feat(moq-video): native per-platform H.264 codecs + zero-copy capture, drop ffmpeg (#1691)

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: Jakub Perżyło <jakub.perzylo@swmansion.com>

Author: 3 people

Cargo.lock

@@ -26,12 +26,16 @@ pub enum PublishFormat {
 #[cfg(feature = "capture")]
 #[derive(clap::Args, Clone)]
 pub struct CaptureArgs {
-	/// Camera device. Platform-specific: an avfoundation index/name on macOS,
-	/// a `/dev/videoN` path on Linux, or a dshow device name on Windows.
-	/// Omit to use the default camera.
-	#[arg(long)]
+	/// Camera device. Platform-specific: an AVFoundation `uniqueID` on macOS, or
+	/// a camera index / `/dev/videoN` path on Linux. Omit to use the default
+	/// camera. Ignored with `--screen`.
+	#[arg(long, conflicts_with = "screen")]
 	pub camera: Option<String>,
 
+	/// Capture a display (whole screen) instead of a camera. macOS only.
+	#[arg(long)]
+	pub screen: bool,
+
 	/// Requested capture width. The camera snaps to its nearest supported mode.
 	#[arg(long)]
 	pub width: Option<u32>,
@@ -52,7 +56,7 @@ pub struct CaptureArgs {
 	#[arg(long, conflicts_with = "software")]
 	pub hardware: bool,
 
-	/// Force the software encoder (libx264).
+	/// Force the software encoder (openh264).
 	#[arg(long)]
 	pub software: bool,
 
@@ -99,7 +103,7 @@ enum Source {
 	/// Decode a container read from stdin (or an HLS playlist).
 	Stream(PublishDecoder),
 	/// Capture from local devices. The per-medium producers are built on their
-	/// own capture threads (camera via ffmpeg, microphone via cpal), publishing
+	/// own capture threads (native camera/screen capture, microphone via cpal), publishing
 	/// onto the shared broadcast + catalog; [`Publish::run`] drives them
 	/// concurrently.
 	#[cfg(feature = "capture")]
@@ -220,7 +224,11 @@ impl Publish {
 impl CaptureArgs {
 	fn video_config(&self) -> moq_video::capture::Config {
 		let mut config = moq_video::capture::Config::default();
-		config.device = self.camera.clone();
+		if self.screen {
+			config.source = moq_video::capture::Source::Display;
+		} else {
+			config.device = self.camera.clone();
+		}
 		config.width = self.width;
 		config.height = self.height;
 		config.framerate = self.fps;

rs/moq-cli/src/publish.rs

@@ -15,17 +15,55 @@ categories = ["multimedia", "multimedia::video", "multimedia::encoding"]
 [lib]
 doctest = false
 
+[features]
+# NVIDIA NVENC hardware encoder (Linux only). Off by default; release builds
+# for Linux enable it. cudarc loads CUDA dynamically (no build-time CUDA), and
+# the NVENC API lives in the driver, so this is not a build dependency.
+nvenc = ["dep:nvidia-video-codec-sdk", "cudarc/dynamic-loading", "cudarc/cuda-12020"]
+# Intel/AMD VAAPI hardware encoder (Linux only). Off by default; release builds
+# for Linux enable it. Links libva (stable soname); pairs with a V4L2 dmabuf
+# capture for a fully zero-copy path.
+vaapi = ["dep:cros-codecs"]
+
 [dependencies]
 anyhow = "1"
 bytes = "1"
-# H.264 encode plus webcam capture (libavdevice). One system dependency
-# covers both, and unlike a pure-Rust encoder it exposes the platform
-# hardware encoders (h264_videotoolbox / h264_nvenc / h264_vaapi) we want
-# for real-time capture. The `device` feature pulls in libavdevice for
-# avfoundation / v4l2 / dshow camera input.
-ffmpeg-next = { version = "8", features = ["device"] }
+# VAAPI H.264 encoder + dmabuf import. Gated behind `vaapi` + a linux cfg; pinned
+# (pre-1.0 churns). Pulls libva (cros-libva) when the feature is on.
+cros-codecs = { version = "=0.0.6", optional = true, default-features = false, features = ["vaapi"] }
+cudarc = { version = "0.16", optional = true, default-features = false, features = ["driver"] }
 moq-mux = { workspace = true }
 moq-net = { workspace = true }
+# Pure-Rust webcam capture (avfoundation / v4l2 / msmf), decoding MJPEG/YUYV to
+# RGBA. Platform input backends are enabled per-target below.
+nokhwa = { version = "0.10.11", default-features = false, features = ["decoding"] }
+# NVENC backend, gated behind the `nvenc` feature + a linux cfg. Listed here
+# (not under a target table) so the feature can reference it on any host; the
+# code that uses it is `cfg(target_os = "linux")`.
+nvidia-video-codec-sdk = { version = "0.4", optional = true }
+# Vendored, statically-linked software H.264 fallback (no system dependency).
+openh264 = "0.9.3"
 thiserror = "2"
 tokio = { workspace = true, features = ["rt", "macros"] }
 tracing = "0.1"
+# RGBA -> I420 color conversion (replaces ffmpeg swscale).
+yuv = "0.8.14"
+
+[target.'cfg(target_os="linux")'.dependencies]
+nokhwa = { version = "0.10.11", features = ["input-v4l"] }
+
+[target.'cfg(target_os="macos")'.dependencies]
+block2 = "0.6.2"
+dispatch2 = "0.3.1"
+nokhwa = { version = "0.10.11", features = ["input-avfoundation"] }
+objc2 = "0.6.4"
+objc2-av-foundation = "0.3.2"
+objc2-core-foundation = "0.3.2"
+objc2-core-media = "0.3.2"
+objc2-core-video = "0.3.2"
+objc2-foundation = "0.3.2"
+objc2-screen-capture-kit = "0.3.2"
+objc2-video-toolbox = "0.3.2"
+
+[target.'cfg(target_os="windows")'.dependencies]
+nokhwa = { version = "0.10.11", features = ["input-msmf"] }