We're excited to announce Talents, a new way to build your automations that's dramatically faster than UI-based browser automations.

The Speed Problem with Browser Automation

UI-based Browser automation is powerful. You can automate virtually anything a human can do on the web: click buttons, fill forms, navigate pages, extract data. But it comes with inherent overhead:

UI Rendering: Every page load, every element interaction requires the browser to render and respond
LLM Processing: Each step requires interpreting instructions and understanding the current page state
Sequential Execution: Actions must wait for pages to load and elements to become interactive

For many use cases, this overhead is acceptable. But what if your workflow is stable, your target website's API hasn't changed, and you just need to run the same automation thousands of times? That's where Talents come in.

What Are Talents?

Talents are API endpoints generated directly from your UI-based automation workflow runs. Instead of replaying browser interactions step-by-step, Talents intelligently replay the underlying network calls that your workflow captured, transforming inputs and outputs exactly as your original workflow did.

Think of it this way:

Workflows = Automation through the UI (flexible, visual, AI-powered)
Talents = Automation through the wire (fast, deterministic, API-driven)

How Talents Work

1. Build and Run Your Workflow as Usual

Build your workflow using the Workflow Builder and run it from the UI just like you always do. While triggering the run, toggle "Prepare for Talent generation" to true. With this, Witrium's browser captures every network request and response behind the scenes.

2. Click "Generate Talent"

After a successful run, you'll see a new "Generate Talent" button on the run results page. One click triggers our AI to analyze:

Your workflow's natural language instructions
The network calls captured during execution
The data transformations between input and output

3. Get Your Talent Endpoint

Within a few minutes, you have a production-ready API endpoint that:

Replays the network calls from your workflow run in a deterministic manner, with variable inputs.
Transforms data according to your original instructions
Returns output in the same format as your workflow, with outputs corresponding to the inputs.

No UI interaction. No LLM interpretation at runtime. Just fast, direct network execution.

The Performance Advantage

Let's put this in perspective with a real example:

Metric	Traditional Workflow	Talent
Execution Time	150-180 seconds	8-15 seconds
LLM Calls	Per instruction step	Zero at runtime

Talents are typically 10-20x faster than their equivalent workflows. For high-volume use cases like scraping thousands of pages, monitoring hundreds of endpoints, or powering real-time integrations, this difference is transformative.

Same Stealth, Same Reliability

You might wonder: "If Talents skip the UI, do I lose Witrium's stealth capabilities?"

No. Talents run within the same browsers as workflows. This means:

✅ Session Management: Talents can use your saved login states
✅ Provider Configurations: All your proxy and browser settings apply
✅ Fingerprint Consistency: Requests maintain the same browser fingerprint
✅ Cookie Handling: Authentication cookies are preserved and sent correctly

From the target website's perspective, it's indistinguishable from a regular browser request. You can still use the same stealth capabilities as you would with workflows.

When to Use Talents vs. Workflows

Use Talents When:

Your automation is stable and well-tested
You need high-volume, repeated execution
Speed is critical (real-time data feeds, rapid monitoring)
You're integrating Witrium into existing data pipelines
The target website's structure rarely changes

Stick with Workflows When:

You're building or iterating on an automation
The target website frequently changes its UI or API
Your automation requires dynamic decision-making
You need visual debugging and step-by-step execution
The task involves complex, multi-page navigation

Pro tip: Use workflows to build and refine your automation, then generate Talents for production execution at scale.

Getting Started with Talents

Step 1: Complete a Successful Workflow Run

Navigate to any of your workflows and execute a run. On the run execution modal toggle "Prepare for Talent generation" to true. Make sure it completes successfully with the expected output.

Step 2: Generate Your Talent

On the run results page, click the "Generate Talent" button. Our AI will analyze the run and create your Talent endpoint within a few minutes.

Step 3: Use Your Talent

Once generated, you can invoke your Talent via:

cURL:

curl -X POST 'https://api.witrium.com/v1/talents/<TALENT_ID>/run' \
  -H 'Content-Type: application/json' \
  -H 'X-Witrium-Key: YOUR_API_KEY' \
  -d '{
    "args": {
      "search_term": "laptop stands"
    },
    "files": [],
    "use_states": [],
    "preserve_state": "my-saved-state",
    "keep_session_alive": false,
    "use_existing_session": null
  }'

Python SDK:

from witrium import AsyncWitriumClient
from witrium.types import TalentRunOptionsSchema

async with AsyncWitriumClient(api_token="YOUR_API_KEY") as client:
    result = await client.run_talent(
        talent_id="<TALENT_ID>",
        options=TalentRunOptionsSchema(
            args={
                "search_term": "laptop stands"
            },
            use_states=["state-id"],  # Optional: list of saved states
            preserve_state="new-state-name",  # Optional: state name to save
            keep_session_alive=False,  # Optional: keep browser running
            use_existing_session=None  # Optional: run ID of existing session
        )
    )

JavaScript/TypeScript SDK:

import { WitriumClient } from '@witrium/witrium';

const client = new WitriumClient("YOUR_API_KEY");

const result = await client.runTalent("<TALENT_ID>", {
  args: {
    "search_term": "laptop stands"
  },
  files: [],
  useStates: ["state-id"],  // Optional: list of saved states
  preserveState: "new-state-name",  // Optional: state name to save
  keepSessionAlive: false,  // Optional: keep browser running
  useExistingSession: "talent-invocation-id"  // Optional: run ID of existing session
});

Step 4: Scale with Confidence

Talents are designed for high-throughput execution. Run them in parallel, integrate them into your pipelines, and process data at speeds that weren't possible with browser-based automation.

The Best of Both Worlds

Talents represent a fundamental evolution in how we think about web automation:

Build visually with workflows. No code required, just natural language instructions.
Test interactively with live browser sessions and step-by-step debugging.
Deploy efficiently with Talents for API-speed execution at scale.

You get the flexibility and power of browser automation for development, combined with the speed and reliability of direct API calls for production.

What's Next

We're just getting started with Talents. On our roadmap:

Talent refinement to improve the accuracy of the Talent
Talent versioning to manage changes over time

Start Building Talents Today

Talents (beta) are available now for all Witrium users. Head to any completed workflow run and click "Generate Talent" to experience the speed difference for yourself.

Have questions about Talents or want to share how you're using them? Reach out to our team at support@witrium.com.

Talents are a new feature and we're actively improving them based on user feedback. If you encounter any issues or have suggestions, we'd love to hear from you.

Introducing Talents: 10x Faster Automation Without the UI