Transformation Agent: AI-Powered Image Generation and Editing

Summary

Transform, repurpose and reimagine existing assets for any market, season, campaign or audience, so your teams can meet growing content demands within the resource and production constraints they have. Transformation Agents remove, replace or generate new backgrounds, expand and resize images and generate and edit images with a prompt.

Originals are always preserved: transformed output is saved as a new asset with an indicator marking it as AI-created.

Available features

  • Generate Image: create a unique image from a natural language prompt.
  • Remove Background: instantly isolate subjects from their backgrounds.
  • Replace Background: change the background behind your subject using text prompts, reference images, or solid color HEX codes.
  • Resize (Crop & Extend): adapt images to predefined presets or custom pixel dimensions while preserving quality and core content.
  • Drop Shadow: add depth and polish to isolated objects.
  • Smart Edit: use natural language commands (for example, "remove object", "add gloves", "change location") to make targeted changes to an image.
  • Iterative editing and Version History: build on your image step-by-step by layering prompts and tools, with full freedom to jump back to earlier session versions and branch off into new variations.
  • Custom Naming: set a uniform naming convention for all assets created with the Transformation agent. 
  • Preview image
  • Brand guidelines and reference assets: inform transformations and secure brand guidelines with up to 14 reference images and Brand Guideline PDF documents.  

The dual-workflow system

Bynder AI supports two distinct workflows, depending on whether you need automated efficiency at scale or interactive creative editing:

  • Automated Transformation Agents (bulk/scale): built in the Agent Library as a reusable blueprint. Chain single-action steps into a named agent that can run on multiple assets simultaneously, by different users.
  • On-the-fly interactive editing (single asset): executed directly from an individual asset's detail page, giving you manual control over the image editing tools.
     

Configuring rights and permissions

Administrators set up who can manage and run agents via Settings > Users & Rights > Permission Management. Inside a permission profile, scroll to the Bynder AI section to toggle:

  • Manage Agents: design and manage reusable Transformation Agents in the Agent Library.
  • Run Agents: execute preconfigured agents on selected assets.
  • GenAI capabilities: individual tools can be granted or restricted per profile: Edit with Prompt, Replace Background, Remove Background, Resize Image, and Generate Image with Prompt.

Users also need standard permissions to download and upload original assets to use these tools.

How?

Workflow 1: Building and running reusable Transformation Agents

Use this workflow to establish repeatable, brand-compliant pipelines for batches of assets.

Step 1: Create the agent blueprint

  1. Navigate to AI Agents to open the Agent Library.
  2. Click Create agent and select Transformation.
  3. Set the functional capabilities: Remove Background, Replace Background, Shadow, Resize & Extend, or Smart Edit.
  4. Select a preview image to test the expected results of your agent.
  5. Click custom naming to set a uniform naming convention for all generated assets.

Step 2: Finalize agent details

  1. Name (required): give your agent a distinct name.
  2. Description (required): describe what the pipeline does.
  3. Set status: choose Draft, Available, or Disabled.

Step 3: Run the agent

  1. Select one or more files in the Asset Bank.
  2. From the action bar, select a published Transformation Agent. Users with the Run Agents permission can view and run all published agents.
  3. Track execution under Agent activity > Transformation.
  4. Review, edit, and save approved variations directly to the DAM or the Waiting Room.

Step 4: Scaling with automation rules

Chain published Transformation Agents to automated backend triggers:

  1. Define the trigger: select a system action (for example, When Asset Published).
  2. Define criteria: apply filters to narrow the scope.
  3. Deploy the agent: select Transformation and map to your named agent.
  4. Toggle the status to Active.
  5. Click Publish rule.

Workflow 2: On-the-fly editing and Smart Edit

Direct text-to-image generation

Located on the Asset Overview page of the DAM:

  1. Click the Create AI Image button, located near the standard Upload functionality.
  2. Type a description of the image you want to create (for example, "Create a photo of a team collaborating in a modern office").
  3. Optional parameters:
  • Aspect ratio: define the desired dimensions (for example, Square, Landscape).
  • Reference assets: upload or select images from the DAM to guide the AI on content, brand style, or composition.
  1. The AI returns a single high-quality image in the image editor.
  2. Use Smart Edit to send follow-up prompts, refining the image until it meets your needs.

Editing an existing asset

  1. Open an asset's Asset Detail view and click Edit with AI.
  2. Select a feature from the utility panel: Remove Background, Replace Background, Resize, or Smart Edit.
  3. Configure the change you want to make and apply it. Review the live preview.
  4. Not satisfied? Adjust your input and click Try again to generate a different variation.

Engaging Smart Edit

  1. Click Smart Edit in the utility panel.
  2. Compose your prompt: tell the agent precisely what to add, remove, or substitute.
  3. Optionally inject asset references: use the attachment slot to pin up to 5 images.
  4. Click Generate.

Good to know

  • You can run several transformations in sequence, for example changing an object with Smart Edit and then removing the background.
  • Not happy with a result? Use the History section to jump back to an earlier variation or start fresh from the original asset.

Saving your transformed assets

  • Use the History section to review and compare variations before saving.
  • Once satisfied, click Save/Send to Waiting Room. The generated asset is saved to the DAM or sent to the Waiting Room, depending on your user permissions.
  • The original asset is always preserved; the new asset carries a tag indicator marking it as AI-created.
  • If you're allowed to download original assets from the DAM, you can also download generated assets to use locally.
  • You can save or download multiple variations from the History section, as long as the editor remains open.

Limitations

  • Smart Edit resolution: engaging Smart Edit restricts output resolution to a maximum of 2048x2048 pixels.
  • Other transformations: images generated with Resize, Background Removal, and Background Generation will not exceed a resolution of 5000x5000 pixels.
  • Supported asset types: these features work exclusively on common raster image formats, including JPEG, PNG, Tif, HEIC, Avif and WebP.
  • Maximum assets per agent run: 1,000 images

Pro tips for great results

Mastering the History section

  • Expect the unexpected: stacking multiple AI actions can occasionally yield unpredictable results. Use History to jump back to the original or a preferred earlier variation if a sequence goes off course.
  • Multi-save: you aren't limited to one final image. Save or download multiple distinct variations from your History timeline while the editor session stays open.

Optimizing Smart Edit

  • Frame prompts positively: avoid negative prompts like "don't edit the product." Mentioning elements you want avoided makes the model more sensitive to them, which frequently triggers unintended edits. State what you want: "Change the color of the product to blue."
  • Watch the framing: Smart Edit uses predefined aspect ratios. By default the ratio is set to Auto, which returns the closest ratio to the original image. Select a different ratio from the dropdown to force uniform scaling.
  • Surgical object and logo integration: to add a specific logo, brand mark, or product shot, upload it as a reference attachment and give clear placement instructions in your prompt: "Place the attached logo flat onto the surface of the mug."
  • Lead with strong action verbs: start prompts with explicit commands like Change, Add, Remove, or Replace to set the correct execution mode instantly.
  • Blend lighting for natural results: attached logos or product references can look pasted on. Fix this by prompting how the object should interact with the scene's light: "Integrate the attached product into the foreground, matching the warm golden-hour lighting of the scene."

Flawless backgrounds and shadows

  • Contrast is key: for the cleanest cutouts with Remove Background, use source images with high contrast and clearly defined subject boundaries.
  • Descriptive background prompts: when replacing backgrounds, write specific prompts detailing the exact setting, environment, and lighting style you want.
  • Order of operations: always Resize your image before replacing the background, so the newly generated environment scales to your target dimensions.
  • Realistic depth: the Shadow tool combined with background removal works best on source files that originally have a solid, non-transparent background.
  • Quality review: check subject edges carefully after background removal, especially with complex background details.
  • Asset organization: use clear naming conventions (such as adding _bg-removed or _resized_1080x1080) and apply organizational tags to your new AI assets.

Related Articles

Agent Library: Creating, Running, and Updating Agents

Workspace: Agent Activity - Reviewing, Editing, and Approving Agent Suggestions

Updated