Bypassing HCaptcha Using YesCaptcha Service with Python

HCaptcha is a widely used captcha service that protects websites from malicious bots; unlike Google reCaptcha, it is accessible for international sites. The article first introduces HCaptcha's user interaction, which may present a grid of images requiring selection based on a question.

Readers are guided to experience HCaptcha manually via the demo site https://democaptcha.com/demo-form-eng/hcaptcha.html, preferably in Chrome incognito mode, to understand the full verification flow.

To automate solving, the article recommends using the YesCaptcha platform. After registering at http://www.yescaptcha.com/ and obtaining a clientKey, users can create a HCaptchaClassification task through YesCaptcha's API.

The required API parameters are:

type : "HCaptchaClassification" (10 points)

queries : a list of Base64‑encoded images (without the data URI prefix)

question : the question ID extracted from the captcha challenge

Python code is provided to fetch the captcha images and question using Selenium, convert the images to Base64, and submit a task to YesCaptcha:

headers = {'user-agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/100.0.4896.127 Safari/537.36'}

def get_as_base64(url):
    return base64.b64encode(requests.get(url).content)

def get_captcha_image_base64(data):
    image_result = []
    for d in data:
        img_base64 = get_as_base64(d['datapoint_uri'])
        img_base64 = img_base64.decode('utf-8')
        image_result.append({'url': d['datapoint_uri'], 'task_key': d['task_key'], 'base64': img_base64})
    return image_result

def create_task(question, queries):
    url = 'https://api.yescaptcha.com/createTask'
    data = {
        "clientKey": clientKey,
        "task": {
            "type": "HCaptchaClassification",
            "question": question,
            "queries": queries,
        }
    }
    r = requests.post(url, json=data, timeout=60)
    return r.json()

if __name__ == '__main__':
    with open('data.json', encoding='utf8') as f:
        data = f.read()
    data = json.loads(data)
    question = data.get('requester_question', {}).get('zh')
    tasklist = data.get('tasklist')
    images = get_captcha_image_base64(tasklist)
    queries = [d['base64'] for d in images]
    result = create_task(question=question, queries=queries)
    pprint.pprint(result)

Running the script returns a JSON response containing a solution field with a list of booleans indicating which grid images should be selected, matching the order of the submitted queries. These results can be fed back to Selenium to programmatically click the appropriate images and submit the captcha.

The article concludes that, while the focus is on breaking HCaptcha, the surrounding Selenium automation steps are straightforward and well‑documented elsewhere.