Capture audio from CEF

Julusian · niklaspandersson · commit 0e9f1a492447 · 2024-12-06T23:53:51.000+01:00
diff --git a/src/modules/ffmpeg/CMakeLists.txt b/src/modules/ffmpeg/CMakeLists.txt
@@ -5,6 +5,7 @@ set(SOURCES
 	producer/av_producer.cpp
 	producer/av_input.cpp
 	util/av_util.cpp
+	util/audio_resampler.cpp
 	producer/ffmpeg_producer.cpp
 	consumer/ffmpeg_consumer.cpp
 
@@ -15,6 +16,7 @@ set(HEADERS
 	producer/av_producer.h
 	producer/av_input.h
 	util/av_util.h
+	util/audio_resampler.h
 	producer/ffmpeg_producer.h
 	consumer/ffmpeg_consumer.h
 
diff --git a/src/modules/ffmpeg/util/audio_resampler.cpp b/src/modules/ffmpeg/util/audio_resampler.cpp
@@ -0,0 +1,41 @@
+#include "audio_resampler.h"
+#include "av_assert.h"
+
+extern "C" {
+#include <libavutil/samplefmt.h>
+#include <libswresample/swresample.h>
+}
+
+namespace caspar::ffmpeg {
+
+AudioResampler::AudioResampler(int sample_rate, AVSampleFormat in_sample_fmt)
+{
+    AVChannelLayout channel_layout = AV_CHANNEL_LAYOUT_7POINT1;
+    AVChannelLayout channel_layout_out = AV_CHANNEL_LAYOUT_HEXADECAGONAL;
+
+    SwrContext* raw_ctx = nullptr;
+    FF(swr_alloc_set_opts2(&raw_ctx,
+                           &channel_layout_out,
+                           AV_SAMPLE_FMT_S32,
+                           sample_rate,
+                           &channel_layout,
+                           in_sample_fmt,
+                           sample_rate,
+                           0,
+                           nullptr));
+
+    ctx = std::shared_ptr<SwrContext>(raw_ctx, [](SwrContext* ptr) { swr_free(&ptr); });
+
+    FF_RET(swr_init(ctx.get()), "swr_init");
+}
+
+caspar::array<int32_t> AudioResampler::convert(int frames, const void** src)
+{
+    auto result = caspar::array<int32_t>(frames * 16 * sizeof(int32_t));
+    auto ptr    = result.data();
+    swr_convert(ctx.get(), (uint8_t**)&ptr, frames, reinterpret_cast<const uint8_t**>(src), frames);
+
+    return result;
+}
+
+}; // namespace caspar::ffmpeg
diff --git a/src/modules/ffmpeg/util/audio_resampler.h b/src/modules/ffmpeg/util/audio_resampler.h
@@ -0,0 +1,27 @@
+#include <common/array.h>
+#include <memory>
+
+#pragma once
+
+extern "C" {
+#include <libavutil/samplefmt.h>
+}
+
+struct SwrContext;
+
+namespace caspar::ffmpeg {
+
+class AudioResampler
+{
+    std::shared_ptr<SwrContext> ctx;
+
+  public:
+    AudioResampler(int sample_rate, AVSampleFormat in_sample_fmt);
+
+    AudioResampler(const AudioResampler&)            = delete;
+    AudioResampler& operator=(const AudioResampler&) = delete;
+
+    caspar::array<int32_t> convert(int frames, const void** src);
+};
+
+}; // namespace caspar::ffmpeg
diff --git a/src/modules/html/CMakeLists.txt b/src/modules/html/CMakeLists.txt
@@ -25,7 +25,9 @@ target_include_directories(html PRIVATE
     ..
     ../..
     ${CEF_INCLUDE_PATH}
-    )
+	${FFMPEG_INCLUDE_PATH}
+)
+target_link_libraries(html ffmpeg)
 
 set_target_properties(html PROPERTIES FOLDER modules)
 source_group(sources\\producer producer/*)
diff --git a/src/modules/html/producer/html_producer.cpp b/src/modules/html/producer/html_producer.cpp
@@ -60,13 +60,53 @@
 #include <queue>
 #include <utility>
 
+#include <ffmpeg/util/audio_resampler.h>
+
 #include "../html.h"
 
 namespace caspar { namespace html {
 
+inline std::int_least64_t now()
+{
+    return std::chrono::duration_cast<std::chrono::milliseconds>(
+               std::chrono::high_resolution_clock::now().time_since_epoch())
+        .count();
+}
+
+struct presentation_frame
+{
+    std::int_least64_t timestamp;
+    core::draw_frame   frame;
+
+    explicit presentation_frame(core::draw_frame frame = {}, std::int_least64_t ts = now()) noexcept
+        : timestamp(ts)
+        , frame(std::move(frame))
+    {
+    }
+
+    presentation_frame(presentation_frame&& other) noexcept
+        : timestamp(other.timestamp)
+        , frame(std::move(other.frame))
+    {
+    }
+
+    presentation_frame(const presentation_frame&)            = delete;
+    presentation_frame& operator=(const presentation_frame&) = delete;
+
+    presentation_frame& operator=(presentation_frame&& rhs) noexcept
+    {
+        timestamp = rhs.timestamp;
+        frame     = std::move(rhs.frame);
+        return *this;
+    }
+
+    ~presentation_frame() {}
+};
+
 class html_client
     : public CefClient
     , public CefRenderHandler
+    , public CefAudioHandler
     , public CefLifeSpanHandler
     , public CefLoadHandler
     , public CefDisplayHandler
@@ -80,16 +120,21 @@ class html_client
     caspar::timer                       paint_timer_;
     caspar::timer                       test_timer_;
 
-    spl::shared_ptr<core::frame_factory>                        frame_factory_;
-    core::video_format_desc                                     format_desc_;
-    bool                                                        gpu_enabled_;
-    tbb::concurrent_queue<std::wstring>                         javascript_before_load_;
-    std::atomic<bool>                                           loaded_;
-    std::queue<std::pair<std::int_least64_t, core::draw_frame>> frames_;
-    mutable std::mutex                                          frames_mutex_;
-    const size_t                                                frames_max_size_ = 4;
-    std::atomic<bool>                                           closing_;
-
+    spl::shared_ptr<core::frame_factory> frame_factory_;
+    core::video_format_desc              format_desc_;
+    bool                                 gpu_enabled_;
+    tbb::concurrent_queue<std::wstring>  javascript_before_load_;
+    std::atomic<bool>                    loaded_;
+    std::queue<presentation_frame>       frames_;
+    std::queue<presentation_frame>       audio_frames_;
+    mutable std::mutex                   frames_mutex_;
+    mutable std::mutex                   audio_frames_mutex_;
+    const size_t                         frames_max_size_ = 4;
+    std::atomic<bool>                    closing_;
+
+    std::unique_ptr<ffmpeg::AudioResampler> audioResampler_;
+
+    core::draw_frame   last_video_frame_;
     core::draw_frame   last_frame_;
     std::int_least64_t last_frame_time_;
 
@@ -146,8 +191,21 @@ class html_client
 
     bool try_pop(const core::video_field field)
     {
+        bool result = false;
         std::lock_guard<std::mutex> lock(frames_mutex_);
 
+        core::draw_frame audio_frame;
+        uint64_t audio_frame_timestamp = 0;
+
+        {
+            std::lock_guard<std::mutex> audio_lock(audio_frames_mutex_);
+            if (!audio_frames_.empty()) {
+                audio_frame_timestamp = audio_frames_.front().timestamp;
+                audio_frame = core::draw_frame(std::move(audio_frames_.front().frame));
+                audio_frames_.pop();
+            }
+        }
+
         if (!frames_.empty()) {
             /*
              * CEF in gpu-enabled mode only sends frames when something changes, and interlaced channels
@@ -167,35 +225,43 @@ class html_client
 
                 // Check if the sole buffered frame is too young to have a partner field generated (with a tolerance)
                 auto time_per_frame           = (1000 * 1.5) / format_desc_.fps;
-                auto front_frame_is_too_young = (now_time - frames_.front().first) < time_per_frame;
+                auto front_frame_is_too_young = (now_time - frames_.front().timestamp) < time_per_frame;
 
                 if (follows_gap_in_frames && front_frame_is_too_young) {
                     return false;
                 }
             }
 
-            last_frame_time_ = frames_.front().first;
-            last_frame_      = std::move(frames_.front().second);
+            last_frame_time_ = frames_.front().timestamp;
+            last_video_frame_ = std::move(frames_.front().frame);
+            last_frame_      = last_video_frame_;
             frames_.pop();
 
             graph_->set_value("buffered-frames", (double)frames_.size() / frames_max_size_);
 
-            return true;
+            result = true;
         }
 
-        return false;
+        if (audio_frame) {
+            last_frame_time_ = audio_frame_timestamp;
+            last_frame_      = core::draw_frame::over(last_video_frame_, audio_frame);
+            result = true;
+        }
+
+        return result;
     }
 
     core::draw_frame receive(const core::video_field field)
     {
         if (!try_pop(field)) {
             graph_->set_tag(diagnostics::tag_severity::SILENT, "late-frame");
+            return core::draw_frame::still(last_frame_);
+        } else {
+            return last_frame_;
         }
-
-        return last_frame_;
     }
 
-    core::draw_frame last_frame() const { return last_frame_; }
+    core::draw_frame last_frame() const { return core::draw_frame::still(last_frame_); }
 
     bool is_ready() const
     {
@@ -245,13 +311,6 @@ class html_client
     }
 
   private:
-    std::int_least64_t now()
-    {
-        return std::chrono::duration_cast<std::chrono::milliseconds>(
-                   std::chrono::high_resolution_clock::now().time_since_epoch())
-            .count();
-    }
-
     void GetViewRect(CefRefPtr<CefBrowser> browser, CefRect& rect) override
     {
         CASPAR_ASSERT(CefCurrentlyOn(TID_UI));
@@ -302,8 +361,10 @@ class html_client
         {
             std::lock_guard<std::mutex> lock(frames_mutex_);
 
-            frames_.push(std::make_pair(now(), core::draw_frame(std::move(frame))));
-            while (frames_.size() > 4) {
+            core::draw_frame new_frame = core::draw_frame(std::move(frame));
+
+            frames_.push(presentation_frame(std::move(new_frame)));
+            while (frames_.size() > frames_max_size_) {
                 frames_.pop();
                 graph_->set_tag(diagnostics::tag_severity::WARNING, "dropped-frame");
             }
@@ -353,6 +414,8 @@ class html_client
 
     CefRefPtr<CefRenderHandler> GetRenderHandler() override { return this; }
 
+    CefRefPtr<CefAudioHandler> GetAudioHandler() override { return this; }
+
     CefRefPtr<CefLifeSpanHandler> GetLifeSpanHandler() override { return this; }
 
     CefRefPtr<CefLoadHandler> GetLoadHandler() override { return this; }
@@ -378,7 +441,7 @@ class html_client
 
             {
                 std::lock_guard<std::mutex> lock(frames_mutex_);
-                frames_.push(std::make_pair(now(), core::draw_frame::empty()));
+                frames_.push(presentation_frame());
             }
 
             {
@@ -399,6 +462,41 @@ class html_client
         return false;
     }
 
+    bool GetAudioParameters(CefRefPtr<CefBrowser> browser, CefAudioParameters& params) override
+    {
+        params.channel_layout    = CEF_CHANNEL_LAYOUT_7_1;
+        params.sample_rate       = format_desc_.audio_sample_rate;
+        params.frames_per_buffer = format_desc_.audio_cadence[0];
+        return format_desc_.audio_cadence.size() == 1; // TODO - handle 59.94
+    }
+
+    void OnAudioStreamStarted(CefRefPtr<CefBrowser> browser, const CefAudioParameters& params, int channels) override
+    {
+        audioResampler_ = std::make_unique<ffmpeg::AudioResampler>(params.sample_rate, AV_SAMPLE_FMT_FLTP);
+    }
+    void OnAudioStreamPacket(CefRefPtr<CefBrowser> browser, const float** data, int samples, int64_t pts) override
+    {
+        if (!audioResampler_)
+            return;
+
+        auto audio       = audioResampler_->convert(samples, reinterpret_cast<const void**>(data));
+        auto audio_frame = core::mutable_frame(this, {}, std::move(audio), core::pixel_format_desc());
+
+        {
+            std::lock_guard<std::mutex> lock(audio_frames_mutex_);
+            while (audio_frames_.size() >= frames_max_size_) {
+                audio_frames_.pop();
+            }
+            audio_frames_.push(presentation_frame(core::draw_frame(std::move(audio_frame))));
+        }
+    }
+    void OnAudioStreamStopped(CefRefPtr<CefBrowser> browser) override { audioResampler_ = nullptr; }
+    void OnAudioStreamError(CefRefPtr<CefBrowser> browser, const CefString& message) override
+    {
+        CASPAR_LOG(info) << "[html_producer] OnAudioStreamError: \"" << message.ToString() << "\"";
+        audioResampler_ = nullptr;
+    }
+
     void do_execute_javascript(const std::wstring& javascript)
     {
         html::begin_invoke([=] {
