Some time ago, an audio social software focused on instantaneousness became popular.
Audio social spring is coming?
As a voice-based social software, its gameplay is very simple. In each room, there are three roles: host, guest, and audience. After the host creates a room, chat with the guests, the audience can listen, the three identities can be changed after the host agrees, and the host can also invite the audience to interact with the microphone. They communicated in the form of voice, and they burned as soon as they listened. This is a typical real-time voice chat room scene.
So where is its innovation?
1. Innovation in content and gameplay: KOL celebrities vs. opinion leaders
2. KOL celebrities: the development of grassroots culture, KOL celebrities have the same lifestyle as ordinary people, and are consumed by more users
3. Opinion leader: Through his many years of accumulation and influence, he has completed the industry's supply of goods in a certain industry. He is a star in the industry with his own aura and is sought after by everyone.
Based on the content, gameplay, and subcultural innovation of the audio social circuit, with the improvement of audio quality in the 5G era, it will truly usher in the spring of industry explosion. Compared with text, the volume of voice information is larger and more personalized. The emotions and information contained are richer, and it is foreseeable that more social gameplay and scenes will be created. However, the rapidly erupting demand for real-time audio and video still faces challenges. A mature audio and video technology system has become a necessary guarantee for the rapid development of products. At the same time, the audio social scene has spawned new compliance requirements. How to ensure platform audio content Compliance has become a rigid demand for the steady development of products.
What is a mature technology system?
The construction of a set of audio social networking is not complicated. The original audio-video and real-time interactive technologies that seem to be very advanced have become easy to access.
This set of seemingly complex audio social logic, here is a technical disassembly.
As we deal with complex problems, we will first perform block processing, the technical architecture of audio social can also be disassembled from a business perspective: the voice interaction of opinion leaders, the voice processing of opinion leaders, and the audio acquisition of fans. Look:
1. Voice interaction of opinion leaders
With the blessing of Qiniu Cloud QRTC's real-time mic-linked product, an opinion leader’s topic room is created through the easy-to-use room creation logic. After joining other opinion leaders’ rooms, the opinion leaders use the online mic-linked room to perform real-time voice Interact and communicate on preset topics.
Qiniu Cloud’s QRTC is based on the open source WebRTC, and has been verified by its own R&D capabilities and many customers, ensuring that even though opinion leaders are located in many places, or even in different countries, they can also guarantee real-time communication and interaction delays. In only about 150ms, although opinion leaders cannot meet each other, they are as smooth as face-to-face communication.
2. Voice processing of opinion leaders
After the voice communication content of opinion leaders is optimized and processed in the cloud and information is reviewed, the smooth conversation voice is distributed to the outside through a mature live broadcast distribution network.
In such a cloud processing process, not only the integrity of the information exchanged by opinion leaders is guaranteed, but also the content screening and optimization of the information can be completed.
3. Fan's audio acquisition
Supported by the live broadcast function of Qiniu Cloud, the information exchanges of opinion leaders are presented to fans in the form of voice. Let the fans of opinion leaders listen to the voices of their idols as if they were in a room.
At the same time, many years of technology accumulation on the client allows listeners to obtain the dialogue voices of opinion leaders with the best user experience even though they are in different network environments.
So, from the perspective of access, what is the access of the language chat room?
1. R&D access for opinion leaders:
Here, developers are provided with the SDK content of different systems such as Android, iOS, Web, and Mini Programs. After the SDK is introduced, the following 5 steps can be completed to complete the R&D access of the opinion leader:
Complete audio and video core initialization: used to initialize the core capabilities of Qiniu audio and video interaction in the SDK;
Entering the room: Build a room and realize the reception of opinion leaders. In order to ensure the quality of the communication between opinion leaders, we currently support 14 opinion leaders to communicate at the same time;
Release voice track: monitor and collect the voice information of opinion leaders, and establish calls with other opinion leaders;
Check out: Realize the multi-party perception of opinion leaders after they leave the room;
Destroy: realize resource recovery after the overall process ends.
2. The business logic processing of the server:
After the opinion leader side completes the room creation and room entry operations, the server side implements the live repost logic of the conversation content of multiple opinion leaders through the following three-step processing:
Connect to the server SDK to complete the authentication logic support;
Completion of the support of the callback logic, used to handle the processing of event notifications in different rooms;
Established a confluence retweet task, user opinion leaders exchanged content, and listened to more fans
3. Listening access on the fan side:
Fan-end Qiniu Cloud also provides support for different versions of player SDKs such as Android and iOS. After the SDK dependency is introduced, it supports player initialization and assigns the obtained live broadcast address to the playback link of the player. You can complete the fan listening support of different systems.
Content review under supervision
With the standardization of domestic policy on the management of online platform speech, the content review faced by social platforms has become more and more stringent. Compared with traditional audio content review, the online review of multi-person real-time voice in the chat room scene is very complicated, especially for social products with higher daily activities, and the cost and difficulty of voice content review is greater. Because voice auditing has three basic technical problems in addition to the basic text classification technology, namely:
Speech recognition: Internet speech scenes are often accompanied by strong background sounds, fast speaking speed, unclear words, and serious accents. Compared with ordinary scenes, speech recognition is more difficult;
NLP: Politically-related, pornographic, abusive and other illegal audio expressions are varied and obscure, and require extremely high semantic understanding;
Voiceprint recognition: pornographic content such as groaning and wheezing is easily mixed in dialogue, singing and even background sounds. The voiceprint features are subtle and difficult to distinguish, requiring a strong voiceprint recognition ability.
Qiniu Cloud provides the identification of pornography/advertising/politics/violation and other content, and the ability to recognize wheezing sounds for content review scenarios of real-time audio streams. It also provides two access methods to help customers improve audit efficiency and purify the network environment:
Live streaming review API-suitable for live streaming scenarios. Real-time monitoring, return results within 3 seconds;
File review API-suitable for voice messages, files, and short videos. It can be reviewed first and issued later.
The rapid development of audio social networking benefits from the different advantages of traditional social media such as voice and text. For the proper communication of emotions, Qiniu Cloud, as a leading one-stop cloud platform as a service (PaaS) provider in China, provides a platform for such products. A complete set of mature audio technology system and compliance technology solutions effectively help customers focus on business innovation and achieve rapid growth.