recap

A Playwright-based reCAPTCHA v3 solver with runtime browser-fingerprint rotation and humanized interaction. It drives a real Chromium instance so Google's own reCAPTCHA script collects a genuine, internally consistent fingerprint, which is the only reliable way to earn a high v3 score.

---

How it works

reCAPTCHA v3 never shows a puzzle. It silently collects ~50 environment and behavioral signals, encrypts them, and posts them to Google, which returns a score from 0.0 (bot) to 1.0 (human). recap maximizes that score by making a real browser look like a real user:

generate fingerprint  ->  align timezone/geo to proxy  ->  launch Chromium
        ->  inject anti-detection init-script  ->  warm up browsing history
        ->  humanized mouse / scroll / focus  ->  grecaptcha.execute  ->  token

• Fingerprint rotation: correlated hardware profiles (GPU, cores, memory), resolution-matched DPR, locale-matched voices, screen, UA, battery.
• Anti-detection patches: canvas/WebGL noise, storage.estimate, enriched chrome.runtime, mediaDevices, all applied surgically to stay consistent with the signals reCAPTCHA actually hashes.
• Humanized behavior: bezier-curve mouse movement with entry sweep, decelerating scroll, real (trusted) clicks and focus, multi-step warmup.
• Proxy aware: timezone and geolocation are aligned to the proxy's IP.

---

Install

pip install -r requirements.txt
playwright install chromium

Requires Python 3.10+.

---

Usage

As a library

from recap import Recaptcha

token = Recaptcha(headless=True).solve(
    url="https://example.com/login",
    sitekey="6LcAbwIqAAAAAJvVAhSSJ8qzYsujc7kn1knmSgQX",
    action="submit",
)
print(token)

One-shot convenience wrapper:

from recap import solve

token = solve(
    url="https://example.com/login",
    sitekey="6Lc...",
    action="submit",
    proxy="http://user:pass@host:port",   # optional
)

Options

Option	Default	Description
proxy	None	http://user:pass@host:port (timezone/geo align to it)
headless	True	Run Chromium headless
warmup	True	Build a short browsing history before the target
debug	False	Print per-stage timing
user_data_dir	None	Persistent profile; cookies/trust accumulate over runs
enterprise	False	Use the enterprise grecaptcha API

Tip: a residential proxy plus a persistent user_data_dir raises the score the most, because Google's trust cookies build up across sessions.

As an HTTP server

python server.py            # listens on 0.0.0.0:8778

Synchronous solve:

curl -X POST http://localhost:8778/solve \
  -H "Content-Type: application/json" \
  -d '{"url":"https://example.com","sitekey":"6Lc...","action":"submit"}'

Async job:

# enqueue -> returns {"job_id": "..."}
curl -X POST http://localhost:8778/jobs -H "Content-Type: application/json" \
  -d '{"url":"https://example.com","sitekey":"6Lc..."}'

# poll
curl http://localhost:8778/jobs/<job_id>

Endpoint	Method	Description
/solve	POST	Synchronous solve, returns token
/jobs	POST	Enqueue an async solve
/jobs/{id}	GET	Poll job state
/jobs/{id}	DELETE	Drop a finished job

Hosted endpoint

A live instance is available. You can use this endpoint to solve without running the server yourself:

import requests

session = requests.Session()

solver_url = "https://recapv3.hainahola.com/solve"
solver_data = {
    "url": "https://recaptcha-test.hainahola.com/",
    "sitekey": "6Lfmnz8tAAAAAP06n7wf3K3e5a6-rcTfqQ5iAdd0",
    "action": "submit",
    "timeout": 120,
}

result = session.post(solver_url, json=solver_data, timeout=150).json()
print("SOLVE RESULT:", result)
token = result.get("token")
if not token:
    raise SystemExit(f"solve failed: {result.get('error')}")
print("TOKEN:", token)

verify = session.post(
    "https://recaptcha-test.hainahola.com/verify",
    json={"token": token},
    timeout=15,
).json()
print("VERIFY:", verify)
print("SCORE:", verify.get("score"))

---

Project structure

Recaptchav3/
├── server.py            FastAPI front-end (sync + async jobs)
├── requirements.txt
└── recap/
    ├── engine.py        Recaptcha class + solve(): the browser flow
    ├── fingerprint.py   runtime fingerprint generation
    ├── patches.py       anti-detection init-script
    ├── location.py      proxy IP -> timezone/geo lookup
    └── network.py       proxy string parsing

---

Contributing

Pull requests are welcome. To keep the solver consistent (reCAPTCHA punishes inconsistency far more than any single "bad" value), please:

1. Fork the repo and create a feature branch.
2. Keep changes surgical: touch only what's needed, and make sure any fingerprint/patch change stays internally consistent (e.g. GPU must match the UA, deviceMemory must stay within spec).
3. In your PR description, explain what you changed and why: which signal it affects and how you verified it doesn't create a detectable mismatch.

---

Issues & community

Found a bug, a detection tell, or a score regression? Open an issue with the details, or join the Discord:

---

Disclaimer

This project is for educational and research purposes. You are responsible for complying with the terms of service of any site you interact with.