Funscript AI Generation - VR (& 2D POV now?) - Join the Discord :)

[k00gar]

Thanks for your feedback.

You mean more a Copilot assistant in this style ?

I trained a transformer model yesterday and raw coded this editor with suggestions/predictions based on the last dozen positions scripted, that you can accept with Enter or just override, etc.

The positions on the left are the one I input, on the right of the red bar the suggestions. Disregard the non relevancy of the positions, it is just to showcase the prediction based on the previous positions.

Actually planning on including this type of prediction in my unsupervised approach when both yolo tracking and kalman predictors fail.

[k00gar]

Definitely, if it gets any traction / interest at some point.
Not sure how nor when, the very little help offers I received until now did not seem to make the original project progress, so I actually started back from start with the yolo training approach.

[Rriik]

@k00gar

You mean more a Copilot assistant in this style ?

Not necessarily, i meant a full editor similar to the open funscripter project (where you can not only script manually but also extend the functionality of the editor via plugins) that would also contain a layer of AI features on top. And if it sounds like i basically want OFS but better, it’s because i want OFS but better lol. It was very dissapointing to see that project get canned when it had so much potential, but on the flipside it was not easy to extend it because of the author’s choice of underlying framework/language. But yeah i feel like this community would benefit greatly from something that would take the best of all existing funscripter environments, put them in one package and offer AI features to boot. Probably just a pipe dream though.

[k00gar]

Further refining of the algorithm, footage below of unsupervised funscript generation (green in the graph), comparing to the human made funscript (blue in the graph) for 4 randomly picked sections of the video:

[k00gar]

Still tuning, mouth and hands movements being dealt with better during blowjob/handjob. Still need to test foot job, and adjust amplitude globally. Working on it. Here is the result of a generated funscript vs a human generated script the scripter considers “exaggerated as per his taste” - got most of the strokes right and in sync, some human generated strokes seem out of context when analyzing deeply, and some erroneous interpretation from the algorithm, amplitude needs to be tuned… Fully unsupervised.

[k00gar]

Hoping I can offer generating full funscript files soon to willing testers for feedback and adjustments.

[ellequadro]

Thanks for your great work. I am available to test the script.

[serendipitis]

So is this a fine tuning of a vision transformer or did you generate embeddings for each frame from sort of convnet and treat those embeddings as the sequence?

[k00gar]

It is a yolo11 base model that I trained for 200 epochs with a dataset of 4000+ images that I built and tagged. If this answers your question…

As of generating full video script, I have been struggling with a memory leak for the past few days.

Now in the process of rewriting the code with the following approach:

• Perform the yolo detection for each frame
• Dump the results to a file
• Run the tracking algo on the file

Currently testing…

[k00gar]

First VR video fully parsed using this new pipeline.

It took my computer:

• 45 mins to perform the detection on every single frame of a sized down VR 1920*1080 60FPS video of 46 mins and dump the data to a file (24MB file)
• 10 mins to perform the tracking algo based on the data dumped in the file

Here’s what the results looks like:

[k00gar]

Thanks for your offer, if you have a VR video for which you’d like to generate a funscript, feel free to suggest. I would be glad to get some feedback and assess what needs tuning and what needs a full rehaul…
In the meantime, I just picked a video from the Script request section, and launched the pipeline to see what comes out and see if the user is fine with sharing feedback to adjust and tune the algorithm

[SlowTap]

Out of curiosity, have you tried comparing different runs on the same downsized video, with less and less FPS till it breaks? Just to have a metric on how resilient it is.

[k00gar]

Thank you for your question

To make it short : no, not really.

I actually downsized once a video to 640p (size of the images yolo11 was trained on), but I had the impression the detection were not as good as when used on the 1080p version of that same video.

But, you are absolutely right, that would be a nice benchmark to run, I need to plan that.

Adding on the To Do list…

[k00gar]

For this :

https://discuss.eroscripts.com/t/arporn-angel-of-nurse-passthrough-angel-youngs/213829

The algo comes up with this:

It looks like some sections (flat line), beyond the pussy closeup which was identified properly, have put it at some difficulties. Need to look into it.

ARPorn_Angel Youngs_Angel Of Nurse_4000p_2K_original_FISHEYE190_alpha_kAI_vw8.funscript (190.8 KB)

For this one:

https://discuss.eroscripts.com/t/vanessa-vega-honeymoon-in-vega-milfvr/188180

Same here, need to look into it and tune/troubleshoot.

Milfvr - Honeymoon In Vega - Vanessa Vega_1080p_kAI_vw8.funscript (200.1 KB)

If anyone is brave enough to take a look / test, or want to give a shot on another video, feel free to ask.

[k00gar]

Actually, upon deeper dive-in, looks like a mess compared to the unit-single sections I was parsing before.
Need to understand what broke in the scaling-up-to-full-video process.
Getting tired of this.

[Mnamejeff]

Really amazing work you’re doing here, I applaud and thank you!

[k00gar]

Looks like I finally made some progress out of the swamp I was stuck in.

Took time to rest and think of other stuff, before fully rehauling the tracking algorithm yesterday, while still relying on the yolo model for detection (first stage of the pipeline)

Pipeline : yolo => .json file => tracking algo => funscript => simplifier => .funscript file

Here is a script that looks closer to decent than anything the algo produced before:

SLR_SLR Originals_She Gives The Best Gifts_1920p_35951_MKX200_kAI.funscript (175.9 KB)

For the following video : She Gives the Best Gifts

Still a lot to do, any testing and feedback would be extremely useful.

[ellequadro]

thanks for your work. I think you are an inspiration for those who wanted to try this path but did not know where to start (me for example).

What are the intentions of this project? sell the service or release it for free? When do you think we can start testing independently?

thanks again

[k00gar]

Thank you very much for your message, it helps keep up working on the project, when I sometimes just want to let if off where it is.

That’s a fair question.

Actually, I have been experimenting more to keep my head busy, rather than thinking of other life stuff that keeps me awake at night.

The idea came to me after I submitted my first script to the community. To be more precise, during my second script, that I actually never finished because I got sort of exhausted in the making. I felt like, maybe, there was a way to automate part of the process, and then understood how hard it actually was to do so…

Anyway, it is still at a very early stage, quite broken, and my guess is it would definitely profit from contributions to make it better (detection, tracking logic, etc), either with testing feedback and / or coding suggestions/rehaul, and maybe finetuning of the YOLO model I trained.

It’s just, that I started with no real coding nor ComputerVision experience, so I’m a bit shy to share the not-so-good looking code…

To make it short : I am thinking of making it freely available to whoever might want to experiment it, though maybe not for commercial use.

In the end, I hope this just benefits the community, even if I would not mind being offered a coffee or two if anyone ends up appreciating the ability to generate a funscript for a video he likes, but that is just not why I initiated this project.

[k00gar]

This one required some more tuning, as it is the first time I am trying to generate a funscript with more than one actress :

https://discuss.eroscripts.com/t/czechvr-730-aleya-sun-naomi-bennett-injured-but-horny/204199

Injured, But Horny! 730 - Czech VR Network

730-czechvr-3d-7680x3840-60fps-oculusrift_uhq_h265vw_15.0.funscript (242.6 KB)