PRODUCT REMIX · FOR SELLERS

Your product.
A new look.

Turn your product URL or photo and a competitor reference into new product and ad images—with a similar look, feel and layout.

Powered by ModelArk Managed Agent
Seedream 5.0 Pro

An unbranded blue bottle lying across a bicycle saddle bag with a cord over its body
OUTDOOR / LIFESTYLESeedream 5.0 Pro · Image editing
YOUR PRODUCTAI-generated blue bottle used as the demo product inputThe product to feature
THE REFERENCECompetitor image showing a bottle on a bicycle saddle bag with an outdoor backgroundThe look to adapt

Borrow the look.
Make it your product.

Product listings. Lifestyle scenes. Ad creative.

PRODUCT REMIX

See the direction.
Make it your own.

The reference guides the lighting, setting and layout. Your product stays at the centre.

Source image of the unbranded blue demonstration bottle
01
Demo productAI-generated source image
The competitor reference with cycling props, soft lighting and a dark green background
02
Competitor referenceStyle and layout inspiration
The blue bottle resting horizontally on the saddle bag with visible support and a cord across its body
03
A new imageCreated with Seedream 5.0 Pro
About this example

An unbranded, AI-generated demo product is used as the image input. The reference is analyzed separately; Seedream receives the product image and written style and layout instructions.

Generated in a managed-agent session, then edited with Seedream 5.0 Pro to correct the bottle’s support. The bottle now rests horizontally on the bag with a cord across its body. The saddle partially covers the rear of the bottle.

HOW IT WORKS

Your agent.
BytePlus behind it.

Use Claude Code or Codex to collect the inputs. ModelArk Managed Agent handles image creation and review.

A product URL or photo and competitor reference pass through Claude Code or Codex to ModelArk Managed Agent. Seed 2.1 analyzes and reviews; Seedream 5.0 Pro creates the new image. The result returns to your agent.

Swipe to explore the diagram.

For developers: inputs and integration

The client agent retrieves product images from a supplied URL through its configured browser or web tools, or accepts an uploaded photo. It sends the inputs and task to a managed-agent session through ArkCLI or the Session API.

Seed 2.1 reads the reference’s style and layout, extracts product facts and reviews the result. Seedream 5.0 Pro receives product photos and derived written instructions; the competitor image stays out of generation. The managed agent returns the image and any quality issues.

The configured image model is Pro only. Errors are reported; model, size and quality changes require your decision.

CREATE YOUR NEXT IMAGE

Bring your product.
Choose your reference.

Install Product Remix in Claude Code or Codex.
Connect your own ModelArk Managed Agent.

v1.1.0 · ZIP, 23 KB · Claude Code & Codex

TRY THIS REQUEST
“Create a new product or ad image with a similar look, feel and layout to this reference. Use my product and brand, and flag any quality issues.”

Attach your product photo or URL and the reference image.

Install in your own workspace
  1. Download and extract

    Open a terminal in the extracted byteplus-product-remix folder.

  2. Install for your agent
    python3 install.py --client codex --project .

    Requires Python 3. Replace . with another project folder if needed.

  3. Install ArkCLI and sign in

    ArkCLI connects your agent to ModelArk. Install it first, using Node.js and npm.

    npm install -g @byteplus/ark-cli
    arkcli auth login

    Sign in with your own BytePlus account.

  4. Connect your ModelArk account

    Start a new chat in that project and send this request. The bundled guide covers the managed skill, agent, environment and your own API key.

    Use the product-remix skill. Follow its setup guide to connect this project to my own ModelArk Managed Agent, using Seedream 5.0 Pro and the quality gate.

Can I start with a product URL?

Yes. Your connected agent uses its browser or web tools to retrieve product images from the page. If the page cannot be accessed, upload a product photo.

What should I check in the result?

The skill checks product details, wording, style and physical plausibility. Failed images receive up to two corrections with the same model and size. Review the approved image before publication.

Where does image generation happen?

In your ModelArk Managed Agent session, using the configured skill and API access. This page introduces the solution; it does not generate images directly.