itch.io

Np <3 just glad to be here~

Hi,

I didn't know where else to ask this, but are you aware of three software Purevision? I think it's still not released yet but it's being made by the guys who made Puryfi. (I'm in their discord so i know it's in development but I'm not sure if it's out yet)

 They seem to be using Hotscreen as a benchmark I think? Or they're at least aware of it and are comparing some aspects of Purevision to Hotscreen to aid their development and push it in the right direction.

I just wanted to inform you that this kinda thing is being developed. I'm not trying to portray it as negative or positive, just thought you should know lol. 

But I think Hotscreen has become kind of a pioneer of the monitor censoring thing which is pretty cool👍

yeah censoring more is kinda what a lot of us is here for i bet lol

yeah for safe for beta i did notice individual breast detection but thought it was the same for hotscreen? If its not the same as hotscreen it would be possible soon to achieve nipple censoring.

i would recommend trying to lower the width and height? idk if that would work tho.. im really into the more full body censoring :p

quick question, but does the detection software improve with every update or is it only when mentioned?

Just curious

i understand! and i hope to see how it imporves! 

One last thing to mention as well, when watching an action show with a lot of ecchi, like 7 deadly sins, the model seems detect a lot of smoke and other things as breasts or butts, because I am running the Safe for Beta collection and random smoke clouds and dust become glitched lol

I wanted to bring up that a lot of the time the detection for anime isn't the greatest, even with the finetuned anime model. For some context, I was watching some of the older animes which usually have a lot of ecchi which i wanted to get blurred out, but it wouldn't detect it the best. It could be because it's animation, or maybe because the drawings aren't the most high quality so it has a hard time determining detectable things but it was failing to detect a lot of things more than I expected.

I've found it usually has a hard time detecting:

- male vs female faces

- male vs female chests

- covered female breasts

I also found exposed female breasts to be a bit iffy sometimes only detecting one breast, but those three are the main ones it seems to falter in. 

Overall though for most other use cases it seems to work fine!

Hi,

So sometimes if there hasn't been a detection for a while and its just idling not censoring anything, sometimes it'll just stop detecting things even when detectable images show up. If I restart the app it seems to fix itself and work again.

I'm using the bitmap  recording method if that helps!

i see..i do understand that would be really hard and i don't expect you to take that priority but i like that you have it in mind😁 thanks!

Hi, 

Is there any plans at the moment to make a mobile version as well?

Thanks😅

This update was actually one of the best in terms of features and customization. The UX feels a lot better and the way collections work is a big improvement from last time. In the last update there wasn't really a way to turn off one collection (that I found) other than to turn off each filter one by one, which was really annoying. But now it works a hell of a lot better so good work! 

One of my only gripes is with the detection model, which is obviously not perfect. It seems to fail with a bit of noise, which is unfortunate. I was skimming through this video to see how the new update works: https://www.the-joi-database.com/watch/82f8da30362cf55f5c2c47e3 , and it was a bit disappointing  to barely see the AI work at all. 

Overall though, great UX update! Love how this is developing and I hope to see even more!!

PS: I have no idea why but I'm always stuck on the Bitmap version of detection and I can't get the DXGI to work. If I could get help with that that would be great!

Hi! 

Not sure if you remember me but I kinda reviewed the prequel to this and I want to say that I love what your doing here! 

Just a quick suggestion though, not sure how many other people do this but I lot of times I'll watch a JOI video (usually from joidatabase) and they are almost always anime JOIs. I found that many times the AI model wasn't able to detect a lot of the body parts a lot of the time especially when they were first coming into frame. I'm not sure how this helps but I felt that it was important to give feedback..

Maybe you could use some of the frames from those videos which the AI struggles with to help train it?

Anyway I hope you are doing well and I wish to see more from you in the future!

P.S I also wanted to say that after a while of no detections, the app seems to stop working and not detect anything.. So if there is a way to fix that please let me know!

This is actually really enjoyable! I just got it and it seems to be working right off the bat! But there is no startup to tell you whats going on, like the punishment thing is at 75% for me right now and I had no idea what was going to happen when it hit 100. Maybe that's part of the appeal, idk. The only real concerns or improvements I could say are 2:

1. Improve the detection for hentai, it doesn't work very well with it. I'm not sure exactly how nudenet works but it seems to have trouble detecting smaller images and asses, especially anything clothed.

2. Improve the delay on the overlay. I'm not sure how big an ask this is but as of now it's quite and fails to keep the detections on the right things when I scroll.

Other than that though this is really cool! Great work!