初步支持语音双向对讲webrtc插件

This commit is contained in:
xia-chu
2025-11-25 22:26:20 +08:00
parent 5efe843595
commit 5165ac4f74
5 changed files with 245 additions and 13 deletions

View File

@@ -55,36 +55,36 @@ namespace mediakit {
// k=* (encryption key)
// a=* (zero or more media attribute lines)
enum class RtpDirection {
enum class RtpDirection : int8_t {
invalid = -1,
// 只发送 [AUTO-TRANSLATED:d7e7fdb7]
// Send only
sendonly,
sendonly = 1 << 0,
// 只接收 [AUTO-TRANSLATED:f75ca789]
// Receive only
recvonly,
recvonly = 1 << 1,
// 同时发送接收 [AUTO-TRANSLATED:7f900ba1]
// Send and receive simultaneously
sendrecv,
sendrecv = sendonly | recvonly,
// 禁止发送数据 [AUTO-TRANSLATED:6045b47e]
// Prohibit sending data
inactive
inactive = 0
};
enum class DtlsRole {
enum class DtlsRole : int8_t {
invalid = -1,
// 客户端 [AUTO-TRANSLATED:915417a2]
// Client
active,
active = 1 << 0,
// 服务端 [AUTO-TRANSLATED:03a80b18]
// Server
passive,
passive = 1 << 1,
// 既可作做客户端也可以做服务端 [AUTO-TRANSLATED:5ab1162e]
// Can be used as both client and server
actpass,
actpass = active | passive,
};
enum class SdpType { invalid = -1, offer, answer };
enum class SdpType : int8_t { invalid = -1, offer, answer };
DtlsRole getDtlsRole(const std::string &str);
const char *getDtlsRoleString(DtlsRole role);

171
webrtc/WebRtcTalk.cpp Normal file
View File

@@ -0,0 +1,171 @@
/*
* Copyright (c) 2016-present The ZLMediaKit project authors. All Rights Reserved.
*
* This file is part of ZLMediaKit(https://github.com/ZLMediaKit/ZLMediaKit).
*
* Use of this source code is governed by MIT-like license that can be found in the
* LICENSE file in the root of the source tree. All contributing project authors
* may be found in the AUTHORS file in the root of the source tree.
*/
#include "WebRtcTalk.h"
#include "Util/base64.h"
#include "Common/config.h"
#include "Extension/Factory.h"
#include "Common/MultiMediaSourceMuxer.h"
using namespace std;
using namespace toolkit;
namespace mediakit {
WebRtcTalk::Ptr WebRtcTalk::create(
const EventPoller::Ptr &poller, const RtspMediaSource::Ptr &src, const MediaInfo &info, WebRtcTransport::Role role,
WebRtcTransport::SignalingProtocols signaling_protocols) {
WebRtcTalk::Ptr ret(new WebRtcTalk(poller, src, info), [](WebRtcTalk *ptr) {
ptr->onDestory();
delete ptr;
});
ret->setRole(role);
ret->setSignalingProtocols(signaling_protocols);
ret->onCreate();
return ret;
}
WebRtcTalk::WebRtcTalk(const EventPoller::Ptr &poller, const RtspMediaSource::Ptr &src, const MediaInfo &info)
: WebRtcTransportImp(poller) {
_media_info = info;
_play_src = src;
CHECK(src);
_demuxer = std::make_shared<RtspDemuxer>();
}
void WebRtcTalk::onStartWebRTC() {
auto playSrc = _play_src.lock();
if (!playSrc) {
onShutdown(SockException(Err_shutdown, "rtsp media source was shutdown"));
return;
}
WebRtcTransportImp::onStartWebRTC();
// 不支持simulcast
CHECK(!_answer_sdp->supportSimulcast());
auto sdp = _answer_sdp->toRtspSdp();
_demuxer->loadSdp(sdp);
auto audio_track = _demuxer->getTrack(TrackAudio, false);
// 必须包含音频track
CHECK(audio_track);
audio_track->addDelegate([this](const Frame::Ptr &frame) {
// 发送对讲语音rtp流
_sender->inputFrame(frame);
return true;
});
MediaSourceEvent::SendRtpArgs args;
args.con_type = MediaSourceEvent::SendRtpArgs::kVoiceTalk;
args.recv_stream_vhost = playSrc->getMediaTuple().vhost;
args.recv_stream_app = playSrc->getMediaTuple().app;
args.recv_stream_id = playSrc->getMediaTuple().stream;
auto url_args = Parser::parseArgs(_media_info.params);
args.data_type = static_cast<MediaSourceEvent::SendRtpArgs::DataType>(atoi(url_args["data_type"].data()));
args.only_audio = true;
args.pt = static_cast<uint8_t>(atoi(url_args["pt"].data()));
args.ssrc = url_args["ssrc"];
std::weak_ptr<WebRtcTalk> weak_self = static_pointer_cast<WebRtcTalk>(shared_from_this());
_sender = std::make_shared<RtpSender>(getPoller());
_sender->startSend(*(playSrc->getMuxer()), args, [weak_self](uint16_t local_port, const SockException &ex) {
if (!ex) {
return;
}
if (auto strong_self = weak_self.lock()) {
strong_self->onShutdown(ex);
}
});
_sender->addTrack(audio_track);
_sender->addTrackCompleted();
if (canSendRtp()) {
playSrc->pause(false);
_reader = playSrc->getRing()->attach(getPoller(), true);
weak_ptr<WebRtcTalk> weak_self = static_pointer_cast<WebRtcTalk>(shared_from_this());
weak_ptr<Session> weak_session = static_pointer_cast<Session>(getSession());
_reader->setGetInfoCB([weak_session]() {
Any ret;
ret.set(static_pointer_cast<Session>(weak_session.lock()));
return ret;
});
_reader->setReadCB([weak_self](const RtspMediaSource::RingDataType &pkt) {
auto strong_self = weak_self.lock();
if (!strong_self) {
return;
}
size_t i = 0;
pkt->for_each([&](const RtpPacket::Ptr &rtp) { strong_self->onSendRtp(rtp, ++i == pkt->size()); });
});
_reader->setDetachCB([weak_self]() {
auto strong_self = weak_self.lock();
if (!strong_self) {
return;
}
strong_self->onShutdown(SockException(Err_shutdown, "rtsp ring buffer detached"));
});
_reader->setMessageCB([weak_self](const toolkit::Any &data) {
auto strong_self = weak_self.lock();
if (!strong_self) {
return;
}
if (data.is<Buffer>()) {
auto &buffer = data.get<Buffer>();
// PPID 51: 文本string [AUTO-TRANSLATED:69a8cf81]
// PPID 51: Text string
// PPID 53: 二进制 [AUTO-TRANSLATED:faf00c3e]
// PPID 53: Binary
strong_self->sendDatachannel(0, 51, buffer.data(), buffer.size());
} else {
WarnL << "Send unknown message type to webrtc player: " << data.type_name();
}
});
}
}
void WebRtcTalk::onDestory() {
auto duration = getDuration();
auto bytes_usage = getBytesUsage();
// 流量统计事件广播 [AUTO-TRANSLATED:6b0b1234]
// Traffic statistics event broadcast
GET_CONFIG(uint32_t, iFlowThreshold, General::kFlowThreshold);
auto session = getSession();
if (_reader && session) {
WarnL << "RTC对讲(" << _media_info.shortUrl() << ")结束播放,耗时(s):" << duration;
if (bytes_usage >= iFlowThreshold * 1024) {
NOTICE_EMIT(BroadcastFlowReportArgs, Broadcast::kBroadcastFlowReport, _media_info, bytes_usage, duration, true, *session);
}
}
WebRtcTransportImp::onDestory();
}
void WebRtcTalk::onRtcConfigure(RtcConfigure &configure) const {
WebRtcTransportImp::onRtcConfigure(configure);
auto playSrc = _play_src.lock();
if (playSrc) {
configure.setPlayRtspInfo(playSrc->getSdp());
}
// 不接收视频
configure.video.direction = static_cast<RtpDirection>(static_cast<int8_t>(configure.video.direction) & ~static_cast<int8_t>(RtpDirection::recvonly));
// 开启音频接收
configure.audio.direction = static_cast<RtpDirection>(static_cast<int8_t>(configure.audio.direction) | static_cast<int8_t>(RtpDirection::recvonly));
}
void WebRtcTalk::onRecvRtp(MediaTrack &track, const std::string &rid, RtpPacket::Ptr rtp) {
// rtp解析为音频视频丢弃
if (rtp->type == TrackAudio) {
_demuxer->inputRtp(rtp);
}
}
} // namespace mediakit

56
webrtc/WebRtcTalk.h Normal file
View File

@@ -0,0 +1,56 @@
/*
* Copyright (c) 2016-present The ZLMediaKit project authors. All Rights Reserved.
*
* This file is part of ZLMediaKit(https://github.com/ZLMediaKit/ZLMediaKit).
*
* Use of this source code is governed by MIT-like license that can be found in the
* LICENSE file in the root of the source tree. All contributing project authors
* may be found in the AUTHORS file in the root of the source tree.
*/
#ifndef ZLMEDIAKIT_WEBRTC_TALK_H
#define ZLMEDIAKIT_WEBRTC_TALK_H
#include "WebRtcTransport.h"
#include "Rtsp/RtspMediaSource.h"
#include "Rtsp/RtspDemuxer.h"
#include "Rtp/RtpSender.h"
namespace mediakit {
class WebRtcTalk : public WebRtcTransportImp {
public:
using Ptr = std::shared_ptr<WebRtcTalk>;
static Ptr create(const toolkit::EventPoller::Ptr &poller, const RtspMediaSource::Ptr &src, const MediaInfo &info,
WebRtcTransport::Role role, WebRtcTransport::SignalingProtocols signaling_protocols);
protected:
///////WebRtcTransportImp override///////
void onStartWebRTC() override;
void onDestory() override;
void onRtcConfigure(RtcConfigure &configure) const override;
void onRecvRtp(MediaTrack &track, const std::string &rid, RtpPacket::Ptr rtp) override;
private:
WebRtcTalk(const toolkit::EventPoller::Ptr &poller, const RtspMediaSource::Ptr &src, const MediaInfo &info);
private:
// 媒体相关元数据 [AUTO-TRANSLATED:f4cf8045]
// Media related metadata
MediaInfo _media_info;
// 播放的rtsp源 [AUTO-TRANSLATED:9963eed1]
// Playing rtsp source
std::weak_ptr<RtspMediaSource> _play_src;
// 播放rtsp源的reader对象 [AUTO-TRANSLATED:7b305055]
// Reader object for playing rtsp source
RtspMediaSource::RingType::RingReader::Ptr _reader;
// 解析对讲语音rtp流为帧数据
RtspDemuxer::Ptr _demuxer;
// 打包语音帧数据为特定rtp并回复过去
RtpSender::Ptr _sender;
};
}// namespace mediakit
#endif // ZLMEDIAKIT_WEBRTC_TALK_H

View File

@@ -28,6 +28,7 @@
#include "WebRtcEchoTest.h"
#include "WebRtcPlayer.h"
#include "WebRtcPusher.h"
#include "WebRtcTalk.h"
#include "Rtsp/RtspMediaSourceImp.h"
#define RTP_SSRC_OFFSET 1
@@ -1726,6 +1727,7 @@ void push_plugin(SocketHelper& sender, const WebRtcArgs &args, const onCreateWeb
}
}
template<typename Type>
void play_plugin(SocketHelper &sender, const WebRtcArgs &args, const onCreateWebRtc &cb) {
MediaInfo info(args["url"]);
@@ -1748,7 +1750,7 @@ void play_plugin(SocketHelper &sender, const WebRtcArgs &args, const onCreateWeb
// 还原成rtc目的是为了hook时识别哪种播放协议 [AUTO-TRANSLATED:fe8dd2dc]
// Restore to RTC, the purpose is to identify which playback protocol during hooking
info.schema = "rtc";
auto rtc = WebRtcPlayer::create(EventPollerPool::Instance().getPoller(), src, info,
auto rtc = Type::create(EventPollerPool::Instance().getPoller(), src, info,
WebRtcTransport::Role::PEER, WebRtcTransport::SignalingProtocols::WHEP_WHIP);
cb(*rtc);
});
@@ -1831,7 +1833,9 @@ static onceToken s_rtc_auto_register([]() {
WebRtcPluginManager::Instance().registerPlugin("echo", echo_plugin);
#endif
WebRtcPluginManager::Instance().registerPlugin("push", push_plugin);
WebRtcPluginManager::Instance().registerPlugin("play", play_plugin);
WebRtcPluginManager::Instance().registerPlugin("play", play_plugin<WebRtcPlayer>);
WebRtcPluginManager::Instance().registerPlugin("talk", play_plugin<WebRtcTalk>);
WebRtcPluginManager::Instance().setListener([](SocketHelper& sender, const std::string &type, const WebRtcArgs &args, const WebRtcInterface &rtc) {
setWebRtcArgs(args, const_cast<WebRtcInterface&>(rtc));
});

View File

@@ -66,8 +66,9 @@
</p>
<p>
<label for="method">method(play or push or echo):</label>
<label for="method">method:</label>
<input type="radio" name="method" value="echo" >echo
<input type="radio" name="method" value="talk" >talk
<input type="radio" name="method" value="push" >push
<input type="radio" name="method" value="play" checked = true>play
</p>