feat(add): motion data | Bézier Mimic Tear

[QIN2DIM]

No description provided.

[mevrover]

where can I learn more about motion data and editing it?

[QIN2DIM]

@mevrover I recommend this article, which is easy to understand and suitable for getting started.🧙‍♂️

[ijnrghjkdsmigywneig203]

This will include having a human like curve to click the hcaptcha element and also having some movement being stimulated while clicking the images?

[QIN2DIM]

Not available in headless mode

[ijnrghjkdsmigywneig203]

Right

[ijnrghjkdsmigywneig203]

@QIN2DIM When could we expect this update?

[ijnrghjkdsmigywneig203]

@QIN2DIM check the repo of selenium-cursor - https://github.com/smirad91/SeleniumCursor
Makes realistic like movements to click elements. I couldnt get it to correctly move to the images, instead it would slightly miss them.

[QIN2DIM]

@ijnrghjkdsmigywneig203 lovely cat

[ijnrghjkdsmigywneig203]

@ijnrghjkdsmigywneig203 lovely cat

When do you think this human mimicking movements update will be implemented?

[ijnrghjkdsmigywneig203]

@QIN2DIM When will you add this feature...

[QIN2DIM]

I will not open source for now, this feature is a paid resource in some areas and the price tag is very high. I don't want to upset the balance.

[ijnrghjkdsmigywneig203]

I will not open source for now, this feature is a paid resource in some areas and the price tag is very high. I don't want to upset the balance.

could you give it to me personally?

[ijnrghjkdsmigywneig203]

I am only going to use it for my personal projects, nothing will be released/opensourced.

[ijnrghjkdsmigywneig203]

If not, where can I get started to try and make my own version? @QIN2DIM

[QIN2DIM]

It is very difficult to make the stated " mimic" effect through code alone. The kinematic knowledge and mathematical analysis process involved here is so complex that I don't think it can be done by some high school or college student.

The solution I am using is "reinforcement learning + temporal intent". I added context modifiers to all ActionChains APIs, adding Bezier curve traces that fit the intent between each of the original language APIs regarding mouse actions. The solution is able to recreate the human mouse trajectory to the maximum extent and can spoof ML-based trajectory recognition patches.

If you intend to express some curve equation through code, that's fine, but as I said above, you need to put it into a business context to articulate the movement intent, otherwise it will be flagged by ML. I'm not sure if Discord will fit these trajectories through machine learning, but doing so allows you to generate trajectories that make sense.

[QIN2DIM]

I've observed a lot of open source projects on this, and I think most of them generate mouse trajectories with the purpose of spoofing detectors, rather than mimicking human behavior online. So it's another never-ending game of cat and mouse.👻

I'm looking forward to the next generation of human-computer challenge games, which will be very interesting.

[ijnrghjkdsmigywneig203]

I've observed a lot of open source projects on this, and I think most of them generate mouse trajectories with the purpose of spoofing detectors, rather than mimicking human behavior online. So it's another never-ending game of cat and mouse.👻

I'm looking forward to the next generation of human-computer challenge games, which will be very interesting.

That's what im trying to do, bypass the detection. Not mimick human behavour, I mean't it as the same thing.

[ijnrghjkdsmigywneig203]

It is very difficult to make the stated " mimic" effect through code alone. The kinematic knowledge and mathematical analysis process involved here is so complex that I don't think it can be done by some high school or college student.

The solution I am using is "reinforcement learning + temporal intent". I added context modifiers to all ActionChains APIs, adding Bezier curve traces that fit the intent between each of the original language APIs regarding mouse actions. The solution is able to recreate the human mouse trajectory to the maximum extent and can spoof ML-based trajectory recognition patches.

If you intend to express some curve equation through code, that's fine, but as I said above, you need to put it into a business context to articulate the movement intent, otherwise it will be flagged by ML. I'm not sure if Discord will fit these trajectories through machine learning, but doing so allows you to generate trajectories that make sense.

Thanks, I will look into it👍

[jjhesk]

@QIN2DIM you have basically modified the selenium and adding an additional module for mouse movement that can generate valid motionData for hcap validation detection?