Injectable audio encoders: WebRtcVoiceEngine and company

webrtc/media/engine/webrtcvoiceengine.cc

Index: webrtc/media/engine/webrtcvoiceengine.cc
diff --git a/webrtc/media/engine/webrtcvoiceengine.cc b/webrtc/media/engine/webrtcvoiceengine.cc
index f05c702c5b5a898278b11dbdd6b98d3db2eb5821..807dfe2ab8da2167aa7e7ff90fd92a9995395acd 100644
--- a/webrtc/media/engine/webrtcvoiceengine.cc
+++ b/webrtc/media/engine/webrtcvoiceengine.cc
@@ -37,6 +37,7 @@
 #include "webrtc/media/engine/webrtcmediaengine.h"
 #include "webrtc/media/engine/webrtcvoe.h"
 #include "webrtc/modules/audio_coding/acm2/rent_a_codec.h"
+#include "webrtc/modules/audio_coding/codecs/builtin_audio_encoder_factory.h"
 #include "webrtc/modules/audio_mixer/audio_mixer_impl.h"
 #include "webrtc/modules/audio_processing/include/audio_processing.h"
 #include "webrtc/system_wrappers/include/field_trial.h"
@@ -77,27 +78,9 @@ constexpr int kNackRtpHistoryMs = 5000;
 #error "Set WEBRTC_INTELLIGIBILITY_ENHANCER to either 0 or 1"
 #endif
 
-// Codec parameters for Opus.
-// draft-spittka-payload-rtp-opus-03
-
-// Recommended bitrates:
-// 8-12 kb/s for NB speech,
-// 16-20 kb/s for WB speech,
-// 28-40 kb/s for FB speech,
-// 48-64 kb/s for FB mono music, and
-// 64-128 kb/s for FB stereo music.
-// The current implementation applies the following values to mono signals,
-// and multiplies them by 2 for stereo.
-const int kOpusBitrateNbBps = 12000;
-const int kOpusBitrateWbBps = 20000;
-const int kOpusBitrateFbBps = 32000;
-
-// Opus bitrate should be in the range between 6000 and 510000.
+// For SendSideBwe, Opus bitrate should be in the range between 6000 and 32000.
 const int kOpusMinBitrateBps = 6000;
-const int kOpusMaxBitrateBps = 510000;
-
-// iSAC bitrate should be <= 56000.
-const int kIsacMaxBitrateBps = 56000;
+const int kOpusBitrateFbBps = 32000;
 
 // Default audio dscp value.
 // See http://tools.ietf.org/html/rfc2474 for details.
@@ -138,8 +121,15 @@ bool ValidateStreamParams(const StreamParams& sp) {
 // Dumps an AudioCodec in RFC 2327-ish format.
 std::string ToString(const AudioCodec& codec) {
   std::stringstream ss;
-  ss << codec.name << "/" << codec.clockrate << "/" << codec.channels
-     << " (" << codec.id << ")";
+  ss << codec.name << "/" << codec.clockrate << "/" << codec.channels;
+  if (!codec.params.empty()) {
+    ss << " {";
+    for (const auto& param : codec.params) {
+      ss << " " << param.first << "=" << param.second;
+    }
+    ss << " }";
+  }
+  ss << " (" << codec.id << ")";
   return ss.str();
 }
 
@@ -185,12 +175,6 @@ bool VerifyUniquePayloadTypes(const std::vector<AudioCodec>& codecs) {
   return it == payload_types.end();
 }
 
-// Return true if codec.params[feature] == "1", false otherwise.
-bool IsCodecFeatureEnabled(const AudioCodec& codec, const char* feature) {
-  int value;
-  return codec.GetParam(feature, &value) && value == 1;
-}
-
 rtc::Optional<std::string> GetAudioNetworkAdaptorConfig(
     const AudioOptions& options) {
   if (options.audio_network_adaptor && *options.audio_network_adaptor &&
@@ -202,85 +186,6 @@ rtc::Optional<std::string> GetAudioNetworkAdaptorConfig(
   return rtc::Optional<std::string>();
 }
 
-// Returns integer parameter params[feature] if it is defined. Returns
-// |default_value| otherwise.
-int GetCodecFeatureInt(const AudioCodec& codec,
-                       const char* feature,
-                       int default_value) {
-  int value = 0;
-  if (codec.GetParam(feature, &value)) {
-    return value;
-  }
-  return default_value;
-}
-
-// Use params[kCodecParamMaxAverageBitrate] if it is defined, use codec.bitrate
-// otherwise. If the value (either from params or codec.bitrate) <=0, use the
-// default configuration. If the value is beyond feasible bit rate of Opus,
-// clamp it. Returns the Opus bit rate for operation.
-int GetOpusBitrate(const AudioCodec& codec, int max_playback_rate) {
-  int bitrate = 0;
-  bool use_param = true;
-  if (!codec.GetParam(kCodecParamMaxAverageBitrate, &bitrate)) {
-    bitrate = codec.bitrate;
-    use_param = false;
-  }
-  if (bitrate <= 0) {
-    if (max_playback_rate <= 8000) {
-      bitrate = kOpusBitrateNbBps;
-    } else if (max_playback_rate <= 16000) {
-      bitrate = kOpusBitrateWbBps;
-    } else {
-      bitrate = kOpusBitrateFbBps;
-    }
-
-    if (IsCodecFeatureEnabled(codec, kCodecParamStereo)) {
-      bitrate *= 2;
-    }
-  } else if (bitrate < kOpusMinBitrateBps || bitrate > kOpusMaxBitrateBps) {
-    bitrate = (bitrate < kOpusMinBitrateBps) ? kOpusMinBitrateBps
-                                             : kOpusMaxBitrateBps;
-    std::string rate_source =
-        use_param ? "Codec parameter \"maxaveragebitrate\"" :
-            "Supplied Opus bitrate";
-    LOG(LS_WARNING) << rate_source
-                    << " is invalid and is replaced by: "
-                    << bitrate;
-  }
-  return bitrate;
-}
-
-void GetOpusConfig(const AudioCodec& codec,
-                   webrtc::CodecInst* voe_codec,
-                   bool* enable_codec_fec,
-                   int* max_playback_rate,
-                   bool* enable_codec_dtx,
-                   int* min_ptime_ms,
-                   int* max_ptime_ms) {
-  *enable_codec_fec = IsCodecFeatureEnabled(codec, kCodecParamUseInbandFec);
-  *enable_codec_dtx = IsCodecFeatureEnabled(codec, kCodecParamUseDtx);
-  *max_playback_rate = GetCodecFeatureInt(codec, kCodecParamMaxPlaybackRate,
-                                          kOpusDefaultMaxPlaybackRate);
-  *max_ptime_ms =
-      GetCodecFeatureInt(codec, kCodecParamMaxPTime, kOpusDefaultMaxPTime);
-  *min_ptime_ms =
-      GetCodecFeatureInt(codec, kCodecParamMinPTime, kOpusDefaultMinPTime);
-  if (*max_ptime_ms < *min_ptime_ms) {
-    // If min ptime or max ptime defined by codec parameter is wrong, we use
-    // the default values.
-    *max_ptime_ms = kOpusDefaultMaxPTime;
-    *min_ptime_ms = kOpusDefaultMinPTime;
-  }
-
-  // If OPUS, change what we send according to the "stereo" codec
-  // parameter, and not the "channels" parameter.  We set
-  // voe_codec.channels to 2 if "stereo=1" and 1 otherwise.  If
-  // the bitrate is not specified, i.e. is <= zero, we set it to the
-  // appropriate default value for mono or stereo Opus.
-  voe_codec->channels = IsCodecFeatureEnabled(codec, kCodecParamStereo) ? 2 : 1;
-  voe_codec->rate = GetOpusBitrate(codec, *max_playback_rate);
-}
-
 webrtc::AudioState::Config MakeAudioStateConfig(
     VoEWrapper* voe_wrapper,
     rtc::scoped_refptr<webrtc::AudioMixer> audio_mixer) {
@@ -296,67 +201,14 @@ webrtc::AudioState::Config MakeAudioStateConfig(
 
 class WebRtcVoiceCodecs final {
  public:
-  // TODO(solenberg): Do this filtering once off-line, add a simple AudioCodec
-  // list and add a test which verifies VoE supports the listed codecs.
-  static std::vector<AudioCodec> SupportedSendCodecs() {
-    std::vector<AudioCodec> result;
-    // Iterate first over our preferred codecs list, so that the results are
-    // added in order of preference.
-    for (size_t i = 0; i < arraysize(kCodecPrefs); ++i) {
-      const CodecPref* pref = &kCodecPrefs[i];
-      for (webrtc::CodecInst voe_codec : webrtc::acm2::RentACodec::Database()) {
-        // Change the sample rate of G722 to 8000 to match SDP.
-        MaybeFixupG722(&voe_codec, 8000);
-        // Skip uncompressed formats.
-        if (IsCodec(voe_codec, kL16CodecName)) {
-          continue;
-        }
-
-        if (!IsCodec(voe_codec, pref->name) ||
-            pref->clockrate != voe_codec.plfreq ||
-            pref->channels != voe_codec.channels) {
-          // Not a match.
-          continue;
-        }
-
-        AudioCodec codec(pref->payload_type, voe_codec.plname, voe_codec.plfreq,
-                         voe_codec.rate, voe_codec.channels);
-        LOG(LS_INFO) << "Adding supported codec: " << ToString(codec);
-        if (IsCodec(codec, kIsacCodecName)) {
-          // Indicate auto-bitrate in signaling.
-          codec.bitrate = 0;
-        }
-        if (IsCodec(codec, kOpusCodecName)) {
-          // Only add fmtp parameters that differ from the spec.
-          if (kPreferredMinPTime != kOpusDefaultMinPTime) {
-            codec.params[kCodecParamMinPTime] =
-                rtc::ToString(kPreferredMinPTime);
-          }
-          if (kPreferredMaxPTime != kOpusDefaultMaxPTime) {
-            codec.params[kCodecParamMaxPTime] =
-                rtc::ToString(kPreferredMaxPTime);
-          }
-          codec.SetParam(kCodecParamUseInbandFec, 1);
-          codec.AddFeedbackParam(
-              FeedbackParam(kRtcpFbParamTransportCc, kParamValueEmpty));
-
-          // TODO(hellner): Add ptime, sprop-stereo, and stereo
-          // when they can be set to values other than the default.
-        }
-        result.push_back(codec);
-      }
-    }
-    return result;
-  }
-
-  static bool ToCodecInst(const AudioCodec& in,
-                          webrtc::CodecInst* out) {
+  static bool ToCodecInst(const AudioCodec& in, webrtc::CodecInst* out) {
     for (webrtc::CodecInst voe_codec : webrtc::acm2::RentACodec::Database()) {
       // Change the sample rate of G722 to 8000 to match SDP.
       MaybeFixupG722(&voe_codec, 8000);
       AudioCodec codec(voe_codec.pltype, voe_codec.plname, voe_codec.plfreq,
                        voe_codec.rate, voe_codec.channels);
-      bool multi_rate = IsCodecMultiRate(voe_codec);
+      const bool multi_rate =
+          IsCodec(codec, kIsacCodecName) || IsCodec(codec, kOpusCodecName);
       // Allow arbitrary rates for ISAC to be specified.
       if (multi_rate) {
         // Set codec.bitrate to 0 so the check for codec.Matches() passes.
@@ -383,112 +235,6 @@ class WebRtcVoiceCodecs final {
     return false;
   }
 
-  static bool IsCodecMultiRate(const webrtc::CodecInst& codec) {
-    for (size_t i = 0; i < arraysize(kCodecPrefs); ++i) {
-      if (IsCodec(codec, kCodecPrefs[i].name) &&
-          kCodecPrefs[i].clockrate == codec.plfreq) {
-        return kCodecPrefs[i].is_multi_rate;
-      }
-    }
-    return false;
-  }
-
-  static int MaxBitrateBps(const webrtc::CodecInst& codec) {
-    for (size_t i = 0; i < arraysize(kCodecPrefs); ++i) {
-      if (IsCodec(codec, kCodecPrefs[i].name) &&
-          kCodecPrefs[i].clockrate == codec.plfreq) {
-        return kCodecPrefs[i].max_bitrate_bps;
-      }
-    }
-    return 0;
-  }
-
-  static rtc::ArrayView<const int> GetPacketSizesMs(
-      const webrtc::CodecInst& codec) {
-    for (size_t i = 0; i < arraysize(kCodecPrefs); ++i) {
-      if (IsCodec(codec, kCodecPrefs[i].name)) {
-        size_t num_packet_sizes = kMaxNumPacketSize;
-        for (int index = 0; index < kMaxNumPacketSize; index++) {
-          if (kCodecPrefs[i].packet_sizes_ms[index] == 0) {
-            num_packet_sizes = index;
-            break;
-          }
-        }
-        return rtc::ArrayView<const int>(kCodecPrefs[i].packet_sizes_ms,
-                                         num_packet_sizes);
-      }
-    }
-    return rtc::ArrayView<const int>();
-  }
-
-  // If the AudioCodec param kCodecParamPTime is set, then we will set it to
-  // codec pacsize if it's valid, or we will pick the next smallest value we
-  // support.
-  // TODO(Brave): Query supported packet sizes from ACM when the API is ready.
-  static bool SetPTimeAsPacketSize(webrtc::CodecInst* codec, int ptime_ms) {
-    for (const CodecPref& codec_pref : kCodecPrefs) {
-      if ((IsCodec(*codec, codec_pref.name) &&
-          codec_pref.clockrate == codec->plfreq) ||
-          IsCodec(*codec, kG722CodecName)) {
-        int packet_size_ms = SelectPacketSize(codec_pref, ptime_ms);
-        if (packet_size_ms) {
-          // Convert unit from milli-seconds to samples.
-          codec->pacsize = (codec->plfreq / 1000) * packet_size_ms;
-          return true;
-        }
-      }
-    }
-    return false;
-  }
-
-  static const AudioCodec* GetPreferredCodec(
-      const std::vector<AudioCodec>& codecs,
-      webrtc::CodecInst* out) {
-    RTC_DCHECK(out);
-    // Select the preferred send codec (the first non-telephone-event/CN codec).
-    for (const AudioCodec& codec : codecs) {
-      if (IsCodec(codec, kDtmfCodecName) || IsCodec(codec, kCnCodecName)) {
-        // Skip telephone-event/CN codecs - they will be handled later.
-        continue;
-      }
-
-      // We'll use the first codec in the list to actually send audio data.
-      // Be sure to use the payload type requested by the remote side.
-      // Ignore codecs we don't know about. The negotiation step should prevent
-      // this, but double-check to be sure.
-      if (!ToCodecInst(codec, out)) {
-        LOG(LS_WARNING) << "Unknown codec " << ToString(codec);
-        continue;
-      }
-      return &codec;
-    }
-    return nullptr;
-  }
-
- private:
-  static const int kMaxNumPacketSize = 6;
-  struct CodecPref {
-    const char* name;
-    int clockrate;
-    size_t channels;
-    int payload_type;
-    bool is_multi_rate;
-    int packet_sizes_ms[kMaxNumPacketSize];
-    int max_bitrate_bps;
-  };
-  // Note: keep the supported packet sizes in ascending order.
-  static const CodecPref kCodecPrefs[14];
-
-  static int SelectPacketSize(const CodecPref& codec_pref, int ptime_ms) {
-    int selected_packet_size_ms = codec_pref.packet_sizes_ms[0];
-    for (int packet_size_ms : codec_pref.packet_sizes_ms) {
-      if (packet_size_ms && packet_size_ms <= ptime_ms) {
-        selected_packet_size_ms = packet_size_ms;
-      }
-    }
-    return selected_packet_size_ms;
-  }
-
   // Changes RTP timestamp rate of G722. This is due to the "bug" in the RFC
   // which says that G722 should be advertised as 8 kHz although it is a 16 kHz
   // codec.
@@ -502,66 +248,37 @@ class WebRtcVoiceCodecs final {
   }
 };
 
-const WebRtcVoiceCodecs::CodecPref WebRtcVoiceCodecs::kCodecPrefs[14] = {
-#if WEBRTC_OPUS_SUPPORT_120MS_PTIME
-    {kOpusCodecName, 48000, 2, 111, true, {10, 20, 40, 60, 120},
-     kOpusMaxBitrateBps},
-#else
-    {kOpusCodecName, 48000, 2, 111, true, {10, 20, 40, 60}, kOpusMaxBitrateBps},
-#endif
-    {kIsacCodecName, 16000, 1, 103, true, {30, 60}, kIsacMaxBitrateBps},
-    {kIsacCodecName, 32000, 1, 104, true, {30}, kIsacMaxBitrateBps},
-    // G722 should be advertised as 8000 Hz because of the RFC "bug".
-    {kG722CodecName, 8000, 1, 9, false, {10, 20, 30, 40, 50, 60}},
-    {kIlbcCodecName, 8000, 1, 102, false, {20, 30, 40, 60}},
-    {kPcmuCodecName, 8000, 1, 0, false, {10, 20, 30, 40, 50, 60}},
-    {kPcmaCodecName, 8000, 1, 8, false, {10, 20, 30, 40, 50, 60}},
-    {kCnCodecName, 32000, 1, 106, false, {}},
-    {kCnCodecName, 16000, 1, 105, false, {}},
-    {kCnCodecName, 8000, 1, 13, false, {}},
-    {kDtmfCodecName, 48000, 1, 110, false, {}},
-    {kDtmfCodecName, 32000, 1, 112, false, {}},
-    {kDtmfCodecName, 16000, 1, 113, false, {}},
-    {kDtmfCodecName, 8000, 1, 126, false, {}}
-};
-
 // |max_send_bitrate_bps| is the bitrate from "b=" in SDP.
 // |rtp_max_bitrate_bps| is the bitrate from RtpSender::SetParameters.
 rtc::Optional<int> ComputeSendBitrate(int max_send_bitrate_bps,
                                       rtc::Optional<int> rtp_max_bitrate_bps,
-                                      const webrtc::CodecInst& codec_inst) {
+                                      const webrtc::AudioCodecSpec& spec) {
   // If application-configured bitrate is set, take minimum of that and SDP
   // bitrate.
   const int bps = rtp_max_bitrate_bps
                       ? MinPositive(max_send_bitrate_bps, *rtp_max_bitrate_bps)
                       : max_send_bitrate_bps;
-  const int codec_rate = codec_inst.rate;
-
   if (bps <= 0) {
-    return rtc::Optional<int>(codec_rate);
+    return rtc::Optional<int>(spec.info.default_bitrate_bps);
   }
 
-  if (codec_inst.pltype == -1) {
-    return rtc::Optional<int>(codec_rate);
-    ;
-  }
-
-  if (WebRtcVoiceCodecs::IsCodecMultiRate(codec_inst)) {
-    // If codec is multi-rate then just set the bitrate.
-    return rtc::Optional<int>(
-        std::min(bps, WebRtcVoiceCodecs::MaxBitrateBps(codec_inst)));
-  }
-
-  if (bps < codec_inst.rate) {
+  if (bps < spec.info.min_bitrate_bps) {
     // If codec is not multi-rate and |bps| is less than the fixed bitrate then
     // fail. If codec is not multi-rate and |bps| exceeds or equal the fixed
     // bitrate then ignore.
-    LOG(LS_ERROR) << "Failed to set codec " << codec_inst.plname
+    LOG(LS_ERROR) << "Failed to set codec " << spec.format.name
                   << " to bitrate " << bps << " bps"
-                  << ", requires at least " << codec_inst.rate << " bps.";
+                  << ", requires at least " << spec.info.min_bitrate_bps
+                  << " bps.";
     return rtc::Optional<int>();
   }
-  return rtc::Optional<int>(codec_rate);
+
+  if (spec.info.HasFixedBitrate()) {
+    return rtc::Optional<int>(spec.info.default_bitrate_bps);
+  } else {
+    // If codec is multi-rate then just set the bitrate.
+    return rtc::Optional<int>(std::min(bps, spec.info.max_bitrate_bps));
+  }
 }
 
 }  // namespace
@@ -585,7 +302,10 @@ WebRtcVoiceEngine::WebRtcVoiceEngine(
     const rtc::scoped_refptr<webrtc::AudioDecoderFactory>& decoder_factory,
     rtc::scoped_refptr<webrtc::AudioMixer> audio_mixer,
     VoEWrapper* voe_wrapper)
-    : adm_(adm), decoder_factory_(decoder_factory), voe_wrapper_(voe_wrapper) {
+    : adm_(adm),
+      encoder_factory_(webrtc::CreateBuiltinAudioEncoderFactory()),
+      decoder_factory_(decoder_factory),
+      voe_wrapper_(voe_wrapper) {
   RTC_DCHECK(worker_thread_checker_.CalledOnValidThread());
   LOG(LS_INFO) << "WebRtcVoiceEngine::WebRtcVoiceEngine";
   RTC_DCHECK(voe_wrapper);
@@ -595,13 +315,13 @@ WebRtcVoiceEngine::WebRtcVoiceEngine(
 
   // Load our audio codec list.
   LOG(LS_INFO) << "Supported send codecs in order of preference:";
-  send_codecs_ = WebRtcVoiceCodecs::SupportedSendCodecs();
+  send_codecs_ = CollectCodecs(encoder_factory_->GetSupportedEncoders());
   for (const AudioCodec& codec : send_codecs_) {
     LOG(LS_INFO) << ToString(codec);
   }
 
   LOG(LS_INFO) << "Supported recv codecs in order of preference:";
-  recv_codecs_ = CollectRecvCodecs();
+  recv_codecs_ = CollectCodecs(decoder_factory_->GetSupportedDecoders());
   for (const AudioCodec& codec : recv_codecs_) {
     LOG(LS_INFO) << ToString(codec);
   }
@@ -1104,11 +824,10 @@ webrtc::voe::TransmitMixer* WebRtcVoiceEngine::transmit_mixer() {
   return transmit_mixer_;
 }
 
-AudioCodecs WebRtcVoiceEngine::CollectRecvCodecs() const {
+AudioCodecs WebRtcVoiceEngine::CollectCodecs(
+    const std::vector<webrtc::AudioCodecSpec>& specs) const {
   PayloadTypeMapper mapper;
   AudioCodecs out;
-  const std::vector<webrtc::AudioCodecSpec>& specs =
-      decoder_factory_->GetSupportedDecoders();
 
   // Only generate CN payload types for these clockrates:
   std::map<int, bool, std::greater<int>> generate_cn = {{ 8000,  false },
@@ -1188,12 +907,14 @@ class WebRtcVoiceMediaChannel::WebRtcAudioSendStream
       webrtc::AudioTransport* voe_audio_transport,
       uint32_t ssrc,
       const std::string& c_name,
-      const webrtc::AudioSendStream::Config::SendCodecSpec& send_codec_spec,
+      const rtc::Optional<webrtc::AudioSendStream::Config::SendCodecSpec>&
+          send_codec_spec,
       const std::vector<webrtc::RtpExtension>& extensions,
       int max_send_bitrate_bps,
       const rtc::Optional<std::string>& audio_network_adaptor_config,
       webrtc::Call* call,
-      webrtc::Transport* send_transport)
+      webrtc::Transport* send_transport,
+      const rtc::scoped_refptr<webrtc::AudioEncoderFactory>& encoder_factory)
       : voe_audio_transport_(voe_audio_transport),
         call_(call),
         config_(send_transport),
@@ -1205,13 +926,21 @@ class WebRtcVoiceMediaChannel::WebRtcAudioSendStream
     // TODO(solenberg): Once we're not using FakeWebRtcVoiceEngine anymore:
     // RTC_DCHECK(voe_audio_transport);
     RTC_DCHECK(call);
+    RTC_DCHECK(encoder_factory);
     config_.rtp.ssrc = ssrc;
     config_.rtp.c_name = c_name;
     config_.voe_channel_id = ch;
     config_.rtp.extensions = extensions;
     config_.audio_network_adaptor_config = audio_network_adaptor_config;
+    config_.encoder_factory = encoder_factory;
     rtp_parameters_.encodings[0].ssrc = rtc::Optional<uint32_t>(ssrc);
-    RecreateAudioSendStream(send_codec_spec);
+
+    UpdateAllowedBitrateRange();
+    if (send_codec_spec) {
+      UpdateSendCodecSpec(*send_codec_spec);
+    }
+
+    CreateAudioSendStream();
   }
 
   ~WebRtcAudioSendStream() override {
@@ -1220,56 +949,45 @@ class WebRtcVoiceMediaChannel::WebRtcAudioSendStream
     call_->DestroyAudioSendStream(stream_);
   }
 
-  void RecreateAudioSendStream(
+  void SetSendCodecSpec(
       const webrtc::AudioSendStream::Config::SendCodecSpec& send_codec_spec) {
-    RTC_DCHECK(worker_thread_checker_.CalledOnValidThread());
-    send_codec_spec_ = send_codec_spec;
-    config_.rtp.nack.rtp_history_ms =
-        send_codec_spec_.nack_enabled ? kNackRtpHistoryMs : 0;
-    config_.send_codec_spec = send_codec_spec_;
-    auto send_rate = ComputeSendBitrate(
-        max_send_bitrate_bps_, rtp_parameters_.encodings[0].max_bitrate_bps,
-        send_codec_spec.codec_inst);
-    if (send_rate) {
-      // Apply a send rate that abides by |max_send_bitrate_bps_| and
-      // |rtp_parameters_| when possible. Otherwise use the codec rate.
-      config_.send_codec_spec.codec_inst.rate = *send_rate;
-    }
-    RecreateAudioSendStream();
+    UpdateSendCodecSpec(send_codec_spec);
+    ReconfigureAudioSendStream();
   }
 
-  void RecreateAudioSendStream(
-      const std::vector<webrtc::RtpExtension>& extensions) {
+  void SetRtpExtensions(const std::vector<webrtc::RtpExtension>& extensions) {
     RTC_DCHECK(worker_thread_checker_.CalledOnValidThread());
     config_.rtp.extensions = extensions;
-    RecreateAudioSendStream();
+    ReconfigureAudioSendStream();
   }
 
-  void RecreateAudioSendStream(
+  void SetAudioNetworkAdaptorConfig(
       const rtc::Optional<std::string>& audio_network_adaptor_config) {
     RTC_DCHECK(worker_thread_checker_.CalledOnValidThread());
     if (config_.audio_network_adaptor_config == audio_network_adaptor_config) {
       return;
     }
     config_.audio_network_adaptor_config = audio_network_adaptor_config;
-    RecreateAudioSendStream();
+    UpdateAllowedBitrateRange();
+    ReconfigureAudioSendStream();
   }
 
   bool SetMaxSendBitrate(int bps) {
     RTC_DCHECK(worker_thread_checker_.CalledOnValidThread());
-    auto send_rate =
-        ComputeSendBitrate(bps, rtp_parameters_.encodings[0].max_bitrate_bps,
-                           send_codec_spec_.codec_inst);
+    RTC_DCHECK(config_.send_codec_spec);
+    RTC_DCHECK(audio_codec_spec_);
+    auto send_rate = ComputeSendBitrate(
+        bps, rtp_parameters_.encodings[0].max_bitrate_bps, *audio_codec_spec_);
+
     if (!send_rate) {
       return false;
     }
 
     max_send_bitrate_bps_ = bps;
 
-    if (config_.send_codec_spec.codec_inst.rate != *send_rate) {
-      // Recreate AudioSendStream with new bit rate.
-      config_.send_codec_spec.codec_inst.rate = *send_rate;
-      RecreateAudioSendStream();
+    if (send_rate != config_.send_codec_spec->target_bitrate_bps) {
+      config_.send_codec_spec->target_bitrate_bps = send_rate;
+      ReconfigureAudioSendStream();
     }
     return true;
   }
@@ -1385,20 +1103,24 @@ class WebRtcVoiceMediaChannel::WebRtcAudioSendStream
     if (!ValidateRtpParameters(parameters)) {
       return false;
     }
-    auto send_rate = ComputeSendBitrate(max_send_bitrate_bps_,
-                                        parameters.encodings[0].max_bitrate_bps,
-                                        send_codec_spec_.codec_inst);
-    if (!send_rate) {
-      return false;
+
+    rtc::Optional<int> send_rate;
+    if (audio_codec_spec_) {
+      send_rate = ComputeSendBitrate(max_send_bitrate_bps_,
+                                     parameters.encodings[0].max_bitrate_bps,
+                                     *audio_codec_spec_);
+      if (!send_rate) {
+        return false;
+      }
     }
 
     rtp_parameters_ = parameters;
 
     // parameters.encodings[0].encodings[0].max_bitrate_bps could have changed.
-    if (config_.send_codec_spec.codec_inst.rate != *send_rate) {
+    if (send_rate && send_rate != config_.send_codec_spec->target_bitrate_bps) {
       // Recreate AudioSendStream with new bit rate.
-      config_.send_codec_spec.codec_inst.rate = *send_rate;
-      RecreateAudioSendStream();
+      config_.send_codec_spec->target_bitrate_bps = send_rate;
+      ReconfigureAudioSendStream();
     } else {
       // parameters.encodings[0].active could have changed.
       UpdateSendState();
@@ -1418,55 +1140,77 @@ class WebRtcVoiceMediaChannel::WebRtcAudioSendStream
     }
   }
 
-  void RecreateAudioSendStream() {
+  void UpdateAllowedBitrateRange() {
     RTC_DCHECK(worker_thread_checker_.CalledOnValidThread());
-    if (stream_) {
-      call_->DestroyAudioSendStream(stream_);
-      stream_ = nullptr;
-    }
-    RTC_DCHECK(!stream_);
     if (webrtc::field_trial::IsEnabled("WebRTC-Audio-SendSideBwe")) {
       config_.min_bitrate_bps = kOpusMinBitrateBps;
       config_.max_bitrate_bps = kOpusBitrateFbBps;
       // TODO(mflodman): Keep testing this and set proper values.
       // Note: This is an early experiment currently only supported by Opus.
       if (send_side_bwe_with_overhead_) {
-        auto packet_sizes_ms = WebRtcVoiceCodecs::GetPacketSizesMs(
-            config_.send_codec_spec.codec_inst);
-        if (!packet_sizes_ms.empty()) {
-          int max_packet_size_ms =
-              *std::max_element(packet_sizes_ms.begin(), packet_sizes_ms.end());
-          int min_packet_size_ms =
-              *std::min_element(packet_sizes_ms.begin(), packet_sizes_ms.end());
-
-          // Audio network adaptor will just use 20ms and 60ms frame lengths.
-          // The adaptor will only be active for the Opus encoder.
-          if (config_.audio_network_adaptor_config &&
-              IsCodec(config_.send_codec_spec.codec_inst, kOpusCodecName)) {
-#if WEBRTC_OPUS_SUPPORT_120MS_PTIME
-            max_packet_size_ms = 120;
-#else
-            max_packet_size_ms = 60;
-#endif
-            min_packet_size_ms = 20;
-          }
-
-          // OverheadPerPacket = Ipv4(20B) + UDP(8B) + SRTP(10B) + RTP(12)
-          constexpr int kOverheadPerPacket = 20 + 8 + 10 + 12;
-
-          int min_overhead_bps =
-              kOverheadPerPacket * 8 * 1000 / max_packet_size_ms;
-
-          int max_overhead_bps =
-              kOverheadPerPacket * 8 * 1000 / min_packet_size_ms;
-
-          config_.min_bitrate_bps = kOpusMinBitrateBps + min_overhead_bps;
-          config_.max_bitrate_bps = kOpusBitrateFbBps + max_overhead_bps;
-        }
+        const bool is_opus_with_ana =
+            config_.audio_network_adaptor_config &&
+            !STR_CASE_CMP(config_.send_codec_spec->format.name.c_str(),
+                          kOpusCodecName);
+        const int max_packet_size_ms =
+            (is_opus_with_ana && WEBRTC_OPUS_SUPPORT_120MS_PTIME) ? 120 : 60;
+        // Audio network adaptor will just use 20ms and 60ms frame lengths.
+        // The adaptor will only be active for the Opus encoder.
+        const int min_packet_size_ms = is_opus_with_ana ? 20 : 10;
+
+        // OverheadPerPacket = Ipv4(20B) + UDP(8B) + SRTP(10B) + RTP(12)
+        constexpr int kOverheadPerPacket = 20 + 8 + 10 + 12;
+
+        int min_overhead_bps =
+            kOverheadPerPacket * 8 * 1000 / max_packet_size_ms;
+
+        int max_overhead_bps =
+            kOverheadPerPacket * 8 * 1000 / min_packet_size_ms;
+
+        config_.min_bitrate_bps = kOpusMinBitrateBps + min_overhead_bps;
+        config_.max_bitrate_bps = kOpusBitrateFbBps + max_overhead_bps;
       }
     }
+  }
+
+  void UpdateSendCodecSpec(
+      const webrtc::AudioSendStream::Config::SendCodecSpec& send_codec_spec) {
+    RTC_DCHECK(worker_thread_checker_.CalledOnValidThread());
+    config_.rtp.nack.rtp_history_ms =
+        send_codec_spec.nack_enabled ? kNackRtpHistoryMs : 0;
+    config_.send_codec_spec =
+        rtc::Optional<webrtc::AudioSendStream::Config::SendCodecSpec>(
+            send_codec_spec);
+    auto info =
+        config_.encoder_factory->QueryAudioEncoder(send_codec_spec.format);
+    RTC_DCHECK(info);
+    // If a specific target bitrate has been set for the stream, use that as
+    // the new default bitrate when computing send bitrate.
+    if (send_codec_spec.target_bitrate_bps) {
+      info->default_bitrate_bps = std::max(
+          info->min_bitrate_bps,
+          std::min(info->max_bitrate_bps, *send_codec_spec.target_bitrate_bps));
+    }
+
+    audio_codec_spec_.emplace(
+        webrtc::AudioCodecSpec{send_codec_spec.format, *info});
+
+    config_.send_codec_spec->target_bitrate_bps = ComputeSendBitrate(
+        max_send_bitrate_bps_, rtp_parameters_.encodings[0].max_bitrate_bps,
+        *audio_codec_spec_);
+  }
+
+  void CreateAudioSendStream() {
+    RTC_DCHECK(worker_thread_checker_.CalledOnValidThread());
+    RTC_DCHECK(!stream_);
     stream_ = call_->CreateAudioSendStream(config_);
     RTC_CHECK(stream_);
+  }
+
+  void ReconfigureAudioSendStream() {
+    RTC_DCHECK(worker_thread_checker_.CalledOnValidThread());
+    RTC_DCHECK(stream_);
+    stream_->Reconfigure(config_);
     UpdateSendState();
   }
 
@@ -1488,7 +1232,7 @@ class WebRtcVoiceMediaChannel::WebRtcAudioSendStream
   bool muted_ = false;
   int max_send_bitrate_bps_;
   webrtc::RtpParameters rtp_parameters_;
-  webrtc::AudioSendStream::Config::SendCodecSpec send_codec_spec_;
+  rtc::Optional<webrtc::AudioCodecSpec> audio_codec_spec_;
 
   RTC_DISALLOW_IMPLICIT_CONSTRUCTORS(WebRtcAudioSendStream);
 };
@@ -1694,7 +1438,7 @@ bool WebRtcVoiceMediaChannel::SetSendParameters(
   if (send_rtp_extensions_ != filtered_extensions) {
     send_rtp_extensions_.swap(filtered_extensions);
     for (auto& it : send_streams_) {
-      it.second->RecreateAudioSendStream(send_rtp_extensions_);
+      it.second->SetRtpExtensions(send_rtp_extensions_);
     }
   }
 
@@ -1839,10 +1583,10 @@ bool WebRtcVoiceMediaChannel::SetOptions(const AudioOptions& options) {
     return false;
   }
 
-  rtc::Optional<std::string> audio_network_adatptor_config =
+  rtc::Optional<std::string> audio_network_adaptor_config =
       GetAudioNetworkAdaptorConfig(options_);
   for (auto& it : send_streams_) {
-    it.second->RecreateAudioSendStream(audio_network_adatptor_config);
+    it.second->SetAudioNetworkAdaptorConfig(audio_network_adaptor_config);
   }
 
   LOG(LS_INFO) << "Set voice channel options. Current options: "
@@ -1951,85 +1695,64 @@ bool WebRtcVoiceMediaChannel::SetSendCodecs(
     }
   }
 
-  // Scan through the list to figure out the codec to use for sending, along
-  // with the proper configuration for VAD, CNG, NACK and Opus-specific
-  // parameters.
-  // TODO(solenberg): Refactor this logic once we create AudioEncoders here.
-  webrtc::AudioSendStream::Config::SendCodecSpec send_codec_spec;
-  {
-    send_codec_spec.nack_enabled = send_codec_spec_.nack_enabled;
-
-    // Find send codec (the first non-telephone-event/CN codec).
-    const AudioCodec* codec = WebRtcVoiceCodecs::GetPreferredCodec(
-        codecs, &send_codec_spec.codec_inst);
-    if (!codec) {
-      LOG(LS_WARNING) << "Received empty list of codecs.";
-      return false;
-    }
+  // Scan through the list to figure out the codec to use for sending.
+  rtc::Optional<webrtc::AudioSendStream::Config::SendCodecSpec> send_codec_spec;
+  rtc::Optional<webrtc::AudioCodecInfo> voice_codec_info;
+  for (const AudioCodec& voice_codec : codecs) {
+    if (!(IsCodec(voice_codec, kCnCodecName) ||
+          IsCodec(voice_codec, kDtmfCodecName) ||
+          IsCodec(voice_codec, kRedCodecName))) {
+      webrtc::SdpAudioFormat format(voice_codec.name, voice_codec.clockrate,
+                                    voice_codec.channels, voice_codec.params);
 
-    send_codec_spec.transport_cc_enabled = HasTransportCc(*codec);
-    send_codec_spec.nack_enabled = HasNack(*codec);
-    bitrate_config_ = GetBitrateConfigForCodec(*codec);
-
-    // For Opus as the send codec, we are to determine inband FEC, maximum
-    // playback rate, and opus internal dtx.
-    if (IsCodec(*codec, kOpusCodecName)) {
-      GetOpusConfig(*codec, &send_codec_spec.codec_inst,
-                    &send_codec_spec.enable_codec_fec,
-                    &send_codec_spec.opus_max_playback_rate,
-                    &send_codec_spec.enable_opus_dtx,
-                    &send_codec_spec.min_ptime_ms,
-                    &send_codec_spec.max_ptime_ms);
-    }
+      voice_codec_info = engine()->encoder_factory_->QueryAudioEncoder(format);
+      if (!voice_codec_info) {
+        LOG(LS_WARNING) << "Unknown codec " << ToString(voice_codec);
+        continue;
+      }
 
-    // Set packet size if the AudioCodec param kCodecParamPTime is set.
-    int ptime_ms = 0;
-    if (codec->GetParam(kCodecParamPTime, &ptime_ms)) {
-      if (!WebRtcVoiceCodecs::SetPTimeAsPacketSize(
-          &send_codec_spec.codec_inst, ptime_ms)) {
-        LOG(LS_WARNING) << "Failed to set packet size for codec "
-                        << send_codec_spec.codec_inst.plname;
-        return false;
+      send_codec_spec =
+          rtc::Optional<webrtc::AudioSendStream::Config::SendCodecSpec>(
+              {voice_codec.id, format});
+      if (voice_codec.bitrate > 0) {
+        send_codec_spec->target_bitrate_bps =
+            rtc::Optional<int>(voice_codec.bitrate);
       }
+      send_codec_spec->transport_cc_enabled = HasTransportCc(voice_codec);
+      send_codec_spec->nack_enabled = HasNack(voice_codec);
+      bitrate_config_ = GetBitrateConfigForCodec(voice_codec);
+      break;
     }
+  }
+
+  if (!send_codec_spec)
+    return false;
 
+  RTC_DCHECK(voice_codec_info);
+  if (voice_codec_info->allow_comfort_noise) {
     // Loop through the codecs list again to find the CN codec.
     // TODO(solenberg): Break out into a separate function?
     for (const AudioCodec& cn_codec : codecs) {
-      // Ignore codecs we don't know about. The negotiation step should prevent
-      // this, but double-check to be sure.
-      webrtc::CodecInst voe_codec = {0};
-      if (!WebRtcVoiceEngine::ToCodecInst(cn_codec, &voe_codec)) {
-        LOG(LS_WARNING) << "Unknown codec " << ToString(cn_codec);
-        continue;
-      }
-
       if (IsCodec(cn_codec, kCnCodecName) &&
-          cn_codec.clockrate == codec->clockrate) {
-        // Turn voice activity detection/comfort noise on if supported.
-        // Set the wideband CN payload type appropriately.
-        // (narrowband always uses the static payload type 13).
-        int cng_plfreq = -1;
+          cn_codec.clockrate == send_codec_spec->format.clockrate_hz) {
         switch (cn_codec.clockrate) {
           case 8000:
           case 16000:
           case 32000:
-            cng_plfreq = cn_codec.clockrate;
+            send_codec_spec->cng_payload_type = rtc::Optional<int>(cn_codec.id);
             break;
           default:
             LOG(LS_WARNING) << "CN frequency " << cn_codec.clockrate
                             << " not supported.";
-            continue;
+            break;
         }
-        send_codec_spec.cng_payload_type = cn_codec.id;
-        send_codec_spec.cng_plfreq = cng_plfreq;
         break;
       }
     }
 
     // Find the telephone-event PT exactly matching the preferred send codec.
     for (const AudioCodec& dtmf_codec : dtmf_codecs) {
-      if (dtmf_codec.clockrate == codec->clockrate) {
+      if (dtmf_codec.clockrate == send_codec_spec->format.clockrate_hz) {
         dtmf_payload_type_ = rtc::Optional<int>(dtmf_codec.id);
         dtmf_payload_freq_ = dtmf_codec.clockrate;
         break;
@@ -2041,7 +1764,7 @@ bool WebRtcVoiceMediaChannel::SetSendCodecs(
     send_codec_spec_ = std::move(send_codec_spec);
     // Apply new settings to all streams.
     for (const auto& kv : send_streams_) {
-      kv.second->RecreateAudioSendStream(send_codec_spec_);
+      kv.second->SetSendCodecSpec(*send_codec_spec_);
     }
   } else {
     // If the codec isn't changing, set the start bitrate to -1 which means
@@ -2051,12 +1774,12 @@ bool WebRtcVoiceMediaChannel::SetSendCodecs(
 
   // Check if the transport cc feedback or NACK status has changed on the
   // preferred send codec, and in that case reconfigure all receive streams.
-  if (recv_transport_cc_enabled_ != send_codec_spec_.transport_cc_enabled ||
-      recv_nack_enabled_ != send_codec_spec_.nack_enabled) {
+  if (recv_transport_cc_enabled_ != send_codec_spec_->transport_cc_enabled ||
+      recv_nack_enabled_ != send_codec_spec_->nack_enabled) {
     LOG(LS_INFO) << "Recreate all the receive streams because the send "
                     "codec has changed.";
-    recv_transport_cc_enabled_ = send_codec_spec_.transport_cc_enabled;
-    recv_nack_enabled_ = send_codec_spec_.nack_enabled;
+    recv_transport_cc_enabled_ = send_codec_spec_->transport_cc_enabled;
+    recv_nack_enabled_ = send_codec_spec_->nack_enabled;
     for (auto& kv : recv_streams_) {
       kv.second->RecreateAudioReceiveStream(recv_transport_cc_enabled_,
                                             recv_nack_enabled_);
@@ -2179,7 +1902,7 @@ bool WebRtcVoiceMediaChannel::AddSendStream(const StreamParams& sp) {
   WebRtcAudioSendStream* stream = new WebRtcAudioSendStream(
       channel, audio_transport, ssrc, sp.cname, send_codec_spec_,
       send_rtp_extensions_, max_send_bitrate_bps_, audio_network_adaptor_config,
-      call_, this);
+      call_, this, engine()->encoder_factory_);
   send_streams_.insert(std::make_pair(ssrc, stream));
 
   // At this point the stream's local SSRC has been updated. If it is the first