fake-capture-device で nv12 を出力するように変更する (#422)

* fake-capture-device で nv12 を出力するように変更する

* formatter

* コメントを修正

* blend2d 0.20.0 追従

---------

Co-authored-by: melpon <[email protected]>

Author: voluntas

src/main.cpp

@@ -153,6 +153,7 @@ int main(int argc, char* argv[]) {
           video_config.width = size.width;
           video_config.height = size.height;
           video_config.fps = args.framerate;
+          video_config.force_nv12 = args.force_nv12;
           return FakeVideoCapturer::Create(video_config);
         }
 #endif

src/rtc/fake_video_capturer.cpp

@@ -9,6 +9,7 @@
 #include <sstream>
 
 // WebRTC
+#include <api/video/nv12_buffer.h>
 #include <rtc_base/logging.h>
 #include <third_party/libyuv/include/libyuv.h>
 
@@ -70,7 +71,7 @@ void FakeVideoCapturer::CaptureThread() {
     // 画像を更新
     UpdateImage(now);
 
-    // Blend2D イメージから I420 バッファへ変換
+    // Blend2D イメージから VideoFrameBuffer へ変換
     BLImageData data;
     BLResult result = image_.get_data(&data);
     if (result != BL_SUCCESS) {
@@ -79,14 +80,25 @@ void FakeVideoCapturer::CaptureThread() {
       break;
     }
 
-    webrtc::scoped_refptr<webrtc::I420Buffer> buffer =
-        webrtc::I420Buffer::Create(config_.width, config_.height);
-
-    libyuv::ABGRToI420((const uint8_t*)data.pixel_data, data.stride,
-                       buffer->MutableDataY(), buffer->StrideY(),
-                       buffer->MutableDataU(), buffer->StrideU(),
-                       buffer->MutableDataV(), buffer->StrideV(), config_.width,
-                       config_.height);
+    // 出力フォーマットを選択
+    webrtc::scoped_refptr<webrtc::VideoFrameBuffer> buffer;
+    if (config_.force_nv12) {
+      // NV12 へ変換
+      auto nv12 = webrtc::NV12Buffer::Create(config_.width, config_.height);
+      libyuv::ABGRToNV12((const uint8_t*)data.pixel_data, data.stride,
+                         nv12->MutableDataY(), nv12->StrideY(),
+                         nv12->MutableDataUV(), nv12->StrideUV(), config_.width,
+                         config_.height);
+      buffer = nv12;
+    } else {
+      // 既定は I420
+      auto i420 = webrtc::I420Buffer::Create(config_.width, config_.height);
+      libyuv::ABGRToI420(
+          (const uint8_t*)data.pixel_data, data.stride, i420->MutableDataY(),
+          i420->StrideY(), i420->MutableDataU(), i420->StrideU(),
+          i420->MutableDataV(), i420->StrideV(), config_.width, config_.height);
+      buffer = i420;
+    }
 
     // タイムスタンプを計算
     int64_t timestamp_us =
@@ -150,7 +162,7 @@ void FakeVideoCapturer::DrawAnimations(
   ctx.set_fill_style(BLRgba32(160, 160, 160));
   uint32_t current_frame = frame_counter_;
   ctx.fill_pie(0, 0, width * 0.3, 0,
-              (current_frame % fps) / static_cast<float>(fps) * 2 * M_PI);
+               (current_frame % fps) / static_cast<float>(fps) * 2 * M_PI);
 
   // 円が一周したときにビープ音を鳴らす
   auto fake_audio_capturer = GetAudioCapturer();
@@ -260,7 +272,7 @@ void FakeVideoCapturer::DrawDigitalClock(
 
   // ドット
   ctx.fill_circle(x + colon_width * 0.3, clock_y + digit_height * 0.8,
-                 digit_height * 0.05);
+                  digit_height * 0.05);
   x += colon_width + spacing;
 
   // ミリ秒（3桁、少し小さめに表示）
@@ -385,4 +397,4 @@ void FakeVideoCapturer::DrawColon(BLContext& ctx,
   ctx.fill_circle(x + dot_size, y + height * 0.7, dot_size);
 }
 
-#endif  // USE_FAKE_CAPTURE_DEVICE
+#endif  // USE_FAKE_CAPTURE_DEVICE

src/rtc/fake_video_capturer.h

@@ -27,6 +27,7 @@ class FakeVideoCapturer : public sora::ScalableVideoTrackSource {
     int width = 640;
     int height = 480;
     int fps = 30;
+    bool force_nv12 = false;
   };
 
   static webrtc::scoped_refptr<FakeVideoCapturer> Create(Config config) {
@@ -75,4 +76,4 @@ class FakeVideoCapturer : public sora::ScalableVideoTrackSource {
 
 #endif  // USE_FAKE_CAPTURE_DEVICE
 
-#endif  // RTC_FAKE_VIDEO_CAPTURER_H_
+#endif  // RTC_FAKE_VIDEO_CAPTURER_H_

test/momo.py

@@ -34,6 +34,8 @@ def __init__(
         no_audio_device: bool = False,
         fake_capture_device: bool = True,
         force_i420: bool = False,
+        force_yuy2: bool = False,
+        force_nv12: bool = False,
         hw_mjpeg_decoder: bool | None = None,
         use_libcamera: bool = False,
         use_libcamera_native: bool = False,
@@ -149,6 +151,8 @@ def __init__(
             "no_audio_device": no_audio_device,
             "fake_capture_device": fake_capture_device,
             "force_i420": force_i420,
+            "force_yuy2": force_yuy2,
+            "force_nv12": force_nv12,
             "hw_mjpeg_decoder": hw_mjpeg_decoder,
             "use_libcamera": use_libcamera,
             "use_libcamera_native": use_libcamera_native,
@@ -377,6 +381,10 @@ def _build_args(self, mode: MomoMode, **kwargs: Any) -> list[str]:
             args.append("--fake-capture-device")
         if kwargs.get("force_i420"):
             args.append("--force-i420")
+        if kwargs.get("force_yuy2"):
+            args.append("--force-yuy2")
+        if kwargs.get("force_nv12"):
+            args.append("--force-nv12")
         if kwargs.get("hw_mjpeg_decoder") is not None:
             args.extend(["--hw-mjpeg-decoder", str(int(kwargs["hw_mjpeg_decoder"]))])
         if kwargs.get("use_libcamera"):

test/test_sora_mode_nv12.py

@@ -0,0 +1,46 @@
+import os
+
+import pytest
+
+from momo import Momo, MomoMode
+
+# Sora モードのテストは TEST_SORA_MODE_SIGNALING_URLS が設定されていない場合スキップ
+pytestmark = pytest.mark.skipif(
+    not os.environ.get("TEST_SORA_MODE_SIGNALING_URLS"),
+    reason="TEST_SORA_MODE_SIGNALING_URLS not set in environment",
+)
+
+
+def test_sendonly_with_force_nv12_outbound_rtp(sora_settings, free_port):
+    """Sora モードで fake capture + --force-nv12 を指定し、送信統計が取得できることを確認"""
+
+    with Momo(
+        mode=MomoMode.SORA,
+        metrics_port=free_port,
+        fake_capture_device=True,
+        signaling_urls=sora_settings.signaling_urls,
+        channel_id=sora_settings.channel_id,
+        role="sendonly",
+        audio=True,
+        video=True,
+        force_nv12=True,
+        metadata=sora_settings.metadata,
+    ) as m:
+        # 接続が確立するまで待機
+        assert m.wait_for_connection(), "Failed to establish connection with --force-nv12"
+
+        # outbound-rtp(kind=video) が出現するまでメトリクスを待つ
+        data = m.get_metrics(
+            wait_stats=[{"type": "outbound-rtp", "kind": "video"}],
+            wait_stats_timeout=10,
+        )
+        stats = data.get("stats", [])
+
+        # 送信（映像）の統計が存在し、送信が進んでいることを確認
+        video_outbound = [
+            s for s in stats if s.get("type") == "outbound-rtp" and s.get("kind") == "video"
+        ]
+        assert len(video_outbound) >= 1, "No video outbound-rtp stats found"
+        v = video_outbound[0]
+        assert v.get("packetsSent", 0) > 0
+        assert v.get("bytesSent", 0) > 0