Implement Metal rendering

submodules/Postbox/Sources/TimeBasedCleanup.swift

@@ -170,7 +170,7 @@ private final class TimeBasedCleanupImpl {
         let generalPaths = self.generalPaths
         let shortLivedPaths = self.shortLivedPaths
         let scanOnce = Signal<Never, NoError> { subscriber in
-            DispatchQueue.global(qos: .utility).async {
+            DispatchQueue.global(qos: .background).async {
                 var removedShortLivedCount: Int = 0
                 var removedGeneralCount: Int = 0
                 var removedGeneralLimitCount: Int = 0

submodules/TelegramCallsUI/BUILD

@@ -1,4 +1,44 @@
 load("@build_bazel_rules_swift//swift:swift.bzl", "swift_library")
+load(
+    "@build_bazel_rules_apple//apple:resources.bzl",
+    "apple_resource_bundle",
+    "apple_resource_group",
+)
+load("//build-system/bazel-utils:plist_fragment.bzl",
+    "plist_fragment",
+)
+
+filegroup(
+    name = "TelegramCallsUIMetalResources",
+    srcs = glob([
+        "Resources/**/*.metal",
+    ]),
+    visibility = ["//visibility:public"],
+)
+
+plist_fragment(
+    name = "TelegramCallsUIBundleInfoPlist",
+    extension = "plist",
+    template =
+    """
+    <key>CFBundleIdentifier</key>
+    <string>org.telegram.TelegramCallsUI</string>
+    <key>CFBundleDevelopmentRegion</key>
+    <string>en</string>
+    <key>CFBundleName</key>
+    <string>TelegramCallsUI</string>
+    """
+)
+
+apple_resource_bundle(
+    name = "TelegramCallsUIBundle",
+    infoplists = [
+        ":TelegramCallsUIBundleInfoPlist",
+    ],
+    resources = [
+        ":TelegramCallsUIMetalResources",
+    ],
+)
 
 swift_library(
     name = "TelegramCallsUI",
@@ -6,6 +46,9 @@ swift_library(
     srcs = glob([
         "Sources/**/*.swift",
     ]),
+    data = [
+        ":TelegramCallsUIBundle",
+    ],
     deps = [
     	"//submodules/SSignalKit/SwiftSignalKit:SwiftSignalKit",
         "//submodules/Display:Display",

submodules/TelegramCallsUI/Resources/I420VideoShaders.metal

@@ -0,0 +1,49 @@
+#include <metal_stdlib>
+using namespace metal;
+
+typedef struct {
+    packed_float2 position;
+    packed_float2 texcoord;
+} Vertex;
+
+typedef struct {
+    float4 position[[position]];
+    float2 texcoord;
+} Varyings;
+
+vertex Varyings i420VertexPassthrough(constant Vertex *verticies[[buffer(0)]],
+                                  unsigned int vid[[vertex_id]]) {
+    Varyings out;
+    constant Vertex &v = verticies[vid];
+    out.position = float4(float2(v.position), 0.0, 1.0);
+    out.texcoord = v.texcoord;
+
+    return out;
+}
+
+fragment half4 i420FragmentColorConversion(
+    Varyings in[[stage_in]],
+    texture2d<float, access::sample> textureY[[texture(0)]],
+    texture2d<float, access::sample> textureU[[texture(1)]],
+    texture2d<float, access::sample> textureV[[texture(2)]]) {
+    constexpr sampler s(address::clamp_to_edge, filter::linear);
+    float y;
+    float u;
+    float v;
+    float r;
+    float g;
+    float b;
+    // Conversion for YUV to rgb from http://www.fourcc.org/fccyvrgb.php
+    y = textureY.sample(s, in.texcoord).r;
+    u = textureU.sample(s, in.texcoord).r;
+    v = textureV.sample(s, in.texcoord).r;
+    u = u - 0.5;
+    v = v - 0.5;
+    r = y + 1.403 * v;
+    g = y - 0.344 * u - 0.714 * v;
+    b = y + 1.770 * u;
+
+    float4 out = float4(r, g, b, 1.0);
+
+    return half4(out);
+}

submodules/TelegramCallsUI/Resources/NV12VideoShaders.metal

@@ -0,0 +1,57 @@
+#include <metal_stdlib>
+using namespace metal;
+
+typedef struct {
+    packed_float2 position;
+    packed_float2 texcoord;
+} Vertex;
+
+typedef struct {
+    float4 position[[position]];
+    float2 texcoord;
+} Varyings;
+
+vertex Varyings nv12VertexPassthrough(
+    constant Vertex *verticies[[buffer(0)]],
+    unsigned int vid[[vertex_id]]
+) {
+    Varyings out;
+    constant Vertex &v = verticies[vid];
+    out.position = float4(float2(v.position), 0.0, 1.0);
+    out.texcoord = v.texcoord;
+    return out;
+}
+
+float4 samplePoint(texture2d<float, access::sample> textureY, texture2d<float, access::sample> textureCbCr, sampler s, float2 texcoord) {
+    float y;
+    float2 uv;
+    y = textureY.sample(s, texcoord).r;
+    uv = textureCbCr.sample(s, texcoord).rg - float2(0.5, 0.5);
+
+    // Conversion for YUV to rgb from http://www.fourcc.org/fccyvrgb.php
+    float4 out = float4(y + 1.403 * uv.y, y - 0.344 * uv.x - 0.714 * uv.y, y + 1.770 * uv.x, 1.0);
+    return out;
+}
+
+fragment half4 nv12FragmentColorConversion(
+    Varyings in[[stage_in]],
+    texture2d<float, access::sample> textureY[[texture(0)]],
+    texture2d<float, access::sample> textureCbCr[[texture(1)]]
+) {
+    constexpr sampler s(address::clamp_to_edge, filter::linear);
+
+    float4 out = samplePoint(textureY, textureCbCr, s, in.texcoord);
+
+    return half4(out);
+}
+
+fragment half4 blitFragmentColorConversion(
+    Varyings in[[stage_in]],
+    texture2d<float, access::sample> texture[[texture(0)]]
+) {
+    constexpr sampler s(address::clamp_to_edge, filter::linear);
+
+    float4 out = texture.sample(s, in.texcoord);
+
+    return half4(out);
+}

submodules/TelegramCallsUI/Sources/GroupVideoNode.swift

@@ -25,16 +25,17 @@ final class GroupVideoNode: ASDisplayNode {
     let sourceContainerNode: PinchSourceContainerNode
     private let containerNode: ASDisplayNode
     private let videoViewContainer: UIView
-    private let videoView: PresentationCallVideoView
+    private let videoView: VideoRenderingView
     
     private let backdropVideoViewContainer: UIView
-    private let backdropVideoView: PresentationCallVideoView?
+    private let backdropVideoView: VideoRenderingView?
     private var backdropEffectView: UIVisualEffectView?
     
     private var effectView: UIVisualEffectView?
     private var isBlurred: Bool = false
 
     private var isEnabled: Bool = false
+    private var isBlurEnabled: Bool = false
         
     private var validLayout: (CGSize, LayoutMode)?
     
@@ -47,7 +48,7 @@ final class GroupVideoNode: ASDisplayNode {
     
     public var isMainstageExclusive = false
     
-    init(videoView: PresentationCallVideoView, backdropVideoView: PresentationCallVideoView?) {
+    init(videoView: VideoRenderingView, backdropVideoView: VideoRenderingView?) {
         self.sourceContainerNode = PinchSourceContainerNode()
         self.containerNode = ASDisplayNode()
         self.videoViewContainer = UIView()
@@ -61,7 +62,7 @@ final class GroupVideoNode: ASDisplayNode {
         super.init()
                 
         if let backdropVideoView = backdropVideoView {
-            self.backdropVideoViewContainer.addSubview(backdropVideoView.view)
+            self.backdropVideoViewContainer.addSubview(backdropVideoView)
             self.view.addSubview(self.backdropVideoViewContainer)
             
             let effect: UIVisualEffect
@@ -70,12 +71,12 @@ final class GroupVideoNode: ASDisplayNode {
             } else {
                 effect = UIBlurEffect(style: .dark)
             }
-            let backdropEffectView = UIVisualEffectView(effect: effect)
+            //let backdropEffectView = UIVisualEffectView(effect: effect)
-            self.view.addSubview(backdropEffectView)
+            //self.view.addSubview(backdropEffectView)
-            self.backdropEffectView = backdropEffectView
+            //self.backdropEffectView = backdropEffectView
         }
         
-        self.videoViewContainer.addSubview(self.videoView.view)
+        self.videoViewContainer.addSubview(self.videoView)
         self.addSubnode(self.sourceContainerNode)
         self.containerNode.view.addSubview(self.videoViewContainer)
         self.sourceContainerNode.contentNode.addSubnode(self.containerNode)
@@ -112,7 +113,7 @@ final class GroupVideoNode: ASDisplayNode {
         self.isEnabled = isEnabled
 
         self.videoView.updateIsEnabled(isEnabled)
-        self.backdropVideoView?.updateIsEnabled(isEnabled)
+        self.backdropVideoView?.updateIsEnabled(isEnabled && self.isBlurEnabled)
     }
     
     func updateIsBlurred(isBlurred: Bool, light: Bool = false, animated: Bool = true) {
@@ -150,11 +151,11 @@ final class GroupVideoNode: ASDisplayNode {
             self.backgroundColor = .black
         }
         var snapshotView: UIView?
-        if let snapshot = self.videoView.view.snapshotView(afterScreenUpdates: false) {
+        if let snapshot = self.videoView.snapshotView(afterScreenUpdates: false) {
             snapshotView = snapshot
-            snapshot.transform = self.videoView.view.transform
+            snapshot.transform = self.videoView.transform
-            snapshot.frame = self.videoView.view.frame
+            snapshot.frame = self.videoView.frame
-            self.videoView.view.superview?.insertSubview(snapshot, aboveSubview: self.videoView.view)
+            self.videoView.superview?.insertSubview(snapshot, aboveSubview: self.videoView)
         }
         UIView.transition(with: withBackground ? self.videoViewContainer : self.view, duration: 0.4, options: [.transitionFlipFromLeft, .curveEaseOut], animations: {
             UIView.performWithoutAnimation {
@@ -282,17 +283,17 @@ final class GroupVideoNode: ASDisplayNode {
         rotatedVideoFrame.size.width = ceil(rotatedVideoFrame.size.width)
         rotatedVideoFrame.size.height = ceil(rotatedVideoFrame.size.height)
 
-        self.videoView.view.alpha = 0.995
+        self.videoView.alpha = 0.995
         
         let normalizedVideoSize = rotatedVideoFrame.size.aspectFilled(CGSize(width: 1080.0, height: 1080.0))
-        transition.updatePosition(layer: self.videoView.view.layer, position: rotatedVideoFrame.center)
+        transition.updatePosition(layer: self.videoView.layer, position: rotatedVideoFrame.center)
-        transition.updateBounds(layer: self.videoView.view.layer, bounds: CGRect(origin: CGPoint(), size: normalizedVideoSize))
+        transition.updateBounds(layer: self.videoView.layer, bounds: CGRect(origin: CGPoint(), size: normalizedVideoSize))
         
         let transformScale: CGFloat = rotatedVideoFrame.width / normalizedVideoSize.width
         transition.updateTransformScale(layer: self.videoViewContainer.layer, scale: transformScale)
         
         if let backdropVideoView = self.backdropVideoView {
-            backdropVideoView.view.alpha = 0.995
+            backdropVideoView.alpha = 0.995
 
             let topFrame = rotatedVideoFrame
             
@@ -303,32 +304,34 @@ final class GroupVideoNode: ASDisplayNode {
             rotatedVideoFrame.size.width = ceil(rotatedVideoFrame.size.width)
             rotatedVideoFrame.size.height = ceil(rotatedVideoFrame.size.height)
 
-            let isBlurEnabled = !topFrame.contains(rotatedVideoFrame)
+            self.isBlurEnabled = !topFrame.contains(rotatedVideoFrame)
             
             let normalizedVideoSize = rotatedVideoFrame.size.aspectFilled(CGSize(width: 1080.0, height: 1080.0))
-            if isBlurEnabled {
+
-                self.backdropVideoView?.updateIsEnabled(self.isEnabled)
+            self.backdropVideoView?.updateIsEnabled(self.isEnabled && self.isBlurEnabled)
-                self.backdropVideoView?.view.isHidden = false
+
+            if self.isBlurEnabled {
+                self.backdropVideoView?.isHidden = false
                 self.backdropEffectView?.isHidden = false
             }
-            transition.updatePosition(layer: backdropVideoView.view.layer, position: rotatedVideoFrame.center, force: true, completion: { [weak self] value in
+            transition.updatePosition(layer: backdropVideoView.layer, position: rotatedVideoFrame.center, force: true, completion: { [weak self] value in
                 guard let strongSelf = self, value else {
                     return
                 }
-                if !isBlurEnabled {
+                if !strongSelf.isBlurEnabled {
                     strongSelf.backdropVideoView?.updateIsEnabled(false)
-                    strongSelf.backdropVideoView?.view.isHidden = true
+                    strongSelf.backdropVideoView?.isHidden = true
                     strongSelf.backdropEffectView?.isHidden = false
                 }
             })
-            transition.updateBounds(layer: backdropVideoView.view.layer, bounds: CGRect(origin: CGPoint(), size: normalizedVideoSize))
+            transition.updateBounds(layer: backdropVideoView.layer, bounds: CGRect(origin: CGPoint(), size: normalizedVideoSize))
             
             let transformScale: CGFloat = rotatedVideoFrame.width / normalizedVideoSize.width
 
             transition.updateTransformScale(layer: self.backdropVideoViewContainer.layer, scale: transformScale)
             
             let transition: ContainedViewLayoutTransition = .immediate
-            transition.updateTransformRotation(view: backdropVideoView.view, angle: angle)
+            transition.updateTransformRotation(view: backdropVideoView, angle: angle)
         }
         
         if let backdropEffectView = self.backdropEffectView {
@@ -359,7 +362,7 @@ final class GroupVideoNode: ASDisplayNode {
         }
         
         let transition: ContainedViewLayoutTransition = .immediate
-        transition.updateTransformRotation(view: self.videoView.view, angle: angle)
+        transition.updateTransformRotation(view: self.videoView, angle: angle)
     }
     
     var snapshotView: UIView?

submodules/TelegramCallsUI/Sources/MetalVideoRenderingView.swift

@@ -0,0 +1,652 @@
+import Foundation
+import UIKit
+import AsyncDisplayKit
+import Display
+import SwiftSignalKit
+import AccountContext
+import TelegramVoip
+import AVFoundation
+import Metal
+import MetalPerformanceShaders
+
+private func alignUp(size: Int, align: Int) -> Int {
+    precondition(((align - 1) & align) == 0, "Align must be a power of two")
+
+    let alignmentMask = align - 1
+    return (size + alignmentMask) & ~alignmentMask
+}
+
+private func getCubeVertexData(
+    cropX: Int,
+    cropY: Int,
+    cropWidth: Int,
+    cropHeight: Int,
+    frameWidth: Int,
+    frameHeight: Int,
+    rotation: Int,
+    buffer: UnsafeMutablePointer<Float>
+) {
+    let cropLeft = Float(cropX) / Float(frameWidth)
+    let cropRight = Float(cropX + cropWidth) / Float(frameWidth)
+    let cropTop = Float(cropY) / Float(frameHeight)
+    let cropBottom = Float(cropY + cropHeight) / Float(frameHeight)
+
+    switch rotation {
+    default:
+        var values: [Float] = [
+            -1.0, -1.0, cropLeft, cropBottom,
+            1.0, -1.0, cropRight, cropBottom,
+            -1.0,  1.0, cropLeft, cropTop,
+            1.0,  1.0, cropRight, cropTop
+        ]
+        memcpy(buffer, &values, values.count * MemoryLayout.size(ofValue: values[0]));
+    }
+}
+
+@available(iOS 13.0, *)
+private protocol FrameBufferRenderingState {
+    var frameSize: CGSize? { get }
+
+    func encode(renderingContext: MetalVideoRenderingContext, vertexBuffer: MTLBuffer, renderEncoder: MTLRenderCommandEncoder) -> Bool
+}
+
+@available(iOS 13.0, *)
+private final class BlitRenderingState {
+    static func encode(renderingContext: MetalVideoRenderingContext, texture: MTLTexture, vertexBuffer: MTLBuffer, renderEncoder: MTLRenderCommandEncoder) -> Bool {
+        renderEncoder.setRenderPipelineState(renderingContext.blitPipelineState)
+
+        renderEncoder.setVertexBuffer(vertexBuffer, offset: 0, index: 0)
+
+        renderEncoder.setFragmentTexture(texture, index: 0)
+
+        renderEncoder.drawPrimitives(type: .triangleStrip, vertexStart: 0, vertexCount: 4, instanceCount: 1)
+
+        return true
+    }
+}
+
+@available(iOS 13.0, *)
+private final class NV12FrameBufferRenderingState: FrameBufferRenderingState {
+    private var yTexture: MTLTexture?
+    private var uvTexture: MTLTexture?
+
+    var frameSize: CGSize? {
+        if let yTexture = self.yTexture {
+            return CGSize(width: yTexture.width, height: yTexture.height)
+        } else {
+            return nil
+        }
+    }
+
+    func updateTextureBuffers(renderingContext: MetalVideoRenderingContext, frameBuffer: OngoingGroupCallContext.VideoFrameData.NativeBuffer) {
+        let pixelBuffer = frameBuffer.pixelBuffer
+
+        var lumaTexture: MTLTexture?
+        var chromaTexture: MTLTexture?
+        var outTexture: CVMetalTexture?
+
+        let lumaWidth = CVPixelBufferGetWidthOfPlane(pixelBuffer, 0)
+        let lumaHeight = CVPixelBufferGetHeightOfPlane(pixelBuffer, 0)
+
+        var indexPlane = 0
+        var result = CVMetalTextureCacheCreateTextureFromImage(kCFAllocatorDefault, renderingContext.textureCache, pixelBuffer, nil, .r8Unorm, lumaWidth, lumaHeight, indexPlane, &outTexture)
+        if result == kCVReturnSuccess, let outTexture = outTexture {
+            lumaTexture = CVMetalTextureGetTexture(outTexture)
+        }
+        outTexture = nil
+
+        indexPlane = 1
+        result = CVMetalTextureCacheCreateTextureFromImage(kCFAllocatorDefault, renderingContext.textureCache, pixelBuffer, nil, .rg8Unorm, lumaWidth / 2, lumaHeight / 2, indexPlane, &outTexture)
+        if result == kCVReturnSuccess, let outTexture = outTexture {
+            chromaTexture = CVMetalTextureGetTexture(outTexture)
+        }
+        outTexture = nil
+
+        if let lumaTexture = lumaTexture, let chromaTexture = chromaTexture {
+            self.yTexture = lumaTexture
+            self.uvTexture = chromaTexture
+        } else {
+            self.yTexture = nil
+            self.uvTexture = nil
+        }
+    }
+
+    func encode(renderingContext: MetalVideoRenderingContext, vertexBuffer: MTLBuffer, renderEncoder: MTLRenderCommandEncoder) -> Bool {
+        guard let yTexture = self.yTexture, let uvTexture = self.uvTexture else {
+            return false
+        }
+
+        renderEncoder.setRenderPipelineState(renderingContext.nv12PipelineState)
+
+        renderEncoder.setVertexBuffer(vertexBuffer, offset: 0, index: 0)
+
+        renderEncoder.setFragmentTexture(yTexture, index: 0)
+        renderEncoder.setFragmentTexture(uvTexture, index: 1)
+
+        renderEncoder.drawPrimitives(type: .triangleStrip, vertexStart: 0, vertexCount: 4, instanceCount: 1)
+
+        return true
+    }
+}
+
+@available(iOS 13.0, *)
+private final class I420FrameBufferRenderingState: FrameBufferRenderingState {
+    private var yTexture: MTLTexture?
+    private var uTexture: MTLTexture?
+    private var vTexture: MTLTexture?
+
+    private var lumaTextureDescriptorSize: CGSize?
+    private var lumaTextureDescriptor: MTLTextureDescriptor?
+    private var chromaTextureDescriptor: MTLTextureDescriptor?
+
+    var frameSize: CGSize? {
+        if let yTexture = self.yTexture {
+            return CGSize(width: yTexture.width, height: yTexture.height)
+        } else {
+            return nil
+        }
+    }
+
+    func updateTextureBuffers(renderingContext: MetalVideoRenderingContext, frameBuffer: OngoingGroupCallContext.VideoFrameData.I420Buffer) {
+        let lumaSize = CGSize(width: frameBuffer.width, height: frameBuffer.height)
+
+        if lumaSize != lumaTextureDescriptorSize || lumaTextureDescriptor == nil || chromaTextureDescriptor == nil {
+            self.lumaTextureDescriptorSize = lumaSize
+
+            let lumaTextureDescriptor = MTLTextureDescriptor.texture2DDescriptor(pixelFormat: .r8Unorm, width: frameBuffer.width, height: frameBuffer.height, mipmapped: false)
+            lumaTextureDescriptor.usage = .shaderRead
+            self.lumaTextureDescriptor = lumaTextureDescriptor
+
+            self.yTexture = renderingContext.device.makeTexture(descriptor: lumaTextureDescriptor)
+
+            let chromaTextureDescriptor = MTLTextureDescriptor.texture2DDescriptor(pixelFormat: .r8Unorm, width: frameBuffer.width / 2, height: frameBuffer.height / 2, mipmapped: false)
+            chromaTextureDescriptor.usage = .shaderRead
+            self.chromaTextureDescriptor = chromaTextureDescriptor
+
+            self.uTexture = renderingContext.device.makeTexture(descriptor: chromaTextureDescriptor)
+            self.vTexture = renderingContext.device.makeTexture(descriptor: chromaTextureDescriptor)
+        }
+
+        guard let yTexture = self.yTexture, let uTexture = self.uTexture, let vTexture = self.vTexture else {
+            return
+        }
+
+        frameBuffer.y.withUnsafeBytes { bufferPointer in
+            if let baseAddress = bufferPointer.baseAddress {
+                yTexture.replace(region: MTLRegionMake2D(0, 0, yTexture.width, yTexture.height), mipmapLevel: 0, withBytes: baseAddress, bytesPerRow: frameBuffer.strideY)
+            }
+        }
+
+        frameBuffer.u.withUnsafeBytes { bufferPointer in
+            if let baseAddress = bufferPointer.baseAddress {
+                uTexture.replace(region: MTLRegionMake2D(0, 0, uTexture.width, uTexture.height), mipmapLevel: 0, withBytes: baseAddress, bytesPerRow: frameBuffer.strideU)
+            }
+        }
+
+        frameBuffer.v.withUnsafeBytes { bufferPointer in
+            if let baseAddress = bufferPointer.baseAddress {
+                vTexture.replace(region: MTLRegionMake2D(0, 0, vTexture.width, vTexture.height), mipmapLevel: 0, withBytes: baseAddress, bytesPerRow: frameBuffer.strideV)
+            }
+        }
+    }
+
+    func encode(renderingContext: MetalVideoRenderingContext, vertexBuffer: MTLBuffer, renderEncoder: MTLRenderCommandEncoder) -> Bool {
+        guard let yTexture = self.yTexture, let uTexture = self.uTexture, let vTexture = self.vTexture else {
+            return false
+        }
+
+        renderEncoder.setRenderPipelineState(renderingContext.i420PipelineState)
+
+        renderEncoder.setVertexBuffer(vertexBuffer, offset: 0, index: 0)
+
+        renderEncoder.setFragmentTexture(yTexture, index: 0)
+        renderEncoder.setFragmentTexture(uTexture, index: 1)
+        renderEncoder.setFragmentTexture(vTexture, index: 2)
+
+        renderEncoder.drawPrimitives(type: .triangleStrip, vertexStart: 0, vertexCount: 4, instanceCount: 1)
+
+        return true
+    }
+}
+
+@available(iOS 13.0, *)
+final class MetalVideoRenderingView: UIView, VideoRenderingView {
+    static override var layerClass: AnyClass {
+        return CAMetalLayer.self
+    }
+
+    private var metalLayer: CAMetalLayer {
+        return self.layer as! CAMetalLayer
+    }
+
+    private weak var renderingContext: MetalVideoRenderingContext?
+    private var renderingContextIndex: Int?
+
+    private let blur: Bool
+
+    private let vertexBuffer: MTLBuffer
+
+    private var frameBufferRenderingState: FrameBufferRenderingState?
+    private var blurInputTexture: MTLTexture?
+    private var blurOutputTexture: MTLTexture?
+
+    fileprivate private(set) var isEnabled: Bool = false
+    fileprivate var needsRedraw: Bool = false
+    fileprivate let numberOfUsedDrawables = Atomic<Int>(value: 0)
+
+    private var onFirstFrameReceived: ((Float) -> Void)?
+    private var onOrientationUpdated: ((PresentationCallVideoView.Orientation, CGFloat) -> Void)?
+    private var onIsMirroredUpdated: ((Bool) -> Void)?
+
+    private var didReportFirstFrame: Bool = false
+    private var currentOrientation: PresentationCallVideoView.Orientation = .rotation0
+    private var currentAspect: CGFloat = 1.0
+
+    private var disposable: Disposable?
+
+    init?(renderingContext: MetalVideoRenderingContext, input: Signal<OngoingGroupCallContext.VideoFrameData, NoError>, blur: Bool) {
+        self.renderingContext = renderingContext
+        self.blur = blur
+
+        let vertexBufferArray = Array<Float>(repeating: 0, count: 16)
+        guard let vertexBuffer = renderingContext.device.makeBuffer(bytes: vertexBufferArray, length: vertexBufferArray.count * MemoryLayout.size(ofValue: vertexBufferArray[0]), options: [.cpuCacheModeWriteCombined]) else {
+            return nil
+        }
+        self.vertexBuffer = vertexBuffer
+
+        super.init(frame: CGRect())
+
+        self.renderingContextIndex = renderingContext.add(view: self)
+
+        self.metalLayer.device = renderingContext.device
+        self.metalLayer.pixelFormat = .bgra8Unorm
+        self.metalLayer.framebufferOnly = true
+        self.metalLayer.allowsNextDrawableTimeout = true
+
+        self.disposable = input.start(next: { [weak self] videoFrameData in
+            Queue.mainQueue().async {
+                self?.addFrame(videoFrameData)
+            }
+        })
+    }
+
+    required init?(coder: NSCoder) {
+        fatalError("init(coder:) has not been implemented")
+    }
+
+    deinit {
+        self.disposable?.dispose()
+        if let renderingContext = self.renderingContext, let renderingContextIndex = self.renderingContextIndex {
+            renderingContext.remove(index: renderingContextIndex)
+        }
+    }
+
+    private func addFrame(_ videoFrameData: OngoingGroupCallContext.VideoFrameData) {
+        let aspect = CGFloat(videoFrameData.width) / CGFloat(videoFrameData.height)
+        var isAspectUpdated = false
+        if self.currentAspect != aspect {
+            self.currentAspect = aspect
+            isAspectUpdated = true
+        }
+
+        let videoFrameOrientation = PresentationCallVideoView.Orientation(videoFrameData.orientation)
+        var isOrientationUpdated = false
+        if self.currentOrientation != videoFrameOrientation {
+            self.currentOrientation = videoFrameOrientation
+            isOrientationUpdated = true
+        }
+
+        if isAspectUpdated || isOrientationUpdated {
+            self.onOrientationUpdated?(self.currentOrientation, self.currentAspect)
+        }
+
+        if !self.didReportFirstFrame {
+            self.didReportFirstFrame = true
+            self.onFirstFrameReceived?(Float(self.currentAspect))
+        }
+
+        if self.isEnabled, let renderingContext = self.renderingContext {
+            switch videoFrameData.buffer {
+            case let .native(buffer):
+                let renderingState: NV12FrameBufferRenderingState
+                if let current = self.frameBufferRenderingState as? NV12FrameBufferRenderingState {
+                    renderingState = current
+                } else {
+                    renderingState = NV12FrameBufferRenderingState()
+                    self.frameBufferRenderingState = renderingState
+                }
+                renderingState.updateTextureBuffers(renderingContext: renderingContext, frameBuffer: buffer)
+                self.needsRedraw = true
+            case let .i420(buffer):
+                let renderingState: I420FrameBufferRenderingState
+                if let current = self.frameBufferRenderingState as? I420FrameBufferRenderingState {
+                    renderingState = current
+                } else {
+                    renderingState = I420FrameBufferRenderingState()
+                    self.frameBufferRenderingState = renderingState
+                }
+                renderingState.updateTextureBuffers(renderingContext: renderingContext, frameBuffer: buffer)
+                self.needsRedraw = true
+            default:
+                break
+            }
+        }
+    }
+
+    fileprivate func encode(commandBuffer: MTLCommandBuffer) -> MTLDrawable? {
+        guard let renderingContext = self.renderingContext else {
+            return nil
+        }
+        if self.numberOfUsedDrawables.with({ $0 }) >= 2 {
+            return nil
+        }
+        guard let frameBufferRenderingState = self.frameBufferRenderingState else {
+            return nil
+        }
+
+        guard let frameSize = frameBufferRenderingState.frameSize else {
+            return nil
+        }
+
+        let drawableSize: CGSize
+        if self.blur {
+            drawableSize = frameSize.aspectFitted(CGSize(width: 64.0, height: 64.0))
+        } else {
+            drawableSize = frameSize
+        }
+
+        if self.blur {
+            if let current = self.blurInputTexture, current.width == Int(drawableSize.width) && current.height == Int(drawableSize.height) {
+            } else {
+                let blurTextureDescriptor = MTLTextureDescriptor.texture2DDescriptor(pixelFormat: .bgra8Unorm, width: Int(drawableSize.width), height: Int(drawableSize.height), mipmapped: false)
+                blurTextureDescriptor.usage = [.shaderRead, .shaderWrite, .renderTarget]
+
+                if let texture = renderingContext.device.makeTexture(descriptor: blurTextureDescriptor) {
+                    self.blurInputTexture = texture
+                }
+            }
+
+            if let current = self.blurOutputTexture, current.width == Int(drawableSize.width) && current.height == Int(drawableSize.height) {
+            } else {
+                let blurTextureDescriptor = MTLTextureDescriptor.texture2DDescriptor(pixelFormat: .bgra8Unorm, width: Int(drawableSize.width), height: Int(drawableSize.height), mipmapped: false)
+                blurTextureDescriptor.usage = [.shaderRead, .shaderWrite]
+
+                if let texture = renderingContext.device.makeTexture(descriptor: blurTextureDescriptor) {
+                    self.blurOutputTexture = texture
+                }
+            }
+        }
+
+        if self.metalLayer.drawableSize != drawableSize {
+            self.metalLayer.drawableSize = drawableSize
+
+            getCubeVertexData(
+                cropX: 0,
+                cropY: 0,
+                cropWidth: Int(drawableSize.width),
+                cropHeight: Int(drawableSize.height),
+                frameWidth: Int(drawableSize.width),
+                frameHeight: Int(drawableSize.height),
+                rotation: 0,
+                buffer: self.vertexBuffer.contents().assumingMemoryBound(to: Float.self)
+            )
+        }
+
+
+        guard let drawable = self.metalLayer.nextDrawable() else {
+            return nil
+        }
+
+        if let blurInputTexture = self.blurInputTexture, let blurOutputTexture = self.blurOutputTexture {
+            let renderPassDescriptor = MTLRenderPassDescriptor()
+            renderPassDescriptor.colorAttachments[0].texture = blurInputTexture
+            renderPassDescriptor.colorAttachments[0].loadAction = .clear
+            renderPassDescriptor.colorAttachments[0].clearColor = MTLClearColor(
+                red: 0.0,
+                green: 0.0,
+                blue: 0.0,
+                alpha: 1.0
+            )
+
+            guard let renderEncoder = commandBuffer.makeRenderCommandEncoder(descriptor: renderPassDescriptor) else {
+                return nil
+            }
+
+            let _ = frameBufferRenderingState.encode(renderingContext: renderingContext, vertexBuffer: self.vertexBuffer, renderEncoder: renderEncoder)
+
+            renderEncoder.endEncoding()
+
+            renderingContext.blurKernel.encode(commandBuffer: commandBuffer, sourceTexture: blurInputTexture, destinationTexture: blurOutputTexture)
+
+            let blitPassDescriptor = MTLRenderPassDescriptor()
+            blitPassDescriptor.colorAttachments[0].texture = drawable.texture
+            blitPassDescriptor.colorAttachments[0].loadAction = .clear
+            blitPassDescriptor.colorAttachments[0].clearColor = MTLClearColor(
+                red: 0.0,
+                green: 0.0,
+                blue: 0.0,
+                alpha: 1.0
+            )
+
+            guard let blitEncoder = commandBuffer.makeRenderCommandEncoder(descriptor: blitPassDescriptor) else {
+                return nil
+            }
+
+            let _ = BlitRenderingState.encode(renderingContext: renderingContext, texture: blurOutputTexture, vertexBuffer: self.vertexBuffer, renderEncoder: blitEncoder)
+
+            blitEncoder.endEncoding()
+        } else {
+            let renderPassDescriptor = MTLRenderPassDescriptor()
+            renderPassDescriptor.colorAttachments[0].texture = drawable.texture
+            renderPassDescriptor.colorAttachments[0].loadAction = .clear
+            renderPassDescriptor.colorAttachments[0].clearColor = MTLClearColor(
+                red: 0.0,
+                green: 0.0,
+                blue: 0.0,
+                alpha: 1.0
+            )
+
+            guard let renderEncoder = commandBuffer.makeRenderCommandEncoder(descriptor: renderPassDescriptor) else {
+                return nil
+            }
+
+            let _ = frameBufferRenderingState.encode(renderingContext: renderingContext, vertexBuffer: self.vertexBuffer, renderEncoder: renderEncoder)
+
+            renderEncoder.endEncoding()
+        }
+
+        return drawable
+    }
+
+    func setOnFirstFrameReceived(_ f: @escaping (Float) -> Void) {
+        self.onFirstFrameReceived = f
+        self.didReportFirstFrame = false
+    }
+
+    func setOnOrientationUpdated(_ f: @escaping (PresentationCallVideoView.Orientation, CGFloat) -> Void) {
+        self.onOrientationUpdated = f
+    }
+
+    func getOrientation() -> PresentationCallVideoView.Orientation {
+        return self.currentOrientation
+    }
+
+    func getAspect() -> CGFloat {
+        return self.currentAspect
+    }
+
+    func setOnIsMirroredUpdated(_ f: @escaping (Bool) -> Void) {
+        self.onIsMirroredUpdated = f
+    }
+
+    func updateIsEnabled(_ isEnabled: Bool) {
+        if self.isEnabled != isEnabled {
+            self.isEnabled = isEnabled
+
+            if self.isEnabled {
+                self.needsRedraw = true
+            }
+        }
+    }
+}
+
+@available(iOS 13.0, *)
+class MetalVideoRenderingContext {
+    private final class ViewReference {
+        weak var view: MetalVideoRenderingView?
+
+        init(view: MetalVideoRenderingView) {
+            self.view = view
+        }
+    }
+
+    fileprivate let device: MTLDevice
+    fileprivate let textureCache: CVMetalTextureCache
+    fileprivate let blurKernel: MPSImageGaussianBlur
+
+    fileprivate let blitPipelineState: MTLRenderPipelineState
+    fileprivate let nv12PipelineState: MTLRenderPipelineState
+    fileprivate let i420PipelineState: MTLRenderPipelineState
+
+    private let commandQueue: MTLCommandQueue
+
+    private var displayLink: ConstantDisplayLinkAnimator?
+    private var viewReferences = Bag<ViewReference>()
+
+    init?() {
+        guard let device = MTLCreateSystemDefaultDevice() else {
+            return nil
+        }
+        self.device = device
+
+        var textureCache: CVMetalTextureCache?
+        let _ = CVMetalTextureCacheCreate(kCFAllocatorDefault, nil, self.device, nil, &textureCache)
+        if let textureCache = textureCache {
+            self.textureCache = textureCache
+        } else {
+            return nil
+        }
+
+        let mainBundle = Bundle(for: MetalVideoRenderingView.self)
+
+        guard let path = mainBundle.path(forResource: "TelegramCallsUIBundle", ofType: "bundle") else {
+            return nil
+        }
+        guard let bundle = Bundle(path: path) else {
+            return nil
+        }
+        guard let defaultLibrary = try? self.device.makeDefaultLibrary(bundle: bundle) else {
+            return nil
+        }
+
+        self.blurKernel = MPSImageGaussianBlur(device: self.device, sigma: 3.0)
+
+        func makePipelineState(vertexProgram: String, fragmentProgram: String) -> MTLRenderPipelineState? {
+            guard let loadedVertexProgram = defaultLibrary.makeFunction(name: vertexProgram) else {
+                return nil
+            }
+            guard let loadedFragmentProgram = defaultLibrary.makeFunction(name: fragmentProgram) else {
+                return nil
+            }
+
+            let pipelineStateDescriptor = MTLRenderPipelineDescriptor()
+            pipelineStateDescriptor.vertexFunction = loadedVertexProgram
+            pipelineStateDescriptor.fragmentFunction = loadedFragmentProgram
+            pipelineStateDescriptor.colorAttachments[0].pixelFormat = .bgra8Unorm
+            guard let pipelineState = try? device.makeRenderPipelineState(descriptor: pipelineStateDescriptor) else {
+                return nil
+            }
+
+            return pipelineState
+        }
+
+        guard let blitPipelineState = makePipelineState(vertexProgram: "nv12VertexPassthrough", fragmentProgram: "blitFragmentColorConversion") else {
+            return nil
+        }
+        self.blitPipelineState = blitPipelineState
+
+        guard let nv12PipelineState = makePipelineState(vertexProgram: "nv12VertexPassthrough", fragmentProgram: "nv12FragmentColorConversion") else {
+            return nil
+        }
+        self.nv12PipelineState = nv12PipelineState
+
+        guard let i420PipelineState = makePipelineState(vertexProgram: "i420VertexPassthrough", fragmentProgram: "i420FragmentColorConversion") else {
+            return nil
+        }
+        self.i420PipelineState = i420PipelineState
+
+        guard let commandQueue = self.device.makeCommandQueue() else {
+            return nil
+        }
+        self.commandQueue = commandQueue
+        
+        self.displayLink = ConstantDisplayLinkAnimator(update: { [weak self] in
+            self?.redraw()
+        })
+        self.displayLink?.isPaused = false
+    }
+
+    func updateVisibility(isVisible: Bool) {
+        self.displayLink?.isPaused = !isVisible
+    }
+
+    fileprivate func add(view: MetalVideoRenderingView) -> Int {
+        return self.viewReferences.add(ViewReference(view: view))
+    }
+
+    fileprivate func remove(index: Int) {
+        self.viewReferences.remove(index)
+    }
+
+    private func redraw() {
+        guard let commandBuffer = self.commandQueue.makeCommandBuffer() else {
+            return
+        }
+
+        var drawables: [MTLDrawable] = []
+        var takenViewReferences: [ViewReference] = []
+
+        for viewReference in self.viewReferences.copyItems() {
+            guard let videoView = viewReference.view else {
+                continue
+            }
+
+            if !videoView.needsRedraw {
+                continue
+            }
+            videoView.needsRedraw = false
+
+            if let drawable = videoView.encode(commandBuffer: commandBuffer) {
+                let numberOfUsedDrawables = videoView.numberOfUsedDrawables
+                let _ = numberOfUsedDrawables.modify {
+                    return $0 + 1
+                }
+                takenViewReferences.append(viewReference)
+
+                drawable.addPresentedHandler { _ in
+                    let _ = numberOfUsedDrawables.modify {
+                        return max(0, $0 - 1)
+                    }
+                }
+
+                drawables.append(drawable)
+            }
+        }
+
+        if drawables.isEmpty {
+            return
+        }
+
+        if drawables.count > 10 {
+            print("Schedule \(drawables.count) drawables")
+        }
+
+        commandBuffer.addScheduledHandler { _ in
+            for drawable in drawables {
+                drawable.present()
+            }
+        }
+
+        commandBuffer.commit()
+    }
+}

submodules/TelegramCallsUI/Sources/PresentationGroupCall.swift

@@ -2595,6 +2595,15 @@ public final class PresentationGroupCallImpl: PresentationGroupCall {
             let videoCapturer = OngoingCallVideoCapturer()
             self.videoCapturer = videoCapturer
         }
+
+        if let videoCapturer = self.videoCapturer {
+            self.requestVideo(capturer: videoCapturer)
+        }
+    }
+
+    func requestVideo(capturer: OngoingCallVideoCapturer) {
+        self.videoCapturer = capturer
+        
         self.hasVideo = true
         if let videoCapturer = self.videoCapturer {
             self.genericCallContext?.requestVideo(videoCapturer)
@@ -3190,6 +3199,10 @@ public final class PresentationGroupCallImpl: PresentationGroupCall {
         })
     }
 
+    func video(endpointId: String) -> Signal<OngoingGroupCallContext.VideoFrameData, NoError>? {
+        return self.genericCallContext?.video(endpointId: endpointId)
+    }
+    
     public func loadMoreMembers(token: String) {
         self.participantsContext?.loadMore(token: token)
     }

submodules/TelegramCallsUI/Sources/SampleBufferVideoRenderingView.swift

@@ -0,0 +1,144 @@
+import Foundation
+import UIKit
+import AsyncDisplayKit
+import Display
+import SwiftSignalKit
+import AccountContext
+import TelegramVoip
+import AVFoundation
+
+private func sampleBufferFromPixelBuffer(pixelBuffer: CVPixelBuffer) -> CMSampleBuffer? {
+    var maybeFormat: CMVideoFormatDescription?
+    let status = CMVideoFormatDescriptionCreateForImageBuffer(allocator: kCFAllocatorDefault, imageBuffer: pixelBuffer, formatDescriptionOut: &maybeFormat)
+    if status != noErr {
+        return nil
+    }
+    guard let format = maybeFormat else {
+        return nil
+    }
+
+    var timingInfo = CMSampleTimingInfo(
+        duration: CMTimeMake(value: 1, timescale: 30),
+        presentationTimeStamp: CMTimeMake(value: 0, timescale: 30),
+        decodeTimeStamp: CMTimeMake(value: 0, timescale: 30)
+    )
+
+    var maybeSampleBuffer: CMSampleBuffer?
+    let bufferStatus = CMSampleBufferCreateReadyWithImageBuffer(allocator: kCFAllocatorDefault, imageBuffer: pixelBuffer, formatDescription: format, sampleTiming: &timingInfo, sampleBufferOut: &maybeSampleBuffer)
+
+    if (bufferStatus != noErr) {
+        return nil
+    }
+    guard let sampleBuffer = maybeSampleBuffer else {
+        return nil
+    }
+
+    let attachments: NSArray = CMSampleBufferGetSampleAttachmentsArray(sampleBuffer, createIfNecessary: true)! as NSArray
+    let dict: NSMutableDictionary = attachments[0] as! NSMutableDictionary
+    dict[kCMSampleAttachmentKey_DisplayImmediately as NSString] = true as NSNumber
+
+    return sampleBuffer
+}
+
+final class SampleBufferVideoRenderingView: UIView, VideoRenderingView {
+    static override var layerClass: AnyClass {
+        return AVSampleBufferDisplayLayer.self
+    }
+
+    private var sampleBufferLayer: AVSampleBufferDisplayLayer {
+        return self.layer as! AVSampleBufferDisplayLayer
+    }
+
+    private var isEnabled: Bool = false
+
+    private var onFirstFrameReceived: ((Float) -> Void)?
+    private var onOrientationUpdated: ((PresentationCallVideoView.Orientation, CGFloat) -> Void)?
+    private var onIsMirroredUpdated: ((Bool) -> Void)?
+
+    private var didReportFirstFrame: Bool = false
+    private var currentOrientation: PresentationCallVideoView.Orientation = .rotation0
+    private var currentAspect: CGFloat = 1.0
+
+    private var disposable: Disposable?
+
+    init(input: Signal<OngoingGroupCallContext.VideoFrameData, NoError>) {
+        super.init(frame: CGRect())
+
+        self.disposable = input.start(next: { [weak self] videoFrameData in
+            Queue.mainQueue().async {
+                self?.addFrame(videoFrameData)
+            }
+        })
+
+        self.sampleBufferLayer.videoGravity = .resize
+    }
+
+    required init?(coder: NSCoder) {
+        fatalError("init(coder:) has not been implemented")
+    }
+
+    deinit {
+        self.disposable?.dispose()
+    }
+
+    private func addFrame(_ videoFrameData: OngoingGroupCallContext.VideoFrameData) {
+        let aspect = CGFloat(videoFrameData.width) / CGFloat(videoFrameData.height)
+        var isAspectUpdated = false
+        if self.currentAspect != aspect {
+            self.currentAspect = aspect
+            isAspectUpdated = true
+        }
+
+        let videoFrameOrientation = PresentationCallVideoView.Orientation(videoFrameData.orientation)
+        var isOrientationUpdated = false
+        if self.currentOrientation != videoFrameOrientation {
+            self.currentOrientation = videoFrameOrientation
+            isOrientationUpdated = true
+        }
+
+        if isAspectUpdated || isOrientationUpdated {
+            self.onOrientationUpdated?(self.currentOrientation, self.currentAspect)
+        }
+
+        if !self.didReportFirstFrame {
+            self.didReportFirstFrame = true
+            self.onFirstFrameReceived?(Float(self.currentAspect))
+        }
+
+        if self.isEnabled {
+            switch videoFrameData.buffer {
+            case let .native(buffer):
+                if let sampleBuffer = sampleBufferFromPixelBuffer(pixelBuffer: buffer.pixelBuffer) {
+                    self.sampleBufferLayer.enqueue(sampleBuffer)
+                }
+            default:
+                break
+            }
+        }
+    }
+
+    func setOnFirstFrameReceived(_ f: @escaping (Float) -> Void) {
+        self.onFirstFrameReceived = f
+        self.didReportFirstFrame = false
+    }
+
+    func setOnOrientationUpdated(_ f: @escaping (PresentationCallVideoView.Orientation, CGFloat) -> Void) {
+        self.onOrientationUpdated = f
+    }
+
+    func getOrientation() -> PresentationCallVideoView.Orientation {
+        return self.currentOrientation
+    }
+
+    func getAspect() -> CGFloat {
+        return self.currentAspect
+    }
+
+    func setOnIsMirroredUpdated(_ f: @escaping (Bool) -> Void) {
+        self.onIsMirroredUpdated = f
+    }
+
+    func updateIsEnabled(_ isEnabled: Bool) {
+        self.isEnabled = isEnabled
+    }
+}

submodules/TelegramCallsUI/Sources/VideoRenderingContext.swift

@@ -0,0 +1,61 @@
+import Foundation
+import UIKit
+import AsyncDisplayKit
+import Display
+import SwiftSignalKit
+import AccountContext
+import TelegramVoip
+import AVFoundation
+
+protocol VideoRenderingView: UIView {
+    func setOnFirstFrameReceived(_ f: @escaping (Float) -> Void)
+    func setOnOrientationUpdated(_ f: @escaping (PresentationCallVideoView.Orientation, CGFloat) -> Void)
+    func getOrientation() -> PresentationCallVideoView.Orientation
+    func getAspect() -> CGFloat
+    func setOnIsMirroredUpdated(_ f: @escaping (Bool) -> Void)
+    func updateIsEnabled(_ isEnabled: Bool)
+}
+
+class VideoRenderingContext {
+    private var metalContextImpl: Any?
+
+    @available(iOS 13.0, *)
+    var metalContext: MetalVideoRenderingContext {
+        if let value = self.metalContextImpl as? MetalVideoRenderingContext {
+            return value
+        } else {
+            let value = MetalVideoRenderingContext()!
+            self.metalContextImpl = value
+            return value
+        }
+    }
+
+    func makeView(input: Signal<OngoingGroupCallContext.VideoFrameData, NoError>, blur: Bool) -> VideoRenderingView? {
+        if #available(iOS 13.0, *) {
+            return MetalVideoRenderingView(renderingContext: self.metalContext, input: input, blur: blur)
+        } else {
+            return SampleBufferVideoRenderingView(input: input)
+        }
+    }
+
+    func updateVisibility(isVisible: Bool) {
+        if #available(iOS 13.0, *) {
+            self.metalContext.updateVisibility(isVisible: isVisible)
+        }
+    }
+}
+
+extension PresentationCallVideoView.Orientation {
+    init(_ orientation: OngoingCallVideoOrientation) {
+        switch orientation {
+        case .rotation0:
+            self = .rotation0
+        case .rotation90:
+            self = .rotation90
+        case .rotation180:
+            self = .rotation180
+        case .rotation270:
+            self = .rotation270
+        }
+    }
+}

submodules/TelegramCallsUI/Sources/VoiceChatController.swift

@@ -918,6 +918,7 @@ public final class VoiceChatController: ViewController {
         private var requestedVideoSources = Set<String>()
         private var requestedVideoChannels: [PresentationGroupCallRequestedVideo] = []
 
+        private var videoRenderingContext: VideoRenderingContext
         private var videoNodes: [String: GroupVideoNode] = [:]
         private var wideVideoNodes = Set<String>()
         private var videoOrder: [String] = []
@@ -972,6 +973,8 @@ public final class VoiceChatController: ViewController {
             self.context = call.accountContext
             self.call = call
 
+            self.videoRenderingContext = VideoRenderingContext()
+            
             self.isScheduling = call.schedulePending
                         
             let presentationData = sharedContext.currentPresentationData.with { $0 }
@@ -2332,13 +2335,19 @@ public final class VoiceChatController: ViewController {
                             }))
                         }
                     } else {
-                        strongSelf.call.makeIncomingVideoView(endpointId: endpointId, requestClone: GroupVideoNode.useBlurTransparency, completion: { videoView, backdropVideoView in
+                        if let input = (strongSelf.call as! PresentationGroupCallImpl).video(endpointId: endpointId) {
+                            if let videoView = strongSelf.videoRenderingContext.makeView(input: input, blur: false) {
+                                completion(GroupVideoNode(videoView: videoView, backdropVideoView: strongSelf.videoRenderingContext.makeView(input: input, blur: true)))
+                            }
+                        }
+
+                        /*strongSelf.call.makeIncomingVideoView(endpointId: endpointId, requestClone: GroupVideoNode.useBlurTransparency, completion: { videoView, backdropVideoView in
                             if let videoView = videoView {
                                 completion(GroupVideoNode(videoView: videoView, backdropVideoView: backdropVideoView))
                             } else {
                                 completion(nil)
                             }
-                        })
+                        })*/
                     }
                 }
             }
@@ -3482,7 +3491,29 @@ public final class VoiceChatController: ViewController {
                     guard let strongSelf = self, ready else {
                         return
                     }
-                    strongSelf.call.makeOutgoingVideoView(requestClone: false, completion: { [weak self] view, _ in
+                    let videoCapturer = OngoingCallVideoCapturer()
+                    let input = videoCapturer.video()
+                    if let videoView = strongSelf.videoRenderingContext.makeView(input: input, blur: false) {
+                        let cameraNode = GroupVideoNode(videoView: videoView, backdropVideoView: nil)
+                        let controller = VoiceChatCameraPreviewController(context: strongSelf.context, cameraNode: cameraNode, shareCamera: { [weak self] _, unmuted in
+                            if let strongSelf = self {
+                                strongSelf.call.setIsMuted(action: unmuted ? .unmuted : .muted(isPushToTalkActive: false))
+                                (strongSelf.call as! PresentationGroupCallImpl).requestVideo(capturer: videoCapturer)
+
+                                if let (layout, navigationHeight) = strongSelf.validLayout {
+                                    strongSelf.animatingButtonsSwap = true
+                                    strongSelf.containerLayoutUpdated(layout, navigationHeight: navigationHeight, transition: .animated(duration: 0.4, curve: .spring))
+                                }
+                            }
+                        }, switchCamera: { [weak self] in
+                            Queue.mainQueue().after(0.1) {
+                                self?.call.switchVideoCamera()
+                            }
+                        })
+                        strongSelf.controller?.present(controller, in: .window(.root))
+                    }
+
+                    /*strongSelf.call.makeOutgoingVideoView(requestClone: false, completion: { [weak self] view, _ in
                         guard let strongSelf = self, let view = view else {
                             return
                         }
@@ -3503,7 +3534,7 @@ public final class VoiceChatController: ViewController {
                             }
                         })
                         strongSelf.controller?.present(controller, in: .window(.root))
-                    })
+                    })*/
                 })
             }
         }
@@ -4547,12 +4578,18 @@ public final class VoiceChatController: ViewController {
         
         private var appIsActive = true {
             didSet {
+                if self.appIsActive != oldValue {
                     self.updateVisibility()
+                    self.updateRequestedVideoChannels()
+                }
             }
         }
         private var visibility = false {
             didSet {
+                if self.visibility != oldValue {
                     self.updateVisibility()
+                    self.updateRequestedVideoChannels()
+                }
             }
         }
         
@@ -4574,6 +4611,8 @@ public final class VoiceChatController: ViewController {
                     itemNode.gridVisibility = visible
                 }
             }
+
+            self.videoRenderingContext.updateVisibility(isVisible: visible)
         }
         
         func animateIn() {
@@ -5201,7 +5240,63 @@ public final class VoiceChatController: ViewController {
 
                 if !self.requestedVideoSources.contains(channel.endpointId) {
                     self.requestedVideoSources.insert(channel.endpointId)
-                    self.call.makeIncomingVideoView(endpointId: channel.endpointId, requestClone: GroupVideoNode.useBlurTransparency, completion: { [weak self] videoView, backdropVideoView in
+
+                    let input = (self.call as! PresentationGroupCallImpl).video(endpointId: channel.endpointId)
+                    if let input = input, let videoView = self.videoRenderingContext.makeView(input: input, blur: false) {
+                        let videoNode = GroupVideoNode(videoView: videoView, backdropVideoView: self.videoRenderingContext.makeView(input: input, blur: true))
+
+                        self.readyVideoDisposables.set((combineLatest(videoNode.ready, .single(false) |> then(.single(true) |> delay(10.0, queue: Queue.mainQueue())))
+                        |> deliverOnMainQueue
+                        ).start(next: { [weak self, weak videoNode] ready, timeouted in
+                            if let strongSelf = self, let videoNode = videoNode {
+                                Queue.mainQueue().after(0.1) {
+                                    if timeouted && !ready {
+                                        strongSelf.timeoutedEndpointIds.insert(channel.endpointId)
+                                        strongSelf.readyVideoEndpointIds.remove(channel.endpointId)
+                                        strongSelf.readyVideoEndpointIdsPromise.set(strongSelf.readyVideoEndpointIds)
+                                        strongSelf.wideVideoNodes.remove(channel.endpointId)
+
+                                        strongSelf.updateMembers()
+                                    } else if ready {
+                                        strongSelf.readyVideoEndpointIds.insert(channel.endpointId)
+                                        strongSelf.readyVideoEndpointIdsPromise.set(strongSelf.readyVideoEndpointIds)
+                                        strongSelf.timeoutedEndpointIds.remove(channel.endpointId)
+                                        if videoNode.aspectRatio <= 0.77 {
+                                            strongSelf.wideVideoNodes.insert(channel.endpointId)
+                                        } else {
+                                            strongSelf.wideVideoNodes.remove(channel.endpointId)
+                                        }
+                                        strongSelf.updateMembers()
+
+                                        if let (layout, _) = strongSelf.validLayout, case .compact = layout.metrics.widthClass {
+                                            if let interaction = strongSelf.itemInteraction {
+                                                loop: for i in 0 ..< strongSelf.currentFullscreenEntries.count {
+                                                    let entry = strongSelf.currentFullscreenEntries[i]
+                                                    switch entry {
+                                                    case let .peer(peerEntry, _):
+                                                        if peerEntry.effectiveVideoEndpointId == channel.endpointId {
+                                                            let presentationData = strongSelf.presentationData.withUpdated(theme: strongSelf.darkTheme)
+                                                            strongSelf.fullscreenListNode.transaction(deleteIndices: [], insertIndicesAndItems: [], updateIndicesAndItems: [ListViewUpdateItem(index: i, previousIndex: i, item: entry.fullscreenItem(context: strongSelf.context, presentationData: presentationData, interaction: interaction), directionHint: nil)], options: [.Synchronous], updateOpaqueState: nil)
+                                                            break loop
+                                                        }
+                                                    default:
+                                                        break
+                                                    }
+                                                }
+                                            }
+                                        }
+                                    }
+                                }
+                            }
+                        }), forKey: channel.endpointId)
+                        self.videoNodes[channel.endpointId] = videoNode
+
+                        if let _ = self.validLayout {
+                            self.updateMembers()
+                        }
+                    }
+
+                    /*self.call.makeIncomingVideoView(endpointId: channel.endpointId, requestClone: GroupVideoNode.useBlurTransparency, completion: { [weak self] videoView, backdropVideoView in
                         Queue.mainQueue().async {
                             guard let strongSelf = self, let videoView = videoView else {
                                 return
@@ -5258,7 +5353,7 @@ public final class VoiceChatController: ViewController {
                                 strongSelf.updateMembers()
                             }
                         }
-                    })
+                    })*/
                 }
             }
 
@@ -5366,7 +5461,9 @@ public final class VoiceChatController: ViewController {
 
         private func updateRequestedVideoChannels() {
             Queue.mainQueue().after(0.3) {
-                self.call.setRequestedVideoList(items: self.requestedVideoChannels)
+                let enableVideo = self.appIsActive && self.visibility
+
+                self.call.setRequestedVideoList(items: enableVideo ? self.requestedVideoChannels : [])
                 self.filterRequestedVideoChannels(channels: self.requestedVideoChannels)
             }
         }

submodules/TelegramUI/Sources/PeerInfo/PeerInfoScreen.swift

@@ -2354,7 +2354,9 @@ private final class PeerInfoScreenNode: ViewControllerTracingNode, UIScrollViewD
                 self?.updateProfileVideo(image, asset: asset, adjustments: adjustments)
             }
             galleryController.removedEntry = { [weak self] entry in
-                let _ = self?.headerNode.avatarListNode.listContainerNode.deleteItem(PeerInfoAvatarListItem(entry: entry))
+                if let item = PeerInfoAvatarListItem(entry: entry) {
+                    let _ = self?.headerNode.avatarListNode.listContainerNode.deleteItem(item)
+                }
             }
             strongSelf.hiddenAvatarRepresentationDisposable.set((galleryController.hiddenMedia |> deliverOnMainQueue).start(next: { entry in
                 self?.headerNode.updateAvatarIsHidden(entry: entry)

submodules/TelegramVoip/Sources/GroupCallContext.swift

@@ -219,6 +219,115 @@ public final class OngoingGroupCallContext {
         }
     }
 
+    public final class VideoFrameData {
+        public final class NativeBuffer {
+            public let pixelBuffer: CVPixelBuffer
+
+            init(pixelBuffer: CVPixelBuffer) {
+                self.pixelBuffer = pixelBuffer
+            }
+        }
+
+        public final class NV12Buffer {
+            private let wrapped: CallVideoFrameNV12Buffer
+
+            public var width: Int {
+                return Int(self.wrapped.width)
+            }
+
+            public var height: Int {
+                return Int(self.wrapped.height)
+            }
+
+            public var y: Data {
+                return self.wrapped.y
+            }
+
+            public var strideY: Int {
+                return Int(self.wrapped.strideY)
+            }
+
+            public var uv: Data {
+                return self.wrapped.uv
+            }
+
+            public var strideUV: Int {
+                return Int(self.wrapped.strideUV)
+            }
+
+            init(wrapped: CallVideoFrameNV12Buffer) {
+                self.wrapped = wrapped
+            }
+        }
+
+        public final class I420Buffer {
+            private let wrapped: CallVideoFrameI420Buffer
+
+            public var width: Int {
+                return Int(self.wrapped.width)
+            }
+
+            public var height: Int {
+                return Int(self.wrapped.height)
+            }
+
+            public var y: Data {
+                return self.wrapped.y
+            }
+
+            public var strideY: Int {
+                return Int(self.wrapped.strideY)
+            }
+
+            public var u: Data {
+                return self.wrapped.u
+            }
+
+            public var strideU: Int {
+                return Int(self.wrapped.strideU)
+            }
+
+            public var v: Data {
+                return self.wrapped.v
+            }
+
+            public var strideV: Int {
+                return Int(self.wrapped.strideV)
+            }
+
+            init(wrapped: CallVideoFrameI420Buffer) {
+                self.wrapped = wrapped
+            }
+        }
+
+        public enum Buffer {
+            case native(NativeBuffer)
+            case nv12(NV12Buffer)
+            case i420(I420Buffer)
+        }
+
+        public let buffer: Buffer
+        public let width: Int
+        public let height: Int
+        public let orientation: OngoingCallVideoOrientation
+
+        init(frameData: CallVideoFrameData) {
+            if let nativeBuffer = frameData.buffer as? CallVideoFrameNativePixelBuffer {
+                self.buffer = .native(NativeBuffer(pixelBuffer: nativeBuffer.pixelBuffer))
+            } else if let nv12Buffer = frameData.buffer as? CallVideoFrameNV12Buffer {
+                self.buffer = .nv12(NV12Buffer(wrapped: nv12Buffer))
+            } else if let i420Buffer = frameData.buffer as? CallVideoFrameI420Buffer {
+                self.buffer = .i420(I420Buffer(wrapped: i420Buffer))
+            } else {
+                preconditionFailure()
+            }
+
+            self.width = Int(frameData.width)
+            self.height = Int(frameData.height)
+            self.orientation = OngoingCallVideoOrientation(frameData.orientation)
+        }
+    }
+    
     private final class Impl {
         let queue: Queue
         let context: GroupCallThreadLocalContext
@@ -615,6 +724,27 @@ public final class OngoingGroupCallContext {
             })
         }
 
+        func video(endpointId: String) -> Signal<OngoingGroupCallContext.VideoFrameData, NoError> {
+            let queue = self.queue
+            return Signal { [weak self] subscriber in
+                let disposable = MetaDisposable()
+
+                queue.async {
+                    guard let strongSelf = self else {
+                        return
+                    }
+                    let innerDisposable = strongSelf.context.addVideoOutput(withEndpointId: endpointId) { videoFrameData in
+                        subscriber.putNext(OngoingGroupCallContext.VideoFrameData(frameData: videoFrameData))
+                    }
+                    disposable.set(ActionDisposable {
+                        innerDisposable.dispose()
+                    })
+                }
+
+                return disposable
+            }
+        }
+
         func addExternalAudioData(data: Data) {
             self.context.addExternalAudioData(data)
         }
@@ -778,6 +908,18 @@ public final class OngoingGroupCallContext {
         }
     }
 
+    public func video(endpointId: String) -> Signal<OngoingGroupCallContext.VideoFrameData, NoError> {
+        return Signal { subscriber in
+            let disposable = MetaDisposable()
+            self.impl.with { impl in
+                disposable.set(impl.video(endpointId: endpointId).start(next: { value in
+                    subscriber.putNext(value)
+                }))
+            }
+            return disposable
+        }
+    }
+
     public func addExternalAudioData(data: Data) {
         self.impl.with { impl in
             impl.addExternalAudioData(data: data)

submodules/TelegramVoip/Sources/OngoingCallContext.swift

@@ -451,6 +451,27 @@ public final class OngoingCallVideoCapturer {
         }
         self.impl.submitPixelBuffer(pixelBuffer, rotation: videoRotation.orientation)
     }
+
+    public func video() -> Signal<OngoingGroupCallContext.VideoFrameData, NoError> {
+        let queue = Queue.mainQueue()
+        return Signal { [weak self] subscriber in
+            let disposable = MetaDisposable()
+
+            queue.async {
+                guard let strongSelf = self else {
+                    return
+                }
+                let innerDisposable = strongSelf.impl.addVideoOutput { videoFrameData in
+                    subscriber.putNext(OngoingGroupCallContext.VideoFrameData(frameData: videoFrameData))
+                }
+                disposable.set(ActionDisposable {
+                    innerDisposable.dispose()
+                })
+            }
+
+            return disposable
+        }
+    }
 }
 
 extension OngoingCallThreadLocalContextWebrtc: OngoingCallThreadLocalContextProtocol {

submodules/TgVoipWebrtc/PublicHeaders/TgVoipWebrtc/OngoingCallThreadLocalContext.h

@@ -110,6 +110,60 @@ typedef NS_ENUM(int32_t, OngoingCallDataSavingWebrtc) {
 #endif
 @end
 
+@interface GroupCallDisposable : NSObject
+
+- (void)dispose;
+
+@end
+
+@protocol CallVideoFrameBuffer
+
+@end
+
+@interface CallVideoFrameNativePixelBuffer : NSObject<CallVideoFrameBuffer>
+
+@property (nonatomic, readonly) CVPixelBufferRef _Nonnull pixelBuffer;
+
+@end
+
+@interface CallVideoFrameNV12Buffer : NSObject<CallVideoFrameBuffer>
+
+@property (nonatomic, readonly) int width;
+@property (nonatomic, readonly) int height;
+
+@property (nonatomic, strong, readonly) NSData * _Nonnull y;
+@property (nonatomic, readonly) int strideY;
+
+@property (nonatomic, strong, readonly) NSData * _Nonnull uv;
+@property (nonatomic, readonly) int strideUV;
+
+@end
+
+@interface CallVideoFrameI420Buffer : NSObject<CallVideoFrameBuffer>
+
+@property (nonatomic, readonly) int width;
+@property (nonatomic, readonly) int height;
+
+@property (nonatomic, strong, readonly) NSData * _Nonnull y;
+@property (nonatomic, readonly) int strideY;
+
+@property (nonatomic, strong, readonly) NSData * _Nonnull u;
+@property (nonatomic, readonly) int strideU;
+
+@property (nonatomic, strong, readonly) NSData * _Nonnull v;
+@property (nonatomic, readonly) int strideV;
+
+@end
+
+@interface CallVideoFrameData : NSObject
+
+@property (nonatomic, strong, readonly) id<CallVideoFrameBuffer> _Nonnull buffer;
+@property (nonatomic, readonly) int width;
+@property (nonatomic, readonly) int height;
+@property (nonatomic, readonly) OngoingCallVideoOrientationWebrtc orientation;
+
+@end
+
 @interface OngoingCallThreadLocalContextVideoCapturer : NSObject
 
 - (instancetype _Nonnull)initWithDeviceId:(NSString * _Nonnull)deviceId keepLandscape:(bool)keepLandscape;
@@ -131,6 +185,8 @@ typedef NS_ENUM(int32_t, OngoingCallDataSavingWebrtc) {
 - (void)submitPixelBuffer:(CVPixelBufferRef _Nonnull)pixelBuffer rotation:(OngoingCallVideoOrientationWebrtc)rotation;
 #endif
 
+- (GroupCallDisposable * _Nonnull)addVideoOutput:(void (^_Nonnull)(CallVideoFrameData * _Nonnull))sink;
+
 @end
 
 @interface OngoingCallThreadLocalContextWebrtc : NSObject
@@ -291,6 +347,7 @@ typedef NS_ENUM(int32_t, OngoingGroupCallRequestedVideoQuality) {
 - (void)switchAudioOutput:(NSString * _Nonnull)deviceId;
 - (void)switchAudioInput:(NSString * _Nonnull)deviceId;
 - (void)makeIncomingVideoViewWithEndpointId:(NSString * _Nonnull)endpointId requestClone:(bool)requestClone completion:(void (^_Nonnull)(UIView<OngoingCallThreadLocalContextWebrtcVideoView> * _Nullable, UIView<OngoingCallThreadLocalContextWebrtcVideoView> * _Nullable))completion;
+- (GroupCallDisposable * _Nonnull)addVideoOutputWithEndpointId:(NSString * _Nonnull)endpointId sink:(void (^_Nonnull)(CallVideoFrameData * _Nonnull))sink;
 
 - (void)addExternalAudioData:(NSData * _Nonnull)data;
 

submodules/TgVoipWebrtc/Sources/OngoingCallThreadLocalContext.mm

@@ -32,6 +32,9 @@
 
 #import "VideoCaptureInterfaceImpl.h"
 
+#include "sdk/objc/native/src/objc_frame_buffer.h"
+#import "components/video_frame_buffer/RTCCVPixelBuffer.h"
+
 @implementation OngoingCallConnectionDescriptionWebrtc
 
 - (instancetype _Nonnull)initWithConnectionId:(int64_t)connectionId hasStun:(bool)hasStun hasTurn:(bool)hasTurn ip:(NSString * _Nonnull)ip port:(int32_t)port username:(NSString * _Nonnull)username password:(NSString * _Nonnull)password {
@@ -221,10 +224,216 @@
 
 @end
 
+@interface GroupCallDisposable () {
+    dispatch_block_t _block;
+}
+
+@end
+
+@implementation GroupCallDisposable
+
+- (instancetype)initWithBlock:(dispatch_block_t _Nonnull)block {
+    self = [super init];
+    if (self != nil) {
+        _block = [block copy];
+    }
+    return self;
+}
+
+- (void)dispose {
+    if (_block) {
+        _block();
+    }
+}
+
+@end
+
+@implementation CallVideoFrameNativePixelBuffer
+
+- (instancetype)initWithPixelBuffer:(CVPixelBufferRef)pixelBuffer {
+    self = [super init];
+    if (self != nil) {
+        assert(pixelBuffer != nil);
+
+        _pixelBuffer = CVPixelBufferRetain(pixelBuffer);
+    }
+    return self;
+}
+
+- (void)dealloc {
+    CVPixelBufferRelease(_pixelBuffer);
+}
+
+@end
+
+@implementation CallVideoFrameNV12Buffer
+
+- (instancetype)initWithBuffer:(rtc::scoped_refptr<webrtc::NV12BufferInterface>)nv12Buffer {
+    self = [super init];
+    if (self != nil) {
+        _width = nv12Buffer->width();
+        _height = nv12Buffer->height();
+
+        _strideY = nv12Buffer->StrideY();
+        _strideUV = nv12Buffer->StrideUV();
+
+        _y = [[NSData alloc] initWithBytesNoCopy:(void *)nv12Buffer->DataY() length:nv12Buffer->StrideY() * _height deallocator:^(__unused void * _Nonnull bytes, __unused NSUInteger length) {
+            nv12Buffer.get();
+        }];
+
+        _uv = [[NSData alloc] initWithBytesNoCopy:(void *)nv12Buffer->DataUV() length:nv12Buffer->StrideUV() * _height deallocator:^(__unused void * _Nonnull bytes, __unused NSUInteger length) {
+            nv12Buffer.get();
+        }];
+    }
+    return self;
+}
+
+@end
+
+@implementation CallVideoFrameI420Buffer
+
+- (instancetype)initWithBuffer:(rtc::scoped_refptr<webrtc::I420BufferInterface>)i420Buffer {
+    self = [super init];
+    if (self != nil) {
+        _width = i420Buffer->width();
+        _height = i420Buffer->height();
+
+        _strideY = i420Buffer->StrideY();
+        _strideU = i420Buffer->StrideU();
+        _strideV = i420Buffer->StrideV();
+
+        _y = [[NSData alloc] initWithBytesNoCopy:(void *)i420Buffer->DataY() length:i420Buffer->StrideY() * _height deallocator:^(__unused void * _Nonnull bytes, __unused NSUInteger length) {
+            i420Buffer.get();
+        }];
+
+        _u = [[NSData alloc] initWithBytesNoCopy:(void *)i420Buffer->DataU() length:i420Buffer->StrideU() * _height deallocator:^(__unused void * _Nonnull bytes, __unused NSUInteger length) {
+            i420Buffer.get();
+        }];
+
+        _v = [[NSData alloc] initWithBytesNoCopy:(void *)i420Buffer->DataV() length:i420Buffer->StrideV() * _height deallocator:^(__unused void * _Nonnull bytes, __unused NSUInteger length) {
+            i420Buffer.get();
+        }];
+    }
+    return self;
+}
+
+@end
+
+@interface CallVideoFrameData () {
+}
+
+@end
+
+@implementation CallVideoFrameData
+
+- (instancetype)initWithBuffer:(id<CallVideoFrameBuffer>)buffer frame:(webrtc::VideoFrame const &)frame {
+    self = [super init];
+    if (self != nil) {
+        _buffer = buffer;
+
+        _width = frame.width();
+        _height = frame.height();
+
+        switch (frame.rotation()) {
+            case webrtc::kVideoRotation_0: {
+                _orientation = OngoingCallVideoOrientation0;
+                break;
+            }
+            case webrtc::kVideoRotation_90: {
+                _orientation = OngoingCallVideoOrientation90;
+                break;
+            }
+            case webrtc::kVideoRotation_180: {
+                _orientation = OngoingCallVideoOrientation180;
+                break;
+            }
+            case webrtc::kVideoRotation_270: {
+                _orientation = OngoingCallVideoOrientation270;
+                break;
+            }
+            default: {
+                _orientation = OngoingCallVideoOrientation0;
+                break;
+            }
+        }
+    }
+    return self;
+}
+
+@end
+
+namespace {
+
+class GroupCallVideoSinkAdapter : public rtc::VideoSinkInterface<webrtc::VideoFrame> {
+public:
+    GroupCallVideoSinkAdapter(void (^frameReceived)(webrtc::VideoFrame const &)) {
+        _frameReceived = [frameReceived copy];
+    }
+
+    void OnFrame(const webrtc::VideoFrame& nativeVideoFrame) override {
+        @autoreleasepool {
+            if (_frameReceived) {
+                _frameReceived(nativeVideoFrame);
+            }
+        }
+    }
+
+private:
+    void (^_frameReceived)(webrtc::VideoFrame const &);
+};
+
+}
+
+@interface GroupCallVideoSink : NSObject {
+    std::shared_ptr<GroupCallVideoSinkAdapter> _adapter;
+}
+
+@end
+
+@implementation GroupCallVideoSink
+
+- (instancetype)initWithSink:(void (^_Nonnull)(CallVideoFrameData * _Nonnull))sink {
+    self = [super init];
+    if (self != nil) {
+        void (^storedSink)(CallVideoFrameData * _Nonnull) = [sink copy];
+
+        _adapter.reset(new GroupCallVideoSinkAdapter(^(webrtc::VideoFrame const &videoFrame) {
+            id<CallVideoFrameBuffer> mappedBuffer = nil;
+
+            if (videoFrame.video_frame_buffer()->type() == webrtc::VideoFrameBuffer::Type::kNative) {
+                id<RTC_OBJC_TYPE(RTCVideoFrameBuffer)> nativeBuffer = static_cast<webrtc::ObjCFrameBuffer *>(videoFrame.video_frame_buffer().get())->wrapped_frame_buffer();
+                if ([nativeBuffer isKindOfClass:[RTC_OBJC_TYPE(RTCCVPixelBuffer) class]]) {
+                    RTCCVPixelBuffer *pixelBuffer = (RTCCVPixelBuffer *)nativeBuffer;
+                    mappedBuffer = [[CallVideoFrameNativePixelBuffer alloc] initWithPixelBuffer:pixelBuffer.pixelBuffer];
+                }
+            } else if (videoFrame.video_frame_buffer()->type() == webrtc::VideoFrameBuffer::Type::kNV12) {
+                rtc::scoped_refptr<webrtc::NV12BufferInterface> nv12Buffer = (webrtc::NV12BufferInterface *)videoFrame.video_frame_buffer().get();
+                mappedBuffer = [[CallVideoFrameNV12Buffer alloc] initWithBuffer:nv12Buffer];
+            } else if (videoFrame.video_frame_buffer()->type() == webrtc::VideoFrameBuffer::Type::kI420) {
+                rtc::scoped_refptr<webrtc::I420BufferInterface> i420Buffer = (webrtc::I420BufferInterface *)videoFrame.video_frame_buffer().get();
+                mappedBuffer = [[CallVideoFrameI420Buffer alloc] initWithBuffer:i420Buffer];
+            }
+
+            if (storedSink && mappedBuffer) {
+                storedSink([[CallVideoFrameData alloc] initWithBuffer:mappedBuffer frame:videoFrame]);
+            }
+        }));
+    }
+    return self;
+}
+
+- (std::shared_ptr<rtc::VideoSinkInterface<webrtc::VideoFrame>>)sink {
+    return _adapter;
+}
+
+@end
 
 @interface OngoingCallThreadLocalContextVideoCapturer () {
     bool _keepLandscape;
     std::shared_ptr<std::vector<uint8_t>> _croppingBuffer;
+
+    int _nextSinkId;
+    NSMutableDictionary<NSNumber *, GroupCallVideoSink *> *_sinks;
 }
 
 @end
@@ -312,6 +521,32 @@ tgcalls::VideoCaptureInterfaceObject *GetVideoCaptureAssumingSameThread(tgcalls:
 
 #endif
 
+- (GroupCallDisposable * _Nonnull)addVideoOutput:(void (^_Nonnull)(CallVideoFrameData * _Nonnull))sink {
+    int sinkId = _nextSinkId;
+    _nextSinkId += 1;
+
+    GroupCallVideoSink *storedSink = [[GroupCallVideoSink alloc] initWithSink:sink];
+    _sinks[@(sinkId)] = storedSink;
+
+    auto sinkReference = [storedSink sink];
+
+    tgcalls::StaticThreads::getThreads()->getMediaThread()->PostTask(RTC_FROM_HERE, [interface = _interface, sinkReference]() {
+        interface->setOutput(sinkReference);
+    });
+
+    __weak OngoingCallThreadLocalContextVideoCapturer *weakSelf = self;
+    return [[GroupCallDisposable alloc] initWithBlock:^{
+        dispatch_async(dispatch_get_main_queue(), ^{
+            __strong OngoingCallThreadLocalContextVideoCapturer *strongSelf = weakSelf;
+            if (!strongSelf) {
+                return;
+            }
+
+            [strongSelf->_sinks removeObjectForKey:@(sinkId)];
+        });
+    }];
+}
+
 - (void)switchVideoInput:(NSString * _Nonnull)deviceId {
     std::string resolvedId = deviceId.UTF8String;
     if (_keepLandscape) {
@@ -1076,6 +1311,9 @@ private:
     OngoingCallThreadLocalContextVideoCapturer *_videoCapturer;
 
     void (^_networkStateUpdated)(GroupCallNetworkState);
+
+    int _nextSinkId;
+    NSMutableDictionary<NSNumber *, GroupCallVideoSink *> *_sinks;
 }
 
 @end
@@ -1097,6 +1335,8 @@ private:
     if (self != nil) {
         _queue = queue;
 
+        _sinks = [[NSMutableDictionary alloc] init];
+        
         _networkStateUpdated = [networkStateUpdated copy];
         _videoCapturer = videoCapturer;
         
@@ -1477,6 +1717,31 @@ private:
     }
 }
 
+- (GroupCallDisposable * _Nonnull)addVideoOutputWithEndpointId:(NSString * _Nonnull)endpointId sink:(void (^_Nonnull)(CallVideoFrameData * _Nonnull))sink {
+    int sinkId = _nextSinkId;
+    _nextSinkId += 1;
+
+    GroupCallVideoSink *storedSink = [[GroupCallVideoSink alloc] initWithSink:sink];
+    _sinks[@(sinkId)] = storedSink;
+
+    if (_instance) {
+        _instance->addIncomingVideoOutput(endpointId.UTF8String, [storedSink sink]);
+    }
+
+    __weak GroupCallThreadLocalContext *weakSelf = self;
+    id<OngoingCallThreadLocalContextQueueWebrtc> queue = _queue;
+    return [[GroupCallDisposable alloc] initWithBlock:^{
+        [queue dispatch:^{
+            __strong GroupCallThreadLocalContext *strongSelf = weakSelf;
+            if (!strongSelf) {
+                return;
+            }
+
+            [strongSelf->_sinks removeObjectForKey:@(sinkId)];
+        }];
+    }];
+}
+
 - (void)addExternalAudioData:(NSData * _Nonnull)data {
     if (_instance) {
         std::vector<uint8_t> samples;

submodules/TgVoipWebrtc/tgcalls

@@ -1 +1 @@
-Subproject commit ed88e80ae5cc57538c4f515496d4393e7d84e685
+Subproject commit ef796349808b187b80b75ea1876b940f2882fcbb