pixiv seems not allowing username/password login now

[roytam1]

server returning JSON:

{"has_error":true,"errors":{"system":{"message":"The grant type is unauthorized for this client_id","code":1508}},"error":"invalid_grant"}

[roytam1]

background information:

In pixiv android Version 5.0.234 in 2021-02-08, there is a line:

• From the beginning of February 2021, you'll have to update your app to the latest version if you're running ver.5.0.219 or earlier.

and for testing, I tried to install 5.0.220.1 apk and tried to login, and it pops out custom browser window and jumps to pixiv standard login form.

[roytam1]

a cross reference is found: #140 (comment)

[kilosonc]

I encountered the same problem

HTTP 400: {"has_error":true,"errors":{"system":{"message":"Invalid grant_type parameter or parameter missing","code":1508}},"error":"invalid_grant"}

I want to use set_auth, but I don't know which is access token, there're to many pairs in cookie

[roytam1]

I want to use set_auth, but I don't know which is access token, there're to many pairs in cookie

they're not exist in cookies or IndexedDB or Local Storage.

[evazion]

In my project I had to revert back to using the Ajax API. The problem is that the grant_type=password authentication method is no longer supported. The Pixiv app now logs in through https://accounts.pixiv.net/login, but this page is protected by Google reCAPTCHA, which seems impossible to circumvent.

I think that as long as you have a refresh token, you can keep using it, but getting one is tricky since you have to login with the mobile app and sniff traffic to get it.

https://github.com/danbooru/danbooru/blob/39cc3ed5cf913499093d2f641d70d7682a14fa42/app/logical/pixiv_ajax_client.rb

[emesh0620]

If password authentication is disabled, is there no choice but to scrape with BeautifulSoup from now on? Unbelievable

[Mouwoo]

经测试该账号和密码可以pixiv官网正常登录

[emesh0620]

Last year, you said that you are trying not to register with a leaked password, but you are also considering methods such as multi-factor authentication, so it is time to think about some fundamental measures. It may be. (Link to Japanese)

https://inside.pixiv.blog/2020/01/22/180000

[mrwan2546]

Pixiv was changed grant_type password to authorization_code. But it's so hard to login because pixiv using login with OAuth 2.0

[ZipFile]

Pixiv now follows RFC7636. We won't be able to use (old) password login again, but there is a way to automate token retrieval without traffic sniffing. I've implemented a simple script to retrieve access/refresh tokens. Right now it requires single manual step inspecting dev console, but in general, entire flow could be fully automated using browser extensions: you can add request handler and extract code param from the url.

[reimu1234]

Update the semi-automatic script, get the code through chromedriver

更新了半自动脚本，通过chromedriver获取code

https://gist.github.com/upbit/6edda27cb1644e94183291109b8a5fde

请问手机端能拿到吗 pc 的试了拿到code后就过不去了 开了全局

[upbit]

Update the semi-automatic script, get the code through chromedriver
更新了半自动脚本，通过chromedriver获取code
https://gist.github.com/upbit/6edda27cb1644e94183291109b8a5fde

请问手机端能拿到吗 pc 的试了拿到code后就过不去了 开了全局

如果能输出code，可以参考requests的代理配置，对95行的post请求增加代理

[reimu1234]

Update the semi-automatic script, get the code through chromedriver
更新了半自动脚本，通过chromedriver获取code
https://gist.github.com/upbit/6edda27cb1644e94183291109b8a5fde

请问手机端能拿到吗 pc 的试了拿到code后就过不去了 开了全局

如果能输出code，可以参考requests的代理配置，对95行的post请求增加代理

都试了才来打扰😂

[upbit]

试试换个代理？这个和ZipFile的原始脚本一样，都用requests直接请求的。或者试下 Mapaler 的GUI工具

ps: 手机端没试过，如果你用term跑python应该可以；纯浏览器会打开app，无法拦截code

[Mapaler]

@yzwik 手机端用浏览器获取code会打开P站APP，除非你把P站APP卸载掉，自己写一个注册pixiv协议的APP来获取code。
获取的登陆code只有几十秒的生存时间，需要尽快使用，不然就过期了。你电脑获取的话，需要马上转移到手机上使用。
或者你可以电脑上登陆了后，把刷新code拿去直接添加刷新code，刷新code的生存时间非常长。

[txperl]

Update the semi-automatic script, get the code through chromedriver
更新了半自动脚本，通过chromedriver获取code
https://gist.github.com/upbit/6edda27cb1644e94183291109b8a5fde

请问手机端能拿到吗 pc 的试了拿到code后就过不去了 开了全局

如果能输出code，可以参考requests的代理配置，对95行的post请求增加代理

都试了才来打扰😂

可以通过 pixivpy bypass SNI 的方式，先获取 API 的真实 IP，然后直接请求，就不需要代理了。

参考脚本：
https://github.com/txperl/PixivBiu/blob/24119157facdef4eb31ab6492eb32b583d01c7bf/app/core/biu/login_token.py

参考实现：
https://github.com/txperl/PixivBiu/blob/24119157facdef4eb31ab6492eb32b583d01c7bf/app/core/biu/main.py#L164

[reimu1234]

试试换个代理？这个和ZipFile的原始脚本一样，都用requests直接请求的。或者试下 Mapaler 的GUI工具

ps: 手机端没试过，如果你用term跑python应该可以；纯浏览器会打开app，无法拦截code

确实是代理的问题 换了个节点解决😂 感谢大佬帮助

[TianyiShi2001]

如果是有reCAPTCHA的话那就不可能全自动了吧

不过cookie不是太快过期的话倒也不太影响使用

[eggplants]

https://gist.github.com/upbit/6edda27cb1644e94183291109b8a5fde

I have written a library/CLI based on this script to get tokens headlessly in selenium.
Check: https://github.com/eggplants/get-pixivpy-token