SAM 3 Online Workspace

The SAM 3 Online Workspace is a professional-grade image segmentation tool by SegmentLens. Unlike the local Playground, it leverages powerful cloud computing resources and supports advanced text prompts and AI Agent modes, providing a more precise and intelligent segmentation experience.

Go to SAM 3 Online Workspace →

Features

• Cloud Processing: Powered by high-performance servers. No need to download large AI models to your local browser—get started instantly.
• Text Prompt Segmentation: Simply enter English prompts (e.g., person, cat, yellow car), and the AI will automatically identify and extract the targets.
• AI Agent Mode: Supports natural language descriptions. For example, enter "the largest animal on the left" or "all buildings in the background," and the AI will intelligently understand and complete the task.
• Saved workspace results are available in cloud history, subject to the retention policy below.
• Batch Processing & Merging: Supports multi-selecting objects for merging and exporting, providing efficient productivity tools.
• AI Erase & AI Upscale: Remove people or objects and fill in the background, or enlarge the whole image 2x or sharpen only the objects you pick. Every result is a new version on the image's timeline. See AI Erase, AI Upscale and the Timeline.

How to Use

1. Upload Image

Upload an image to the cloud workspace (up to 30 MB). The workspace resizes processing inputs; standard exports are about 1024 px, while paid HD export crops the source image. A subscription does not change these limits.

2. Choose Segmentation Mode

The Online Workspace offers two core segmentation methods:

Visual Mode (Click-to-Segment)

When you know exactly which area you want to segment, Visual Mode lets you guide the model with point prompts directly on the image.

• Open the Visual tab in the sidebar.
• The control panel has two layers:
  • Mode toggle (top): Add points to place new prompts, or Browse objects to inspect already-segmented masks.
  • Point type (sub-control, only visible while adding): Include (green +) marks an area to keep, Exclude (red −) marks an area to remove.
• Click on the image to drop a point. The marker drawn on the canvas is exactly the same icon shown in the picker, so you always know what you placed.
• Combine include and exclude points to refine the mask — for example, one include point on a person's body, then an exclude point on a held object you want to ignore.
• Use Clear all points to start the prompt over, or Clear masks to remove the current segmentation result and try again.
• When you have at least one point, click Visual Segment to run the model. Each visual segmentation consumes credits per the usage table.

Quick Mode

• Enter the names of objects in English in the search box.
• Supports multiple keywords separated by commas (e.g., mountain, sky, lake).
• Click Find Objects, and the AI will instantly locate and segment these objects.

AI Agent Mode

• Switch to the AI Agent tab.
• Describe your requirements using natural language (e.g., Identify all the players wearing red jerseys).
• The AI will reason, analyze the image content, and precisely locate the matching objects.

3. Manage and Export Objects

• Object List: View all identified objects in the right panel.
• Smart Filtering: Quickly filter content using the confidence slider or category filters.
• Merge & Export: Select multiple objects and use the Merge feature to combine them.
• Transparent PNG: Download high-definition segmentation results with transparent backgrounds in one click.

Credits and Permissions

• Credit Consumption: The Online Workspace uses high-performance cloud resources. Each segmentation task (finding objects or asking the AI) consumes a certain amount of credits.
• New User Benefits: New users get free initial credits upon registration to experience the full range of features.
• Subscriptions provide monthly credit grants; one-time packages do not renew. Both currently allow API use and use the same processing limits. HD, batch export and the commercial-license file are charged separately.

FAQ

Why do prompts need to be in English?

Currently, the underlying multi-modal recognition models are most accurate with English. We recommend using simple English nouns or phrases for the best results.

Is my data secure?

Saved cloud-workspace images and results remain in your history; they are not automatically deleted after 24 hours. The SAM3 history cleanup job targets non-sample records older than 7 days for users without an active/trialing subscription or a completed one-time/credit-package purchase. Records exempt from this job are retained without a fixed automatic expiry. This cleanup does not cover video, 3D, sticker or other uploads. You can delete saved history in the app or contact support about data removal; cleanup timing depends on successful job execution.

What if I run out of credits?

You can click Get More Credits in the interface to top up or earn reward credits by participating in community activities.