Retrying CAPTCHAs until you succeed¶

If you're trying to automatically break a CAPTCHA, you aren't always going to get it right on the first try! Let's use my CAPTCHA breaker test site to see how that plays out in practice.

If you haven't gone through the CAPTCHA breaking tutorial yet, I recommend you start there!

Prepare our pieces¶

Get to the CAPTCHA¶

In [63]:

import time
import asyncio
from playwright.async_api import async_playwright

playwright = await async_playwright().start()
browser = await playwright.chromium.launch(headless = False)
page = await browser.new_page()

import time import asyncio from playwright.async_api import async_playwright playwright = await async_playwright().start() browser = await playwright.chromium.launch(headless = False) page = await browser.new_page()

We'll visit the page. I'm going to click the "Generate" button just so we can pretend we did some sort of setup (entering text, etc).

In [64]:

await page.goto('https://jsoma.github.io/captcha-breaker-tester')
await page.locator("[value='Generate']").click()
# Wait for the CAPTCHA to generate
time.sleep(1)

await page.goto('https://jsoma.github.io/captcha-breaker-tester') await page.locator("[value='Generate']").click() # Wait for the CAPTCHA to generate time.sleep(1)

Condense the breaking code¶

If our CAPTCHA breaking code is split across multiple cells, we need to move our CAPTCHA breaking code into a single cell.

In [65]:

from deskew import determine_skew
from wand.image import Image
import numpy as np
import pytesseract

# Save the CAPTCHA
await page.locator('#captcha-holder > img').screenshot(path='captcha.png')

guess = pytesseract.image_to_string('captcha.png').strip()
print("The guess is", guess)

from deskew import determine_skew from wand.image import Image import numpy as np import pytesseract # Save the CAPTCHA await page.locator('#captcha-holder > img').screenshot(path='captcha.png') guess = pytesseract.image_to_string('captcha.png').strip() print("The guess is", guess)

The guess is 65kRtH

Add the testing code¶

You need some way to check whether the CAPTCHA worked or not. In this case, we're looking for the #result div says "Correct!".

In [66]:

await page.locator("#answer").fill(guess)
await page.locator("#test-answer").click()

# There are better ways to do this, but
# os.sleep is the most flexible!
time.sleep(1)

# Something to test if the CAPTCHA was correct or not
correct = await page.locator("#result").inner_text() == "Correct!"
if correct:
    print("It was right!")
else:
    print("It was wrong!")

await page.locator("#answer").fill(guess) await page.locator("#test-answer").click() # There are better ways to do this, but # os.sleep is the most flexible! time.sleep(1) # Something to test if the CAPTCHA was correct or not correct = await page.locator("#result").inner_text() == "Correct!" if correct: print("It was right!") else: print("It was wrong!")

It was right!

Combine the code¶

This is a rough breakdown of how you might implement the CAPTCHA breaker. The important part is a test for success that allows you to break out of the loop.

We use an attempts variable instead of while True because if you don't get it done in 30 tries you probably need to adjust your CAPTCHA-breaking code!

In [67]:

# Visit the page and do any setup you need
await page.goto('https://jsoma.github.io/captcha-breaker-tester')

attempts = 30

# Keep trying to break the CAPTCHA until attempts run out 
while attempts > 0:
    # Convert the CAPTCHA to text
    await page.locator('#captcha-holder > img').screenshot(path='captcha.png')
    guess = pytesseract.image_to_string('captcha.png').strip()
    print("Guessing", guess)

    # Input the answer
    await page.locator("#answer").fill(guess)

    # Test for success
    await page.locator("#test-answer").click()
    time.sleep(1)
    correct = await page.locator("#result").inner_text() == "Correct!"

    # If correct, exit
    # If not, try again
    if correct:
        print("Successful!")
        break
    else:
        # Can't figure out this captcha
        # generate a new one by clicking the button
        print(f"Not successful, {attempts} left")
        attempts = attempts - 1
        await page.locator("[value='Generate']").click()
        time.sleep(1)

# We're out of the loop, but maybe we failed to break it
if attempts == 0:
    raise Exception("Failed to break the CAPTCHA")

print("We finished successfully")

# Visit the page and do any setup you need await page.goto('https://jsoma.github.io/captcha-breaker-tester') attempts = 30 # Keep trying to break the CAPTCHA until attempts run out while attempts > 0: # Convert the CAPTCHA to text await page.locator('#captcha-holder > img').screenshot(path='captcha.png') guess = pytesseract.image_to_string('captcha.png').strip() print("Guessing", guess) # Input the answer await page.locator("#answer").fill(guess) # Test for success await page.locator("#test-answer").click() time.sleep(1) correct = await page.locator("#result").inner_text() == "Correct!" # If correct, exit # If not, try again if correct: print("Successful!") break else: # Can't figure out this captcha # generate a new one by clicking the button print(f"Not successful, {attempts} left") attempts = attempts - 1 await page.locator("[value='Generate']").click() time.sleep(1) # We're out of the loop, but maybe we failed to break it if attempts == 0: raise Exception("Failed to break the CAPTCHA") print("We finished successfully")

Guessing 2TalmE
Not successful, 30 left
Guessing 29399
Not successful, 29 left
Guessing IFg3bR
Not successful, 28 left
Guessing z6tAnX
Successful!
We finished successfully

In [ ]: