Skip to main content

Overview

PhoneClaw’s voice interface lets you create automation workflows by simply describing what you want to accomplish. The system uses speech recognition and AI to generate ClawScript code automatically.
Voice generation is the fastest way to create automations. Just say “open Twitter and post a tweet every hour” and PhoneClaw generates the complete script with scheduling.

How Voice Generation Works

  1. You speak a natural language command
  2. PhoneClaw transcribes your speech
  3. An AI agent analyzes your request
  4. ClawScript code is generated automatically
  5. The script is saved and ready to run or schedule

Setting Up Voice Commands

1
Grant Microphone Permission
2
When you first launch PhoneClaw:
3
  • Tap the microphone button at the bottom of the screen
  • Android will prompt for microphone permission
  • Tap Allow to enable voice input
  • 4
    If you accidentally denied permission, go to Settings → Apps → PhoneClaw → Permissions → Microphone and enable it.
    5
    Start Voice Input
    6
    PhoneClaw uses push-to-talk for voice commands:
    7
  • Tap the 🎤 Tap to Speak button
  • Button turns red with ”🔴 Listening…”
  • Speak your automation command clearly
  • Release or tap again to stop recording
  • 8
    View Generated Scripts
    9
    After speaking:
    10
  • PhoneClaw processes your command
  • Generated code appears in the Generation History tab
  • You can review, edit, or run the script immediately
  • Scripts are automatically saved with timestamps
  • Voice Command Examples

    Simple Automation

    Generate a basic automation workflow:
    "Open Twitter and click the blue post button"
    

    Scheduled Task

    Create an automation that runs on a schedule:
    "Open Twitter and click the blue post button every hour"
    
    When you say “every hour,” “daily,” or “every Monday,” PhoneClaw automatically generates the correct cron expression and schedules the task.

    Multi-Step Workflow

    Generate complex workflows with multiple steps:
    "Open Instagram, upload a photo from gallery, add the caption 'Hello World', and post it"
    

    Conditional Logic

    Generate scripts with branching:
    "Check my notifications and if there's a new message, open it and reply"
    

    Voice Command Patterns

    Action Verbs

    Start commands with clear action verbs:
    • “Open” - Launch apps or navigate
    • “Click” - Tap buttons or elements
    • “Type” - Enter text in fields
    • “Upload” - Select files/photos
    • “Check” - Read or verify content
    • “Wait” - Add delays

    Location Descriptors

    Include position information:
    • “in the top right corner”
    • “at the bottom of the screen”
    • “in the navigation bar”
    • “below the title”
    • “next to the profile picture”

    Visual Characteristics

    Describe what elements look like:
    • “the blue send button”
    • “the red notification badge”
    • “the plus icon”
    • “the heart-shaped like button”
    • “the three-dot menu”

    Timing Keywords

    Specify when to run:
    • “every hour”0 * * * *
    • “every day at 9am”0 9 * * *
    • “every Monday”0 0 * * 1
    • “every 30 minutes”*/30 * * * *
    • “once a week”0 0 * * 0
    You don’t need to know cron syntax! Just use natural language timing and PhoneClaw generates the correct schedule.

    Advanced Voice Commands

    Multi-App Workflows

    "Open Twitter, copy the first tweet, then open Instagram and paste it as a new post"
    
    Generates a script that:
    • Switches between Twitter and Instagram
    • Extracts content from one app
    • Inputs it into another
    • Handles timing between app switches

    Loop Operations

    "Like the first 10 posts on Instagram"
    
    Generates a script with:
    • A loop that runs 10 times
    • Click action for the like button
    • Scroll action to next post
    • Delays between actions

    Data Extraction

    "Read all the notification messages and speak them aloud"
    
    Generates a script that:
    • Opens notifications
    • Uses magicScraper to extract text
    • Iterates through multiple notifications
    • Speaks each one with speakText

    Editing Generated Scripts

    Voice-generated scripts can be edited:
    1
    View the Script
    2
  • Go to the Generation History tab
  • Find your recently generated script
  • Tap to view the full code
  • 3
    Make Adjustments
    4
    Common edits:
    5
    // Increase delays if actions are too fast
    delay(1000) → delay(2000)
    
    // Make descriptions more specific
    magicClicker("Button") → magicClicker("Blue send button in bottom right")
    
    // Add error handling
    try {
      magicClicker("Submit button")
    } catch (error) {
      speakText("Error clicking button")
    }
    
    // Add confirmation checks
    const result = magicScraper("Did the action succeed?")
    if (result.includes("success")) {
      speakText("Action completed")
    }
    
    6
    Re-generate If Needed
    7
    If the script isn’t quite right:
    8
  • Tap the microphone again
  • Rephrase your command with more detail
  • Compare the new generated version
  • Use the version that works best
  • Troubleshooting Voice Commands

    Command Not Recognized

    • Speak clearly and at a moderate pace
    • Use shorter, simpler sentences
    • Avoid background noise
    • Try rephrasing with more common words

    Wrong Script Generated

    • Be more specific about element descriptions
    • Include visual details (color, shape, icon)
    • Specify exact locations (top, bottom, left, right)
    • Break complex commands into smaller ones

    Microphone Button Not Working

    • Check microphone permission in Settings
    • Restart PhoneClaw app
    • Test with another voice app to verify microphone works
    • Ensure device volume is not muted

    Generated Code Has Errors

    • View the Generation History to see exact code
    • Edit the script manually to fix issues
    • Test commands individually before combining
    • Check logs for specific error messages

    Best Practices

    Be Specific

    "Click the button"
    

    Use Natural Language

    "Execute magicClicker function on element with class button-submit"
    

    Include Timing

    "Post to Twitter"
    

    Break Down Complex Tasks

    Instead of:
    "Open Instagram, upload 5 photos from gallery with captions, add hashtags, tag locations, and post them all"
    
    Try:
    "Open Instagram and start a new post"
    
    Then:
    "Select a photo from gallery and add the caption with hashtags"
    
    Then:
    "Post the image"
    

    Real-World Examples

    Here are commands from actual PhoneClaw demos:

    TikTok Video Upload

    "Open TikTok, upload a video from gallery, add the song 'Popular Track', 
    and post it with the caption 'Check this out'"
    

    Email 2FA Code Extraction

    "Open Gmail, find the verification code in the latest email, 
    and read it aloud"
    

    Captcha Solving

    "Look at the captcha image and tell me what text is shown"
    

    Waymo Twitter Automation

    "Post a tweet saying 'Riding in a Waymo' every 30 minutes"
    
    These examples are from actual PhoneClaw demo videos showing real-world automation scenarios.

    Next Steps

    Schedule Your Automations

    Learn cron expressions and recurring tasks

    Multi-App Workflows

    Chain actions across multiple apps

    ClawScript Reference

    See all available functions