自定义验证码类型：向 CaptchaAI 提交不寻常的挑战

遇到滑块、旋转、拼图这类没有标准 API 的验证码，做法只有一个：截图并附上文字说明，作为图片识别任务提交给 CaptchaAI 的 ocr.captchaai.com 接口。国内的极验（GeeTest）、网易易盾、腾讯防水墙大量使用这类交互，本文用一套 Python 函数逐一覆盖。

常见的非标准验证码类型

通用方法：截图 + 说明提交

识别接口对所有视觉类验证码用同一套调用：

• 截图转成 base64；
• 用 textinstructions 附上说明；
• 轮询 res.php 取结果。

import requests
import base64
import time
import os

API_KEY = os.environ["CAPTCHAAI_API_KEY"]


def solve_custom_captcha(image_b64, instructions):
    """Solve any visual CAPTCHA using image + text instructions."""
    resp = requests.post("https://ocr.captchaai.com/in.php", data={
        "key": API_KEY,
        "method": "base64",
        "body": image_b64,
        "textinstructions": instructions,
        "json": 1,
    }, timeout=30)

    result = resp.json()
    if result.get("status") != 1:
        raise RuntimeError(result.get("request"))

    task_id = result["request"]

    time.sleep(10)
    for _ in range(30):
        resp = requests.get("https://ocr.captchaai.com/res.php", params={
            "key": API_KEY, "action": "get",
            "id": task_id, "json": 1,
        }, timeout=15)
        data = resp.json()
        if data.get("status") == 1:
            return data["request"]
        if data["request"] != "CAPCHA_NOT_READY":
            raise RuntimeError(data["request"])
        time.sleep(5)

    raise TimeoutError("Solve timeout")

下面几节都复用它。

滑块定位验证码

让接口只返回 X 方向像素偏移，再用 Selenium 的 ActionChains 拖动。

# slider_captcha.py
from selenium.webdriver.common.by import By
from selenium.webdriver.common.action_chains import ActionChains


def solve_slider_captcha(driver, captcha_selector):
    """Screenshot slider CAPTCHA and solve via CaptchaAI."""
    captcha = driver.find_element(By.CSS_SELECTOR, captcha_selector)
    image_b64 = captcha.screenshot_as_base64

    result = solve_custom_captcha(
        image_b64,
        "What pixel position should the slider be dragged to? "
        "Return only the X offset number."
    )

    try:
        offset = int(result)
    except ValueError:
        return False

    # Drag slider to position
    slider = driver.find_element(By.CSS_SELECTOR, ".slider-handle")
    ActionChains(driver).click_and_hold(slider).move_by_offset(offset, 0).release().perform()

    return True

极验 v3 等标准滑块，优先走专用的 GeeTest v3 方法。

图像旋转验证码

# rotation_captcha.py


def solve_rotation_captcha(driver, captcha_selector):
    """Solve rotation CAPTCHA."""
    captcha = driver.find_element(By.CSS_SELECTOR, captcha_selector)
    image_b64 = captcha.screenshot_as_base64

    result = solve_custom_captcha(
        image_b64,
        "How many degrees should this image be rotated clockwise "
        "to be in the correct upright orientation? Return only the number."
    )

    try:
        degrees = int(result)
    except ValueError:
        return False

    # Click rotation button the correct number of times
    rotate_btn = driver.find_element(By.CSS_SELECTOR, ".rotate-button")
    clicks = degrees // 90  # Each click rotates 90 degrees

    for _ in range(clicks):
        rotate_btn.click()
        time.sleep(0.3)

    return True

按顺序点击验证码

# order_captcha.py


def solve_order_captcha(driver, captcha_selector, item_selector):
    """Solve click-in-order CAPTCHA."""
    captcha = driver.find_element(By.CSS_SELECTOR, captcha_selector)
    image_b64 = captcha.screenshot_as_base64

    result = solve_custom_captcha(
        image_b64,
        "What is the correct order? Return as comma-separated "
        "numbers (1-indexed) representing positions left-to-right, top-to-bottom."
    )

    # Parse order
    try:
        order = [int(x.strip()) for x in result.split(",")]
    except ValueError:
        return False

    # Click items in order
    items = driver.find_elements(By.CSS_SELECTOR, item_selector)
    for idx in order:
        if 1 <= idx <= len(items):
            items[idx - 1].click()
            time.sleep(0.5)

    return True

音频验证码

部分验证码提供音频备选：下载转 base64 提交，注明是音频并要求转写。

# audio_captcha.py
import requests


def solve_audio_captcha(audio_url):
    """Download and solve an audio CAPTCHA."""
    # Download audio
    resp = requests.get(audio_url, timeout=30)
    audio_b64 = base64.b64encode(resp.content).decode("ascii")

    # Submit as image with instructions
    # CaptchaAI may support audio via the base64 method
    result = solve_custom_captcha(
        audio_b64,
        "This is an audio CAPTCHA. Transcribe the spoken characters."
    )
    return result

自定义验证码小部件

碰上无法归类的小部件，按三步处理：

• 截图并抓取可见的提示文字；
• 连同截图提交识别；
• 结果写回输入框。

# custom_widget.py
from selenium import webdriver
from selenium.webdriver.common.by import By


def handle_custom_widget(driver, widget_selector):
    """Handle an unknown custom CAPTCHA widget."""

    # Step 1: Screenshot the entire widget
    widget = driver.find_element(By.CSS_SELECTOR, widget_selector)
    image_b64 = widget.screenshot_as_base64

    # Step 2: Get any visible instructions
    try:
        instructions_el = widget.find_element(By.CSS_SELECTOR, ".instructions, .prompt, p")
        visible_instructions = instructions_el.text
    except Exception:
        visible_instructions = "Solve this CAPTCHA"

    # Step 3: Submit with descriptive instructions
    result = solve_custom_captcha(
        image_b64,
        f"CAPTCHA instructions: {visible_instructions}. "
        f"Return the answer text."
    )

    # Step 4: Try to submit result
    try:
        input_el = widget.find_element(By.CSS_SELECTOR, "input")
        input_el.clear()
        input_el.send_keys(result)
    except Exception:
        # No input — try clicking based on result
        driver.execute_script("""
            var input = document.querySelector('input[name*="captcha"]');
            if (input) input.value = arguments[0];
        """, result)

    return result

页面提示文字一并提交，比硬编码更准。

自动检测验证码类型

有多个站点时，先扫 HTML 判断类型并路由。

# detector.py
import re


def detect_captcha_type(page_html):
    """Detect which CAPTCHA type is on a page."""
    checks = {
        "recaptcha_v2": r'data-sitekey.*g-recaptcha',
        "recaptcha_v3": r'recaptcha/api\.js\?render=',
        "turnstile": r'cf-turnstile|challenges\.cloudflare\.com/turnstile',
        "geetest": r'gt\b.*challenge|geetest',
        "bls": r'method.*bls|bls-captcha',
        "image_text": r'captcha.*\.(png|jpg|gif|jpeg)',
        "slider": r'slider.*captcha|slide.*verify',
        "audio": r'audio.*captcha|captcha.*audio',
    }

    detected = []
    for captcha_type, pattern in checks.items():
        if re.search(pattern, page_html, re.IGNORECASE):
            detected.append(captcha_type)

    return detected if detected else ["unknown"]

标准类型（reCAPTCHA、Turnstile、GeeTest v3）走各自专用方法；unknown 时才用截图方案。

合规提示：只在自有或已授权的站点操作，遵守 robots 协议与《个人信息保护法》。

常见问题排查

常见问题

滑块、旋转验证码识别成功率高吗？

取决于截图清晰度和说明精确度，站点差异很大；没有供应商能保证 100% 识别，请自行实测。

提交自定义验证码用哪个接口？

用 ocr.captchaai.com 的 in.php 提交、res.php 轮询，method 设为 base64，说明放进 textinstructions。

音频验证码能走图片接口吗？

可以先尝试：音频转 base64，用 base64 方法提交，说明里注明是音频。

自定义验证码怎么计费？

CaptchaAI 按并发线程计费，而非按次；套餐含当月不限次识别，如 BASIC（$15/月，5 线程），自定义类型不加价。

相关指南

• 多字符图片验证码识别策略
• Base64 编码最佳实践

把任何验证码交给识别接口 —— 免费开始使用 CaptchaAI。