heestand-xyz / pixelkit Goto Github PK

Hi all,

We are trying to use the CameraPIX together with a MOV that has an alpha channel.

` let camera = CameraPIX()
camera.camera = .front
let overlay = VideoPIX()
overlay.load(fileNamed: "FreeSmokeYoutube", withExtension: "mov") { done in
let blend = BlendPIX()
blend.blendMode = .addWithAlpha
blend.placement = .aspectFill
blend.extend = .mirror
blend.inputA = camera
blend.inputB = overlay

                self.record = RecordPIX()
                self.record?.input = blend

                try? self.record?.startRec(name: "tempvid")
            }`

The outcome though, it's really low frame rate. This is basically using the default configuration so, what could it be?

Thanks

.gitmodules contains:

https://github.com/heestand-xyz/MetlShaders.git

but it looks like it should be:

https://github.com/heestand-xyz/MetalShaders.git

('Metal' instead of 'Metl')

Hi Hexagons,

Trying to install the package and getting the following error. Can you please advise.

because PixelKit >=1.0.13 contains incompatible tools version and root depends on PixelKit 1.1.2..<1.2.0, version solving failed.

I'd love PixelKit,but,I find that ,Unable to find a pod with name, author, summary, or description matching PixelKit

how it's possible to export Green Screen to asset or video, I use the example here
Green Screen in Swift & PixelKit
http://blog.hexagons.se/blog/green-screen-in-swift-pixelkit/

I'm using the following code to create an initial alpha layer called threshold and then multiplying it with border color. The code below is only run once.

 let fullscreenBackground = ColorPIX(at: .fullscreen)
    fullscreenBackground.color = .clear
    
    let blended: BlendPIX = fullscreenBackground & imagePix
    
    let mask: ReorderPIX = ReorderPIX()
    mask.inputA = blended
    mask.inputB = blended
    mask.redChannel = .alpha
    mask.greenChannel = .alpha
    mask.blueChannel = .alpha
    mask.alphaChannel = .one
    
    threshold = blended._blur(0.1)._threshold(0.1)._lumaToAlpha()
    
    let borderColor = LiveColor(color)
    let colorEdge = threshold! * borderColor
    
    let final: PIX = colorEdge & imagePix
    final.delegate = self

Later by changing the color I'm calling this following method on a debounce of 2 seconds:

guard let threshold = threshold else { return }
    let borderColor = LiveColor(color)
    let colorEdge = threshold * borderColor
    
    let final: PIX = colorEdge & imagePix
    final.delegate = self

M_1_PI_F is used. Tho this is 1.0 / PI. M_PI_F needs to be used.

I'm sorry about this issue and sorry about those things,

but when I export the video using it export only the video without the audio channel,

Note if I did that flag to true,
RecordPIX.recordAudio = true

it will record the microphone audio along with video audio,
is it possible to export the video ad audio without enabling the microphone?

Hi,

first off, congrats for the great work done.

I want to apply flarePIX into Scenekit via ScenePIX , but ScenePIX is undocumented.

I see

public func setup(cameraNode: SCNNode, scene: SCNScene, sceneView: SCNView)

what is the correct way to apply scenePIX to Scenekit?

Thanks.

Hi,

trying to setup this on scenekit , I think I am missing one or two steps for the init

Crash on the below in ScenePIX.swift , line 91
renderer.render(atTime: 0, viewport: viewport, commandBuffer: commandBuffer, passDescriptor: renderPassDescriptor)
<--- renderer seems nil <--- Thread 1: Fatal error: Unexpectedly found nil while implicitly unwrapping an Optional value

Log:
RenderKit ready to render.
PixelKit #00 WARNING [1] ScenePIX Render >>> Metal View res not set
PixelKit #1 INFO [1] ScenePIX Res >>> Applied: custom(w: 1500, h: 1000) [1500x1000]
Fatal error: Unexpectedly found nil while implicitly unwrapping an Optional value: file PixelKit/ScenePIX.swift, line 91

Code
DispatchQueue.global(qos: .background).async {
let scenePix = ScenePIX(at: .custom(w: 1500, h: 1000))
scenePix.setup(cameraNode: self.sceneView.currentPointOfView()!, scene: self.sceneView.currentScene()!, sceneView: self.sceneView)
let res: Resolution = .custom(w: 1500, h: 1000)
let circle = CirclePIX(at: res)
circle.radius = 0.45
circle.bgColor = .clear
let finalPix: PIX = scenePix * circle
finalPix.view.frame = self.view.bounds
self.view.addSubview(finalPix.view)
}

May I have an hint were to look please as there are no readme or examples around on scenePIX?

Thank you very much.

on iOS 16 I get

CAMetalLayer ignoring invalid setDrawableSize width=0.000000 height=0.000000

or

2023-07-10 15:15:22.745864-0400 app[41287:14957907] [Graphics] Invalid size provided to UIGraphicsBeginImageContext(): size={353, 0}, scale=3.000000
2023-07-10 15:15:22.746013-0400 app[41287:14957907] [Unknown process name] CGContextGetType: invalid context 0x0. If you want to see the backtrace, please set CG_CONTEXT_SHOW_BACKTRACE environmental variable.
-----> (1059.0, 0.0) (128.0, 128.0) (353.0, 0.0)

code:

struct Modified<V: View>: View {
	let child: V

	var body: some View {
		let view = ViewPIX { child }
		let pix = view.pixEdge()
		return PixelView(pix: pix)
	}
}

Couple of files in this project do not conform to NodeMetal protocol.
Specifically metalFileName property introduced here is missing for multiple files ex. MetalMergerEffectPIX.

I've added PixelKit to my Podfile and installed it. Xcode does pick it up and I can import it and use the UIKit related part without problems. However, I can't use SwiftUI related types as those don't get resolved by Xcode.

On iOS 16, all UIViewRepresentables have to be structs.

Do you have an example of using video recording?

I was hacking together a demo where I had to do portrait segmentation (Zoom like background removal) without chroma key.
I decided to use Deeplabv3 and found this project: https://github.com/jsharp83/MetalCamera which has all the parts properly setup. With a few minor changes I was able to macOS compatible as well.

This would make a great addition to PixelKit.

My use case is: remove the background from realtime video and merge it with a custom image/video.

Relevant File:
https://github.com/jsharp83/MetalCamera/blob/master/MetalCamera/Classes/CoreML/CoreMLClassifierHandler.swift

What I am trying to achieve is the following:

Use the CameraPix and remove background (green screen), blend it with a custom background, get IOSurface backed CVPixelBuffer from the resultant PIX and publish them to an external process.

Hi heestand,

Thanks for the fix regarding the swift version, it still won't build, here are the errors, I have linked to the source files:

Sorry it's very ugly 😅, do you think it will be possible to remediate those issue to be compatible with an older SDK than Big Sur?

Thanks

Where is the link?

Hi, is it planned to add the function of generating text in the form of a texture using  Metal API?. What methods do you know for generating text using metal? I would be very happy if you gave an example. My goal is to overlay the effect of fire on the text. thanks for the answer

This is a very good framework, I want to save the processed photos to the album, but can't find the corresponding method, I see the code is directly loaded View:

           let image = ImagePIX()
            image.image = self.myImage
            let edge = EdgePIX()
            edge.input = image
            edge.strength = 3
            edge.view.frame = myImageView.bounds
            self.view.addSubview(edge.view)

Can't get image information of edge.view

Hi,

Sorry I was out of the swift/spm loop for two years, as I was still on Mojave, and a lot has changed, querying google gives various options none seems to work.

First of I'm under Catalina and using XCode 12.4 (AFAIK the last version compatible with Catalina).
My project is a basic SwiftUI macOS app.

In XCode I went to File > Swift Packages > Add Package Dependency... and search for PixelKit.

In the version selection panel I tried a few things:

• Exact == 2.0.4
• Up to Next Major from 2.0.4
• branch "main"

Both returned the following:

Very interesting project.

Thanks