Add ability to inject JavaScript

[kuhnchris]

Description

This is the 2nd attempt of PR #137 - ability to inject javascript into a websource

Motivation and Context

87

How Has This Been Tested?

Tested locally, also some other users who require the functionality tested the change.

Types of changes

• New feature (non-breaking change which adds functionality)

Checklist:

• [X] My code has been run through clang-format.
• [X] I have read the contributing document.
• [X] My code is not on the master branch.
• [X] The code has been tested.
• [ ] All commit messages are properly formatted and commits squashed where appropriate.
• [ ] I have included updates to all appropriate documentation.

[kuhnchris]

@RytoEX FYI, the ubuntu workflow is using clang12, I used clang13 - I took the liberty to downgrade to clang12 for this, but please make sure to let us know if clang12 is fixed right now - thanks.

[RytoEX]

@RytoEX FYI, the ubuntu workflow is using clang12, I used clang13 - I took the liberty to downgrade to clang12 for this, but please make sure to let us know if clang12 is fixed right now - thanks.

We require using clang-format 12.0.0 at the moment because that's what our CI workflows use. The formatted output produced by clang-format changes between any given version. We plan to update the clang-format version requirement, hopefully within the next few months.

[kuhnchris]

I would prefer if this setting's boolean state was logged. I don't think we need to log the full JavaScript string, but I would want to know that a user enabled it on a source and if the JavaScript string is empty or not.

How would you like to log it, just as a string somewhere "The user activated javascript in the browser source %s" ?

[kuhnchris]

Added the requested logging, change in logic and sorting.

[willlllllio]

Any chance this makes it into OBS 28? 🤔

Ideally with injecting in OnLoadStart instead of OnLoadEnd to be able to reliably hook/change browser APIs before any scripts on the page run, like mentioned on the original thread.

[kuhnchris]

Any chance this makes it into OBS 28? thinking

Ideally with injecting in OnLoadStart instead of OnLoadEnd to be able to reliably hook/change browser APIs before any scripts on the page run, like mentioned on the original thread.

I'd love to, but at this point I gave up. I know a couple of users simply compiled this themselves, but aside from that I guess the main way of communicating would be to go to one of the chatting platforms (i guess discord) and poke relevant people.

[ScriptPup]

Any chance there's a pre-built version with the injection PR or cheat-sheet somewhere for how-to somewhere? Or maybe even an add-on version with an alternate browser with the feature? I see in the README that I can't build the browser as a standalone and it has to be part of the OBS studio build. I'm not at all a C++ dev and have been banging my head on the wall trying to figure out how to build with this feature but haven't gotten anywhere...

I was super excited when I searched around for this and found someone had already solved it... Then super sad to find that it's apparently just been stonewalled for what appears to be like 4 years?? I can totally understand wanting to have plenty of warnings for users that they can expose themselves if they don't know what they're doing with custom JS... But why not at least have the option when the work's already been done?

[kuhnchris]

Well, I did talk to the people over at Discord regarding this topic quite a while ago and actually was asked to hold my horses until they updated to a newer CEF (Chrome Embedded Framework) version as the sandbox changed quite a bit. As there still isn't any update on this, I might aswell try to build it again. For what OS do you need this? Windows, Linux, MacOS-X?

[ScriptPup]

Wow that would be fantastic! I'm using windows.

Thanks!!

---

From: KuhnChris @.> Sent: Friday, September 30, 2022, 4:39 AM To: obsproject/obs-browser @.> Cc: scriptpup @.>; Comment @.> Subject: Re: [obsproject/obs-browser] #137 Try #2 - Injecting Javascript (PR #354)

Well, I did talk to the people over at Discord regarding this topic quite a while ago and actually was asked to hold my horses until they updated to a newer CEF (Chrome Embedded Framework) version as the sandbox changed quite a bit. As there still isn't any update on this, I might aswell try to build it again. For what OS do you need this? Windows, Linux, MacOS-X?

— Reply to this email directly, view it on GitHubhttps://github.com/obsproject/obs-browser/pull/354#issuecomment-1263283674, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ABPAZPU36RAPG66PP2P3PL3WA2RLLANCNFSM5PKN66JA. You are receiving this because you commented.Message ID: @.***>

[kuhnchris]

Wow that would be fantastic! I'm using windows. Thanks!! […]

Usually I loathe to distribute binaries, as one can never be sure that they are safe, but since it was requested a couple of times: obsBrowserWithJavascript-OBS-28.0.2.zip This is basically the Release/ directory after building it in VS Studio 2022 Community Edition. I tested it against the current 28.0.2 Release of OBS from here on github, and it works as expected. Reminder: Activate the "Javascript Active" checkbox below the javascript input text box, and, for any javascript error, make sure to check the logs. Logs can be found here:

To recreate this, simply follow the entire build process outlined here: https://obsproject.com/wiki/build-instructions-for-windows And, before compiling everything, replace the plugins/obs-browser folder with my fork.

Let me know if that helps you! Cheers

[Fenrirthviti]

Just as a note, we definitely haven't forgotten about this and the request that users have for it.

We have made steps towards properly supporting this safely with the recent CEF update, but keep in mind that it is our opinion that injecting arbitrary JS is very risky, and we're not 100% sold on supporting it at this point, unless it can be done very safely.

We haven't had time to get around to experimenting with the CEF sanboxing feature, which I feel comfortable saying is going to be bare minimum requirement for this to be merged and released by us.

[willlllllio]

our opinion that injecting arbitrary JS is very risky, and we're not 100% sold on supporting it at this point, unless it can be done very safely

Doesn't injected JS just run in the same JS context as the page it's injected in, and just the same as any normal browser source page in general? What's the difference between a user injecting JS from somewhere on the internet vs setting a browser source url from somewhere on the internet, security wise? Apart from the added possibility of lifting tokens/info from the page isn't that basically the same?

And I get the concern about that but even every actual browser I'm aware of lets users open a dev console and do whatever, and the general workarounds I've seen for people needing to inject JS into OBS web sources were either using --remote-debugging-port and injecting via that or installing a modified browser source binary from random people, both respectively more insecure and much much much more insecure than this. And even in general the usual OBS plugin process is just installing dlls/libs or lua/py scripts from the internet, also both significantly more dangerous.

And I don't mean to be rude here sorry if I come off like that, I get you're all very busy and have way more pressing issues and features on the docket than JS injection into browser sources, but it's just the seemingly large concern about the security of this specific feature I don't really get in the context of everything else in OBS.

[ScriptPup]

Thank you very very much @kuhnchris! The binaries work perfectly. I understand your point about it being better to build from source... This is definitely an area I'm lacking in and need to learn how to properly build C++ from source myself. That being said, for the current problem I was facing the plugins you posted work perfectly! 🙏

@Fenrirthviti I'm glad to hear it's not dead! This would be a great out-of-the-box feature, and I also totally understand not wanting to release something that the owning team isn't 100% confident in (especially if it's related to security). I don't think I've ever gotten replies from maintainers or even thread creators this quickly before - This community is clearly fantastic!

Thank you again internet strangers for helping me solve my problem, although you owed me absolutely nothing :)

[kuhnchris]

Just as a note, we definitely haven't forgotten about this and the request that users have for it.

We have made steps towards properly supporting this safely with the recent CEF update, but keep in mind that it is our opinion that injecting arbitrary JS is very risky, and we're not 100% sold on supporting it at this point, unless it can be done very safely.

We haven't had time to get around to experimenting with the CEF sanboxing feature, which I feel comfortable saying is going to be bare minimum requirement for this to be merged and released by us.

Thanks for that info, I haven't really been following around. Regarding CEF Sandboxing: is there an open PR for that? Checking the GitLab real quick seems to makes this appear as a real easy implementation, seeing https://bitbucket.org/chromiumembedded/cef/wiki/SandboxSetup - basically having the new binary dropped in place (kind of like the obs_browser_page(.exe)) and initializing CEFSandbox via cef_sandbox_info() and chugging that pointer into the CEFInitialize over at obs-browser-plugin.cpp (and probabbly remove the no_sandbox property). Is there any open PR for that or is it being worked on internally?

Thanks again and all the best, Chris

[Fenrirthviti]

There is no open PR, and we haven't begun internal testing. Enabling it isn't the issue, it's testing all the use cases of browser source currently to ensure there are no unintended side effects that is the blocker.

[kuhnchris]

There is no open PR, and we haven't begun internal testing. Enabling it isn't the issue, it's testing all the use cases of browser source currently to ensure there are no unintended side effects that is the blocker.

Pardon my naivity, but, if there is such a focus on use cases, why isn't there a unit-test/test framework in place that covers them? Is there any way we could help collect the required use-cases? If so, we could technically help with creating those for you guys, especially if "implementing isn't the issue". I haven't seen any roadmap or anything regarding this, so the community is kind of in the air, but it could be me not looking in the right place to begin with, so I'll excuse myself before-hand.

[Fenrirthviti]

The answer to all of that is the same: Nobody has spent time on it. The reasons for that are varied and numerous. As with most open-source projects, people tend to work on things that interest them. With only a single full time developer on OBS as of the writing of this message, we need to be careful with how we assign that resource, and we have no direct control over what anyone else on the project wants to work on. Every new feature, every pull request, and every suggestion is the most important thing to the person who wants it. We have to weight and prioritize based on things that help as much of the majority of the userbase as we can, which is not easy.

If you, or anyone else, are offering to put that work in to help push this forward, I'm sure it would be welcomed.

[kuhnchris]

The answer to all of that is the same: Nobody has spent time on it. The reasons for that are varied and numerous. As with most open-source projects, people tend to work on things that interest them. With only a single full time developer on OBS as of the writing of this message, we need to be careful with how we assign that resource, and we have no direct control over what anyone else on the project wants to work on. Every new feature, every pull request, and every suggestion is the most important thing to the person who wants it. We have to weight and prioritize based on things that help as much of the majority of the userbase as we can, which is not easy.

If you, or anyone else, are offering to put that work in to help push this forward, I'm sure it would be welcomed.

Most definitly, since I'm not the only "active" contribute over here, it would make sense to channel these efforts. People do want this plugin to work, and not just feature, but many more features. If we can help with some ground work I'm sure we could take off some "unloved and pushed aside" tasks like creating tests or managing a testplan, or, even just for the heck of it, just hosting a kanban board and throwing in what actually is on the table and cannot be worked at. At this time I think, yes, there is/was only 1 person working on the plugin, and there are volunteers who would like to help (yes I even count myself as one of them), but we are kind of lost where to begin or where to help, to be quite frank. Is there anyone we can directly go to, or should we use discord or some other chat medium to get issues like this moved quicker? Especially since it obviously is on some kind of internal time plan, but the exposing to the community is a bit lacking. Again, no critism, just trying to help out, especially since this PR has been floating around since... checks clock 2018, and the original request (#87) floating around as long as August 2017. We definitly would love to help you guys, but are kind of lacking the communication to do so.

[RytoEX]

Things are getting cluttered here, and I wouldn't want any reviewer having to read through a bunch of comments about project governance or priorities. If there are any questions about project governance or priorities, please reach out to us on our Discord. Myself (RytoEX#5545) or @Fenrirthviti (Len#9999) can answer any questions about project priorities, how things get decided, or what the best path forward is for tackling a specific item.

Moving forward, let's keep discussion focused on the actual code and concerns that we've previously raised. On a high level, something that is required for us to even consider merging this is making sure that we can enable the CEF/Chromium sandbox on all platforms (Windows, macOS, Linux/*nix) without breaking any functionality (service integration, browser docks/panels, browser sources, alerts, local files, locally served web pages, etc.). You, or anyone else, are welcome to investigate and try to tackle those tasks in a separate PR, or in multiple PRs if necessary for scope/testing. Do not try to tack on the sandbox to this PR. Note, that this is not a guarantee that we will simply merge any PR that appears and claims to solve those problems. We would need to see a demonstration that the PR both addresses our requirement (enabling the sandbox) and does not break anything (show that you've done the testing, ideally by providing test files/cases that we can verify).

Now that we've touched on project governance and priorities and our prerequisites for this specific item... To answer the specifics of why we delayed looking into this, there were major architectural changes that we wanted to address first before revisiting CEF sandboxing (see: obsproject/obs-studio#4560, obsproject/obs-studio#5155, and obsproject/obs-browser#290) as well as updating CEF (first from CEF 75/3770 to 95/4638; then from 95/4638 to 103/5060; see obsproject/obs-browser#323). Hopefully, that answers the bulk of the questions of "what is this waiting on" from our side of things.

On the specifics of this PR, you would need to rebase the changes on to the current master branch of this repository, for starters. As-is, this is not mergeable.

Finally, please keep comments on this PR on-topic to the specifics of this PR and its code and implementation requirements/details. If this PR devolves into off-topic discussion, it will be locked.

[kuhnchris]

Things are getting cluttered here, and I wouldn't want any reviewer having to read through a bunch of comments about project governance or priorities. If there are any questions about project governance or priorities, please reach out to us on our Discord. Myself (RytoEX#5545) or @Fenrirthviti (Len#9999) can answer any questions about project priorities, how things get decided, or what the best path forward is for tackling a specific item.

Moving forward, let's keep discussion focused on the actual code and concerns that we've previously raised. On a high level, something that is required for us to even consider merging this is making sure that we can enable the CEF/Chromium sandbox on all platforms (Windows, macOS, Linux/*nix) without breaking any functionality (service integration, browser docks/panels, browser sources, alerts, local files, locally served web pages, etc.). You, or anyone else, are welcome to investigate and try to tackle those tasks in a separate PR, or in multiple PRs if necessary for scope/testing. Do not try to tack on the sandbox to this PR. Note, that this is not a guarantee that we will simply merge any PR that appears and claims to solve those problems. We would need to see a demonstration that the PR both addresses our requirement (enabling the sandbox) and does not break anything (show that you've done the testing, ideally by providing test files/cases that we can verify).

Now that we've touched on project governance and priorities and our prerequisites for this specific item... To answer the specifics of why we delayed looking into this, there were major architectural changes that we wanted to address first before revisiting CEF sandboxing (see: obsproject/obs-studio#4560, obsproject/obs-studio#5155, and #290) as well as updating CEF (first from CEF 75/3770 to 95/4638; then from 95/4638 to 103/5060; see #323). Hopefully, that answers the bulk of the questions of "what is this waiting on" from our side of things.

On the specifics of this PR, you would need to rebase the changes on to the current master branch of this repository, for starters. As-is, this is not mergeable.

Finally, please keep comments on this PR on-topic to the specifics of this PR and its code and implementation requirements/details. If this PR devolves into off-topic discussion, it will be locked.

Thanks for clarifying - let's get back on topic with this PR then, separate discussions somewhere else, I agree.

Ad clarifying the sandboxing/CEF upgrade issue: OK, thanks, that makes a lot more sense now.

Ad this PR: Wait, it isn't mergable? I did merge current master into javascript yesterday when building it, but maybe I did something wrong? The only thing I see broken here is the clang fix - please advise if I miss something here.

Again, it wasn't my intend to off-topic the discussion here - so let's keep it at the javascript injection topic.

[RytoEX]

Ad this PR: Wait, it isn't mergable? I did merge current master into javascript yesterday when building it, but maybe I did something wrong? The only thing I see broken here is the clang fix - please advise if I miss something here.

I currently see this PR attempting to merge 43 commits.

This is likely caused by this merge commit:

Do not use merge commits to update branches. Rebase the "javascript" branch on the current obs-browser master branch. Additionally, squash the changes in to one single commit.

[michael-dm]

Hello ! I know you guys are super busy but do you think this will be merged into next OBS version ? 😄

[kuhnchris]

I assume not, but I will get to rebasing the changes to master hopefully this weekend.

[pkviet]

you didn't rebase properly; as a result there are 43 (!!!) commits ... Please correct this, thank you.

[kuhnchris]

Here we go, single commit, squashed. Feel free to try to compile it.

Edit, also @pkviet no idea where you got 43 commits, but I assume you may have based it from the first commit, but it should be a single one commit now.

[pkviet]

See above the comments from @RytoEX You used a merge commit which added earlier commits. It's now fixed, thanks.

[kuhnchris]

See above the comments from @RytoEX You used a merge commit which added earlier commits. It's now fixed, thanks.

oh that's what you meant, well, I squashed them, so it should be fine, you're welcome.

[kuhnchris]

@Warchamp7 so we should close/discard this PR then too, since you closed #87 ?

[TezStream]

Wow that would be fantastic! I'm using windows. Thanks!! […]

Usually I loathe to distribute binaries, as one can never be sure that they are safe, but since it was requested a couple of times: obsBrowserWithJavascript-OBS-28.0.2.zip This is basically the Release/ directory after building it in VS Studio 2022 Community Edition. I tested it against the current 28.0.2 Release of OBS from here on github, and it works as expected. Reminder: Activate the "Javascript Active" checkbox below the javascript input text box, and, for any javascript error, make sure to check the logs. Logs can be found here:

To recreate this, simply follow the entire build process outlined here: https://obsproject.com/wiki/build-instructions-for-windows And, before compiling everything, replace the plugins/obs-browser folder with my fork.

Let me know if that helps you! Cheers

Hello, if this possible to use this obs obs 27.2.4 ? this is working fine on 28 but i use 27 please anyone share link for obs 27, Thanks

[Fenrirthviti]

Closing this per comments in #87

[Dav-Edward]

Would it be possible to get this as an OBS extension then? I and a lot of people really need this functionality and using the Chrome Debugger mode is outright silly and impractical for doing daily streams having to manage browsers to tweak things every time.

With more and more apps denying us the ability to make our own things such as Discord Stream Kit still having the Discord RPC API locked away from the public and their terrible HTML design making it impossible for CSS overrides to do the job any more, it's becoming impossible to do anything with Browser Sources as they are.

If it's a separate extension we can install into OBS then would that satisfy the "You are really sure you what to do this" requirement?

Please, I really don't want to have to go to Stream Labs for this one missing feature.

[Warchamp7]

@Dav-Edward Do you have an example use-case? It is my belief that there are far too many security concerns with arbitrary (and potentially silent) javascript injection to outweigh the potential uses. There are certainly ways that a plugin could achieve this behaviour, but I do not see us supporting it out of the box.

Please, I really don't want to have to go to Stream Labs for this one missing feature.

I'm not quite sure I understand how Streamlabs affects this? Streamlabs Desktop is an OBS fork and has the exact same limitations for Browser Sources.

[Dav-Edward]

@Dav-Edward Do you have an example use-case? It is my belief that there are far too many security concerns with arbitrary (and potentially silent) javascript injection to outweigh the potential uses. There are certainly ways that a plugin could achieve this behaviour, but I do not see us supporting it out of the box.

Please, I really don't want to have to go to Stream Labs for this one missing feature.

I'm not quite sure I understand how Streamlabs affects this? Streamlabs Desktop is an OBS fork and has the exact same limitations for Browser Sources.

I thought Stream Lab's version did have the ability to inject Javascript into their Browser Sources based on comments in one of the duplicate pull requests.

As for use case, absolutely I do.

Discord Stream Kit ( https://discord.com/streamkit ) used to be resonably moddable using CSS for people to do things like rearrange positions of user avatars, set custom avatars for the users, and various other features until two things happened. First they optimized the website using some code optimizer which stripped away all ID tags on objects in the page making specific user references impossible so you can't override specific users with specific avatars or positions, and they enabled Strict-XSS which means it's now impossible to do any image overrides as the web page won't let you reference images outside of the Discord domain.

Discord more than 4 years ago released https://discord.com/developers/docs/topics/rpc but then ripped it away again shortly thereafter from the public meaning there is absolutely no way to make your own version of the Discord Stream Kit by writing your own website as you cannot get access to the Discord RPC APIs.

Only one known developer, https://reactive.fugi.tech ever got access to the Discord RPC APIs by grandfathering in and no one else has been able to make an alternative to Discord Stream Kit other than FugiTech. Those two are litterally the only two choices available for any app of any kind to monitor voice activity in Discord voice chat for the purposes of OBS overlays.

All kinds of Twitch\YouTube\etc. VTuber streamers rely on Discord Streamkit to show voice activity of users in their Discord during streams. But now that Stream Kit is so locked down and limited, people cannot set custom avatars for users appearing in their Discord which poses problems as streamers can't manage what images people might set as their Discord avatar which can cause a TOS violation on the streaming platform of a user enters the voice chat with an explicit avatar.

Twitch streamers like myself heavily used the CSS override feature to make it so users could have custom avatars that appear during streams, and strangers would be given a safe, default avatar to represent themselves when they enter the voice chat. It alow allowed for all kinds of creativity for example being able to reposition avatars so that users avatars 'fit together' to make up a scene when in voice chat.

The only way we can get Discord Streamkit to be even remotly useable these days is by starting OBS into Remote Debugging mode and firing up Chome or Firefox to then inject Javascript overrides from the web browser which is a very tedious process every single time we start our streams having to run the overrides on every relevant browser source in OBS.

Attached is an example of what used to be possible to accomplish with CSS but now requires direct intervention with Javascript to accomplish.

If security is the concern, this being made a separate plugin would be perfectly acceptable as a solution. The attack surface would be minimal as only users specifically meaning to use the feature would go out of their way to install the plugin or turn on the functionality.

I and many, many, streamers I know need this functionality as right now FugiTech is our absolute last option now and it's very limited in what modifications you can do as it much like Discord Stream Kit does not use ID references in HTML tags so you can't do anything useful with CSS. FugiTech uses Canvas and Blob objects making CSS useless to attempt to modify or reposition avatars.

[Warchamp7]

I think the real solution to this particular case is petitioning Fugi to add more customizability or Discord to amend their classes to be more friendly to CSS modifications rather than JS injection. That is a particularly overkill solution as I noted previously.

That being said, modifying the Discord Stream Kit via CSS only is still perfectly doable.

Example CSS:

.Voice_voiceState__OCoZh {
    position: relative;
}

/* Hide users default avatars */
img.Voice_avatar__htiqH {
    opacity: 0;
}

/* Default Avatar to show for anyone */
.Voice_user__8fGwX:before {
    content: "";
    background: white url('https://i.imgur.com/JKLwyDg.png') no-repeat;
    background-size: cover;
    border-radius: 100%;
    height: 50px;
    margin-right: 8px;
    width: 50px;
    display: block;
    z-index: 1;
    position: absolute;
    top: 3px;
    left: 3px;
}

/* Avatar Override based on Discord ID of their avatar */
.Voice_avatar__htiqH[src^="https://cdn.discordapp.com/avatars/60593627616784384/"] ~ .Voice_user__8fGwX:before {
    background-image: url('https://i.imgur.com/ii6dsFf.png');
}

/* Override styling specifically when this user is talking */
.Voice_avatarSpeaking__lE\+4m.Voice_avatar__htiqH[src^="https://cdn.discordapp.com/avatars/60593627616784384/"] ~ .Voice_user__8fGwX:before {
    filter: hue-rotate(90deg);
}

In the above case I've hidden peoples normal avatars and created a fake default one using your GitHub profile icon. I've set an override for myself specifically to use my own icon, and made it hue-shifted when I'm talking resulting in the below appearance

Notes: I've censored the peoples names in Photoshop after the fact, the the example CSS links to Imgur as GitHub does not allow hotlinking

[Dav-Edward]

@Warchamp7 There is absolutely no way we can strongarm Discord to fix their website, there has been hundreds of upvotes on the Discord feedback website for them to change the Stream Kit website back and they have not, and asking FugiTech to add a ton of additional features to their website is not sensible.

I'm both impressed and confused how you managed to do what you did. Ever since Discord changed their Stream Kit website to remove the reference IDs we completely lost the ability to hide Flexboxes, Move/Reposition Flexboxes, and override avatars properly.

Before the change we used to be able to write code like this to control the position, animation, and content of the flexboxes:

img.avatar.speaking[data-reactid*="86586182175035392"] { animation-name: Dav-speak!important; animation-duration: 0.8s!important; animation-direction:normal; animation-iteration-count: infinite; }

@keyframes Dav-speak { 0% { transform:translate(0px,0px); } 10% { transform:translate(0px,10px); } 20% { transform:translate(0px,0px); } 50% { transform:translate(0px,-60px); } 100% { transform:translate(0px,0px); } }

.avatar[data-reactid*="86586182175035392"] { content:url(https://i.imgur.com/JKLwyDg.png); border:none!important; bottom:15px!important; border-radius:0!important; height:200px!important; width:200px!important;} /* Davwyn */
.avatar.speaking[data-reactid*="86586182175035392"] { content:url(https://i.imgur.com/JKLwyDg.png); border:none!important; bottom:15px!important; border-radius:0!important; height:200px!important; width:200px!important;}

Then, after their first set of changes, we completely lost the ability to target the flexboxes and had to resort to only changing the images this way:

li>img[class^="Voice_avatar"][src*="86586182175035392"] { content:url(https://i.imgur.com/JKLwyDg.png); border:none!important; bottom:15px!important; border-radius:0!important; height:200px!important; width:200px!important;} /* Davwyn */
li>img[class*="Speaking"][src*="86586182175035392"] { content:url(https://i.imgur.com/JKLwyDg.png); border:none!important; bottom:15px!important; border-radius:0!important; height:200px!important; width:200px!important;}

One part that has me very confused is in your example, were you actually able to reference images OUTSIDE of Discord?? They have Strict-XSS enabled. When we tried to reference external resorces before, the external resources got shot down. How did you do that?

I used to have the script refer to a repo I managed with a good old @import url("https://example.com/StreamBeans.css") method so that my streamer group always had the same current script but that stopped working.

Is it actually plausible to still be able to: -Hide entire flexboxes (not just images, we don't want an empty gap between avatars -Reposition flexboes (like my first script example before Discord broke it) -Change avatar images to externally stored images -Potentially load CSS stored on our resource server

That would be preferable than us having to fire up Chrome Remote Debugging for every single streamer and cracking open the Developer console every time. Javascript access does give us the ability to also potentially hook into states the website has but doesn't use like 'User Self-Muted' and 'User Self-Deafended' states which the website oddly doesn't actually use despite the API calls being in there.

[Dav-Edward]

@Warchamp7 also, if there is a way to resolve this. I will gladly spread the word, as in all sorts of forums, Discords, and Reddit VTubers are streaming into the void beating their heads against the wall trying to get back the ability to override avatars and flexbox locations/visbility.

I'll let everyone know far and wide your solution with credit if this is actually possible. The only way I knew of to target the flexboxes is as follows:

li:has(img[src*="86586182175035392"]) { display:block!important; z-index:50!important; order:50!important; }

However, OBS does not support the :has statement. I asked about it here a while back:

https://www.reddit.com/r/obs/comments/zxr3j1/has_css_support_for_more_advanced_custom_css/

[Warchamp7]

One part that has me very confused is in your example, were you actually able to reference images OUTSIDE of Discord?? They have Strict-XSS enabled. When we tried to reference external resorces before, the external resources got shot down. How did you do that?

Yes my example is tested and working in both Chrome and OBS itself. It's possible they reverted that restriction. If it was re-enabled again, another potential workaround would be to upload the icons to Discord and use the attachment URL. I doubt they would block a subset of their own CDN. A cumbersome workaround, sure, but again doable without JS injection.

-Hide entire flexboxes (not just images, we don't want an empty gap between avatars

As far as I can tell there are no flex box elements in the current version of Discord's Stream Kit, but realistically you could hide any element with an appropriate CSS selector, yes.

-Reposition flexboes (like my first script example before Discord broke it)

As above, there's no flexboxes (Is this simply your understanding of a DOM element?) but yes you could reposition things

-Change avatar images to externally stored images

Based off the above example yes. Plus the potential workaround.

-Potentially load CSS stored on our resource server

I'm not sure why that would be needed unless you are frequently updating this CSS and constantly having to update it on the browser source. In that case it would be possible to write a Lua script or obs-websocket script that could fetch the CSS and update the Browser Sources properties with the new Custom CSS.

I'll let everyone know far and wide your solution with credit if this is actually possible. The only way I knew of to target the flexboxes is as follows:

li:has(img[src*="86586182175035392"]) { display:block!important; z-index:50!important; order:50!important; }

However, OBS does not support the :has statement. I asked about it here a while back:

The has() selector was added to Chromium in version 105. OBS is currently on version 103 of Chromium. This will be updated to a newer version some day, but I cannot say which version that will be or when it will happen (It will not be in OBS 30).

I'd have to know exactly what you're trying to accomplish in order to propose an alternative.

Feel free to hop in the OBS Discord and ping me (@Warchamp7) in #plugins-and-tools or another channel, as I'd like to avoid cluttering up this PR with development work ;D

[Draco18s]

Yes my example is tested and working in both Chrome and OBS itself. It's possible they reverted that restriction. If it was re-enabled again, another potential workaround would be to upload the icons to Discord and use the attachment URL. I doubt they would block a subset of their own CDN. A cumbersome workaround, sure, but again doable without JS injection.

Joke's on you, they recently added expiration tokens to upload links. So this solution is only good for 3 to 12 months.

https://www.reddit.com/r/discordapp/comments/16uy0an/not_sure_if_this_is_real/

[Warchamp7]

Yes my example is tested and working in both Chrome and OBS itself. It's possible they reverted that restriction. If it was re-enabled again, another potential workaround would be to upload the icons to Discord and use the attachment URL. I doubt they would block a subset of their own CDN. A cumbersome workaround, sure, but again doable without JS injection.

Joke's on you, they recently added expiration tokens to upload links. So this solution is only good for 3 to 12 months.

https://www.reddit.com/r/discordapp/comments/16uy0an/not_sure_if_this_is_real/

Again, it's a worst case scenario solution. It currently does not appear to be needed anyway.