Merge pull request #128 from openvinotoolkit/rhecker/ffmpeg-7

Bump ffmpeg to 7.0.2

Author: RHeckerIntel

openvino_bindings/README.md

@@ -112,7 +112,7 @@ The DLLs (with dependencies) will be in `bazel-bin/windows_bindings.tar`
 [Install OpenVINO Runtime 24.5.0](https://docs.openvino.ai/2024/get-started/install-openvino.html?PACKAGE=OPENVINO_GENAI&VERSION=v_2024_4_0&OP_SYSTEM=MACOS&DISTRIBUTION=ARCHIVE)  with GenAI flavor in `/opt/intel/openvino_24.5.0` and symlink to `/opt/intel/openvino`.
 
 Install OpenCV: `brew install opencv`
-Install ffmpeg: `brew install ffmpeg@6 && brew link ffmpeg@6`
+Install ffmpeg: `brew install ffmpeg@6 && brew link ffmpeg@7`
 
 Run: `bazel build :macos_bindings`
 

openvino_bindings/scripts/setup_ffmpeg.sh

@@ -1,5 +1,5 @@
 #!/bin/bash
-export FFMPEG_VERSION=6.1.2
+export FFMPEG_VERSION=7.0.2
 
 echo "Installing ffmpeg from source"
 rm -rf /tmp/build_ffmpeg
@@ -17,4 +17,4 @@ if [[ "$OSTYPE" == "linux-gnu"* ]]; then
       touch /etc/ld.so.conf.d/ffmpeg.conf
       bash -c  "echo /opt/ffmpeg/lib >> /etc/ld.so.conf.d/ffmpeg.conf"
       ldconfig -v
-fi
+fi

openvino_bindings/src/audio/audio_grabber.cc

@@ -8,8 +8,6 @@
 #include "audio_grabber.h"
 
 AudioGrabber::AudioGrabber(std::string filename): filename(filename) {
-    // Has been removed
-    // v_register_all(); //maybe move
     // Open video file
     if (avformat_open_input(&formatContext, filename.c_str(), nullptr, nullptr) != 0) {
         throw std::runtime_error("Failed to open video file!");
@@ -20,12 +18,11 @@ AudioGrabber::AudioGrabber(std::string filename): filename(filename) {
         throw std::runtime_error("Failed to find stream info!");
     }
 
-
-    AVCodec *codec = nullptr;
+    const AVCodec *codec = nullptr;
     for (unsigned int i = 0; i < formatContext->nb_streams; i++) {
         if (formatContext->streams[i]->codecpar->codec_type == AVMEDIA_TYPE_AUDIO) {
             audioStreamIndex = i;
-            codec = const_cast<AVCodec*>(avcodec_find_decoder(formatContext->streams[i]->codecpar->codec_id));
+            codec = avcodec_find_decoder(formatContext->streams[i]->codecpar->codec_id);
             break;
         }
     }
@@ -36,46 +33,57 @@ AudioGrabber::AudioGrabber(std::string filename): filename(filename) {
         throw std::runtime_error("Audio stream not found!");
     }
 
-
     codecContext = avcodec_alloc_context3(codec);
     avcodec_parameters_to_context(codecContext, formatContext->streams[audioStreamIndex]->codecpar);
     avcodec_open2(codecContext, codec, nullptr);
 
     swrContext = swr_alloc();
-    av_opt_set_int(swrContext, "in_channel_layout", codecContext->channel_layout, 0);
+
+    // FFmpeg 7: use AVChannelLayout instead of channel_layout
+    av_opt_set_chlayout(swrContext, "in_chlayout", &codecContext->ch_layout, 0);
     av_opt_set_int(swrContext, "in_sample_rate", codecContext->sample_rate, 0);
     av_opt_set_sample_fmt(swrContext, "in_sample_fmt", codecContext->sample_fmt, 0);
 
-    av_opt_set_int(swrContext, "out_channel_layout", AV_CH_LAYOUT_MONO, 0);  // Mono output
-    av_opt_set_int(swrContext, "out_sample_rate", 16000, 0);  // 16kHz output sample rate
-    av_opt_set_sample_fmt(swrContext, "out_sample_fmt", AV_SAMPLE_FMT_S16, 0);  // 16-bit PCM output
+    AVChannelLayout out_ch_layout;
+    av_channel_layout_default(&out_ch_layout, 1); // mono
+    av_opt_set_chlayout(swrContext, "out_chlayout", &out_ch_layout, 0);
+    av_opt_set_int(swrContext, "out_sample_rate", 16000, 0);  // 16kHz output
+    av_opt_set_sample_fmt(swrContext, "out_sample_fmt", AV_SAMPLE_FMT_S16, 0); // 16-bit PCM
 
     if (swr_init(swrContext) < 0) {
         throw std::runtime_error("Failed to initialize resampler!");
     }
 }
 
 AudioGrabber::~AudioGrabber() {
-    avcodec_close(codecContext);
-    avformat_close_input(&formatContext);
-    swr_free(&swrContext);
+    if (codecContext) {
+        avcodec_free_context(&codecContext);  // FFmpeg 7 replacement for avcodec_close
+    }
+    if (formatContext) {
+        avformat_close_input(&formatContext);
+    }
+    if (swrContext) {
+        swr_free(&swrContext);
+    }
 }
 
 std::vector<float> AudioGrabber::grab_chunk(uint64_t start_time, uint64_t duration) {
     AVFrame *frame = av_frame_alloc();
     AVPacket packet;
-    std::vector<float> resampledAudio;  // For storing resampled audio data as float
-    // Seek to the starting time (start_time in seconds)
-    int64_t startPts = av_rescale_q(start_time * AV_TIME_BASE, AV_TIME_BASE_Q, formatContext->streams[audioStreamIndex]->time_base);
+    std::vector<float> resampledAudio;
+
+    int64_t startPts = av_rescale_q(start_time * AV_TIME_BASE, AV_TIME_BASE_Q,
+                                    formatContext->streams[audioStreamIndex]->time_base);
     av_seek_frame(formatContext, audioStreamIndex, startPts, AVSEEK_FLAG_BACKWARD);
 
-    int audioDurationPts = av_rescale_q(duration * AV_TIME_BASE, AV_TIME_BASE_Q, formatContext->streams[audioStreamIndex]->time_base);
+    int audioDurationPts = av_rescale_q(duration * AV_TIME_BASE, AV_TIME_BASE_Q,
+                                        formatContext->streams[audioStreamIndex]->time_base);
+
     int decoded = 0;
     while (av_read_frame(formatContext, &packet) >= 0 && decoded < audioDurationPts) {
         if (packet.stream_index == audioStreamIndex) {
             if (avcodec_send_packet(codecContext, &packet) >= 0) {
                 while (avcodec_receive_frame(codecContext, frame) >= 0) {
-                    // Resample the audio to 16kHz
                     int outSamples = swr_get_out_samples(swrContext, frame->nb_samples);
                     std::vector<uint8_t> buffer(outSamples * av_get_bytes_per_sample(AV_SAMPLE_FMT_S16));
                     uint8_t *outBuffer[] = { buffer.data() };
@@ -85,9 +93,8 @@ std::vector<float> AudioGrabber::grab_chunk(uint64_t start_time, uint64_t durati
 
                     int16_t *resampledData = reinterpret_cast<int16_t*>(buffer.data());
 
-                    // Convert resampled int16_t samples to float and store in resampledAudio
                     for (int i = 0; i < resampledSamples; i++) {
-                        float sample = resampledData[i] / 32768.0f;  // Convert to float in range [-1.0, 1.0]
+                        float sample = resampledData[i] / 32768.0f;
                         resampledAudio.push_back(sample);
                     }
                 }
@@ -96,6 +103,7 @@ std::vector<float> AudioGrabber::grab_chunk(uint64_t start_time, uint64_t durati
         }
         av_packet_unref(&packet);
     }
+
     av_frame_free(&frame);
     return resampledAudio;
 }

openvino_bindings/third_party/ffmpeg/mac.BUILD

@@ -8,11 +8,11 @@ cc_library(
         "include/**/*.h",
     ]),
     srcs = [
-        "lib/libavdevice.60.dylib",
-        "lib/libavformat.60.dylib",
-        "lib/libavcodec.60.dylib",
-        "lib/libswresample.4.dylib",
-        "lib/libavutil.58.dylib",
+        "lib/libavdevice.61.dylib",
+        "lib/libavformat.61.dylib",
+        "lib/libavcodec.61.dylib",
+        "lib/libswresample.5.dylib",
+        "lib/libavutil.59.dylib",
     ],
     includes = [
         "lib",

openvino_bindings/third_party/ffmpeg/windows.BUILD

@@ -13,12 +13,12 @@ cc_library(
         "include/libswresample/*.h",
     ]),
     srcs = glob([
-        "bin/avcodec-60.dll",
-        "bin/avutil-58.dll",
-        "bin/avformat-60.dll",
-        "bin/avfilter-9.dll",
-        "bin/swscale-7.dll",
-        "bin/swresample-4.dll",
+        "bin/avcodec-61.dll",
+        "bin/avutil-59.dll",
+        "bin/avformat-61.dll",
+        "bin/avfilter-10.dll",
+        "bin/swscale-8.dll",
+        "bin/swresample-5.dll",
         "lib/*.lib",
     ]),
     includes = [

openvino_bindings/third_party/vcpkg.json

@@ -3,8 +3,5 @@
   "builtin-baseline": "c8582b4d83dbd36e1bebc08bf166b5eb807996b0",
   "dependencies": [
     "ffmpeg"
-  ],
-  "overrides": [
-    { "name": "ffmpeg", "version": "6.1.1" }
   ]
 }