itch.io

in 0.8, when hotscreen gets paused, latency correction get's flushed - would be great if there would be an option to keep it stored, so when hotscreen gets resumend, latency correction is still there

this is a feature idea or mod idea that I think might be possible, but I'm not code-abled enough to make it myself.

My idea is to have some effects that only appears when two (or more) boxes collide with each other - when two detection windows overlap. That would allow to make things like censoring a feature only when another one is present or close. Censoring might be applied either on the overlapped area or whole area of collided boxes.
If anyone has an idea how to make that into mod, please share 

you have to download and unpack 0.8 as new, but you can copy all saved presets from previous versions - just copy them to CUSTOM_DATA 

as of losing track of a body part, you could use a "stay detected during" slider - setting it for .1-.2 sec makes most detection mishaps covered. If you're not using it, increasing screen latency by 1-2 frames makes the detection more consistent
as of body detection, the are different threads about that and someone shared their shader that could do this to some extent

you could possibly do it with a levels collection with quick level progression, that would trigger between a normal view and a view that would be a inverted censor/color with goon/screen takeover. having this with maxed out progression/regression speed (max 20) and a progress multiplier of 5 or more, it would trigger in a second of detection

that's quite cool, though when applied to irl with complex lighting it's not doing right
to be honest, i believe that creating a filter that would work exactly like that (so follow the edge of body then fill it with whatever) would require an another approach, i wonder if some kind of vector output is possible from this kind of neuralnet models - if the detection model would be able to return a vector shape of body, then having this kind of clipping effect should be very easy. I wonder though how much more resource-hungry this model would be, I guess that having a vector image processing would also require a huge change in the whole program's code and the way that image is processed now - but those are only guesses 'cause i've got little to no knowledge how those things work 

having additional detection models like the ones that you mentioned - full body, additional details (mouth, nipples, someone in another thread proposed a chastity/erect penis, individual eyes) it all might be great thing - that would require quite a lot of time to train a model but I believe it might be really possible -or some of that work might even be done already - afaik there's been some research done on few different platforms (eg nudenet on github) 

the circular mosaic is possibly easy to achieve as a shader - you can try to generate it with some llm - both gemini and chatgpt can provide some cool effects; but speaking of shaders, i'd like to know how to do 2 things - how to make a shader scalable (do it can be adjusted to smaller details, just as blur or pixel filters can be scaled) and if it would be possible to sort of mix some effects. For example, i'd like to create an effect that would make an black-n-white screen, but i''d like to have a pixel effect applied to that same area, that would also be monochrome

I think that an option that would allow for release of resources without closing program (closing an overlay and any detections) with something like a 'panic button' could be cool? idk if that would be useful for many people though

Excluding a config window from effects would also be great, as it's now quite easy to lock yourself from it when using a single monitor and invert boxes setting 

this is an old thread now, but I'd like to add that eye detection do have some more issues. It seems that it doesn't follow the "stay detected during [time]" setting and uses only the performance tab setting of eye position update frequency. adaptive censors (such as blur or pixel) has some scaling issues when eyes aren't in horizontal orientation - it get's scaled up when eyes are getting rotated up to 90 degrees. Model has some issues with glasses, that makes it really shaky when adapted to a face equipped with them.

you can do that right away, just look for someone willing to take control by a anydesk or sth like that and you're good to go, I believe that creating a whole service for sharing people's locks might be a bit much to ask a humble dev

that would be cool and i think it should be possible to create a mod that would do that - although I have no practical knowledge as how to do that

seems that it's the case - changing detection rate didn' reduce flickering, but made it shift into higher values - from ~40fps to 70-110fps on the output rate
I checked the obs though and using 4k input 4k output with dxgi made a stable 60fps vid - neither visual or program detected framerate change 

I believe it's a countermeasure necessary to avoid freezing your pc - in v0.6 it was set so any time that your pc lagged, it got set back to 0. In 0.7 it got changed, but I can still notice this changing to 0, especially after some resource devouring actions

I believe that OP is referring to values in the performance tab, and recording they are mentioning is the sampling rate of hotscreen itself

It may be worth noting that switching 4x detection  on or off does not improve performance much - i'd say the change is neglectible if any

hey, i'm using 4070ti super on my 4k display and I have really simmilar experience, although my average hotscreen performance is around 50fps, with both screen recording and body detection rates in 40-50fps range, despite setting fps presets for normal (60) or high (120). - higher fps cap allows for bigger variation of performance, as overlays frame rate range from 40 to 80 fps

neither mine cpu/ram/gpu performance seems strained - 10gb free ram space, cpu under 25% usage, gpu memory under 20%, yet the experience seems like it's on the edge (lol) of it's capabilities

I believe the issue here is about the whole screen glitch effect that's applied on any occasion that hotscreen starts, it's the same effect as the default level progression effect

basing it on v0.6 vs v0.7, you have to copy your named censor presets from here: 
Hotscreen 0.6\custom_ressources\collections
into there:
Hotscreen 0.7\CUSTOM_DATA

however, i'd recommend creating another dir in the CUSTOM_DATA, as the files are recognized by the hotscreen and they're put in neat file tree

this would also help with configuration, as the whole screen obstruction can be hard to work with - choosing a pixel censor makes it really hard to fine tune when the whole screen gets pixelated. Or perhaps an exclusion for config window would be possible?

Hey, I have no idea how complicated that would be code-wise but I was wondering whether it would be possible to have a feature that would recognize the rotation of detected element, that is more of rectangle than a square and apply a censor box with simmilar rotation? That would allow some kind of 'minimalistic' censoring, that would hide a censored object with little to none margin. Once again - I have no idea how much effort that would take, or if that's actually possible within the current neural-net's detection, andhow resource-hungry would this feature be