VK (VKontakte) runs its own CAPTCHA system under the VK ID name. In 2026 the old image CAPTCHA (captcha_sid / captcha_img) gave way to a JavaScript-heavy "not a robot" widget that mixes a proof-of-work, browser fingerprinting and, at times, a slider. You meet it during login, sign-up, password recovery, and whenever the API returns error 14.
How to identifyLink to How to identify
- A page or iframe on
id.vk.comorid.vk.ruwith a path such as/not_robot_captcha - A
#not-robot-captcha-checkboxcheckbox - A large, obfuscated
not_robot_captcha.jsscript - A slider shown to suspicious sessions
- VK API error
error_code: 14witherror_msg: "Captcha needed",redirect_uriandremixstlid - The legacy
captcha_sidandcaptcha_imgfields, only in old integrations
How it worksLink to How it works
- The VK API or the login page returns a captcha error containing a
redirect_uriand a temporary user ID,remixstlid. Legacy clients may also getcaptcha_sidandcaptcha_img. - The client opens the VK ID captcha page, which runs
not_robot_captcha.js. - The script calls
captchaNotRobot.initSessionand thencaptchaNotRobot.check. Along the way it solves a proof-of-work and collects a FingerprintJSbrowser_fpplus device and network signals. - The proof-of-work is sent as
v2.followed by Base64(JSON{hash, nonce}), and the check request also carriesdebug_infoand interaction data. - When everything passes, VK returns a
success_token. - The original API call is then repeated with
success_tokenandremixstlid. The cookie and the parameter must agree. - The token is tied to the session, the browser fingerprint and the IP.
Why it is hardLink to Why it is hard
- The proof-of-work is just one layer. The browser fingerprint has to hold together too.
browser_fpis a FingerprintJS visitor ID built from canvas, WebGL, fonts and more. Plain HTTP can't fake it.- The script is heavily obfuscated and changes often.
- VK looks at
debug_infoand at request timing to spot automation. - Even a valid token is rejected if
remixstlidis missing or belongs to another session.
Free & open-source bypass methodsLink to Free & open-source bypass methods
| Tool | What it does | Success Rate | Link |
|---|---|---|---|
| Playwright + real Chromium | Drives the official widget, ticks the box and intercepts success_token |
60–85% | microsoft/playwright-python |
| Patchright / Camoufox | Hide CDP and headless traces from the FingerprintJS check | 50–80% | Kaliiiiiiiiii-Vinyzu/patchright-python · daijro/camoufox |
| vk-captcha-solver (npm) | A community solver for the VK ID captcha page | experimental | @tripoloski/vk-captcha-solver |
| DrissionPage | Browser automation with tighter control over the VK ID iframe | 50–75% | g1879/DrissionPage |
Practical approachLink to Practical approach
- Use a persistent browser profile and a residential IP in the target region.
- Trigger the captcha the normal way: login, sign-up or a controlled API call.
- Open the
redirect_uri, wait for#not-robot-captcha-checkbox, and click it along a Bezier mouse path. - If a slider follows, solve it with OpenCV template matching and a natural drag.
- Intercept the
captchaNotRobot.checkresponse to getsuccess_token. - Repeat the original request with
success_tokenand theremixstlidcookie and parameter. - Keep the same browser context, cookies, User-Agent and IP throughout.
Practical tipsLink to Practical tips
- Leave
debug_infountouched. A stripped or random value gives you away. initSessionmust run beforecheck. Skip it and you getstatus: BOT.- Hold on to the
remixstlidcookie from the captcha page, because the token alone is not enough. - Let a real browser produce the fingerprint. HTTP-only solvers would have to fully reverse-engineer the fingerprint payload.
- On the slider, the quality of the path matters as much as where the piece ends up.
- VK changes this flow frequently. Pinning a known-good script version is fine for research but not as a long-term setup.