference-driven edits

Specs

An example, start to finish

Получайте бесплатные токены

Other image models

FAQ

Make your first shot with Seedream

Related tools

More AI models

Explore more tools

Seedream 4.5: photoreal 4K
with readable text