RTC: Fix FFmpeg opus audio noisy issue

ossrs · Oct 21, 2023 · 00027b6 · 00027b6
1 parent 4e7c075
commit 00027b6
Show file tree

Hide file tree

Showing 2 changed files with 27 additions and 29 deletions.
diff --git a/trunk/src/app/srs_app_rtc_codec.cpp b/trunk/src/app/srs_app_rtc_codec.cpp
@@ -90,7 +90,6 @@ SrsAudioTranscoder::SrsAudioTranscoder()
     dec_frame_ = NULL;
     dec_packet_ = NULL;
     enc_ = NULL;
-    enc_frame_ = NULL;
     enc_packet_ = NULL;
     swr_ = NULL;
     swr_data_ = NULL;
@@ -123,10 +122,6 @@ SrsAudioTranscoder::~SrsAudioTranscoder()
         avcodec_free_context(&enc_);
     }
 
-    if (enc_frame_) {
-        av_frame_free(&enc_frame_);
-    }
-
     if (enc_packet_) {
         av_packet_free(&enc_packet_);
     }
@@ -242,7 +237,7 @@ srs_error_t SrsAudioTranscoder::init_enc(SrsAudioCodecId dst_codec, int dst_chan
     enc_->channel_layout = av_get_default_channel_layout(dst_channels);
     enc_->bit_rate = dst_bit_rate;
     enc_->sample_fmt = codec->sample_fmts[0];
-    enc_->time_base.num = 1; enc_->time_base.den = 1000; // {1, 1000}
+    enc_->time_base.num = 1; enc_->time_base.den = dst_samplerate; // {1, dst_samplerate}
     if (dst_codec == SrsAudioCodecIdOpus) {
         //TODO: for more level setting
         enc_->compression_level = 1;
@@ -256,19 +251,6 @@ srs_error_t SrsAudioTranscoder::init_enc(SrsAudioCodecId dst_codec, int dst_chan
         return srs_error_new(ERROR_RTC_RTP_MUXER, "Could not open codec");
     }
 
-    enc_frame_ = av_frame_alloc();
-    if (!enc_frame_) {
-        return srs_error_new(ERROR_RTC_RTP_MUXER, "Could not allocate audio encode in frame");
-    }
-
-    enc_frame_->format = enc_->sample_fmt;
-    enc_frame_->nb_samples = enc_->frame_size;
-    enc_frame_->channel_layout = enc_->channel_layout;
-
-    if (av_frame_get_buffer(enc_frame_, 0) < 0) {
-        return srs_error_new(ERROR_RTC_RTP_MUXER, "Could not get audio frame buffer");
-    }
-
     enc_packet_ = av_packet_alloc();
     if (!enc_packet_) {
         return srs_error_new(ERROR_RTC_RTP_MUXER, "Could not allocate audio encode out packet");
@@ -380,25 +362,40 @@ srs_error_t SrsAudioTranscoder::encode(std::vector<SrsAudioFrame*> &pkts)
     if (next_out_pts_ == AV_NOPTS_VALUE) {
         next_out_pts_ = new_pkt_pts_;
     } else {
-        int64_t diff = llabs(new_pkt_pts_ - next_out_pts_);
+        int64_t diff = llabs(new_pkt_pts_ - av_rescale(next_out_pts_, 1000, enc_->time_base.den));
         if (diff > 1000) {
             srs_trace("time diff to large=%lld, next out=%lld, new pkt=%lld, set to new pkt",
                 diff, next_out_pts_, new_pkt_pts_);
-            next_out_pts_ = new_pkt_pts_;
+            next_out_pts_ = av_rescale(new_pkt_pts_, enc_->time_base.den, 1000);
         }
     }
 
-    int frame_cnt = 0;
     while (av_audio_fifo_size(fifo_) >= enc_->frame_size) {
+        AVFrame* enc_frame = av_frame_alloc();
+        if (!enc_frame) {
+            return srs_error_new(ERROR_RTC_RTP_MUXER, "Could not allocate audio encode in frame");
+        }
+
+        enc_frame->format = enc_->sample_fmt;
+        enc_frame->nb_samples = enc_->frame_size;
+        enc_frame->channel_layout = enc_->channel_layout;
+
+        if (av_frame_get_buffer(enc_frame, 0) < 0) {
+            av_frame_free(&enc_frame);
+            return srs_error_new(ERROR_RTC_RTP_MUXER, "Could not get audio frame buffer");
+        }
+
         /* Read as many samples from the FIFO buffer as required to fill the frame.
         * The samples are stored in the frame temporarily. */
-        if (av_audio_fifo_read(fifo_, (void **)enc_frame_->data, enc_->frame_size) < enc_->frame_size) {
+        if (av_audio_fifo_read(fifo_, (void **)enc_frame->data, enc_->frame_size) < enc_->frame_size) {
+            av_frame_free(&enc_frame);
             return srs_error_new(ERROR_RTC_RTP_MUXER, "Could not read data from FIFO");
         }
         /* send the frame for encoding */
-        enc_frame_->pts = next_out_pts_ + av_rescale(enc_->frame_size * frame_cnt, 1000, enc_->sample_rate);
-        ++frame_cnt;
-        int error = avcodec_send_frame(enc_, enc_frame_);
+        enc_frame->pts = next_out_pts_;
+        next_out_pts_ += enc_->frame_size;
+        int error = avcodec_send_frame(enc_, enc_frame);
+        av_frame_free(&enc_frame);
         if (error < 0) {
             return srs_error_new(ERROR_RTC_RTP_MUXER, "Error sending the frame to the encoder(%d,%s)", error,
                 av_make_error_string(err_buf, AV_ERROR_MAX_STRING_SIZE, error));
@@ -419,6 +416,10 @@ srs_error_t SrsAudioTranscoder::encode(std::vector<SrsAudioFrame*> &pkts)
                     av_make_error_string(err_buf, AV_ERROR_MAX_STRING_SIZE, error));
             }
 
+            // rescale time base from sample_rate 1000.
+            enc_packet_->dts = av_rescale(enc_packet_->dts, 1000, enc_->time_base.den); 
+            enc_packet_->pts = av_rescale(enc_packet_->pts, 1000, enc_->time_base.den); 
+
             SrsAudioFrame *out_frame = new SrsAudioFrame;
             char *buf = new char[enc_packet_->size];
             memcpy(buf, enc_packet_->data, enc_packet_->size);
@@ -429,8 +430,6 @@ srs_error_t SrsAudioTranscoder::encode(std::vector<SrsAudioFrame*> &pkts)
         }
     }
 
-    next_out_pts_ += av_rescale(enc_->frame_size * frame_cnt, 1000, enc_->sample_rate);
-
     return srs_success;
 }
 

diff --git a/trunk/src/app/srs_app_rtc_codec.hpp b/trunk/src/app/srs_app_rtc_codec.hpp
@@ -38,7 +38,6 @@ class SrsAudioTranscoder
     AVPacket *dec_packet_;
 
     AVCodecContext *enc_;
-    AVFrame *enc_frame_;
     AVPacket *enc_packet_;
 
     SwrContext *swr_;