Tengxun Cloud Media Processing MPS Efficient Transcoding: How to Cut Video Bills in Half with Black Technology?
In the circle of video, live broadcast, online education or overseas short play, such a classic dialogue is often staged between the technical team and the financial director:
Chief Financial Officer: "How did our CDN bandwidth fee and object storage fee overspend again last month? Most of the company's profits are paid to cloud vendors for network fees!" Technical person in charge: "I can't help it, boss. Now the users are too picky. The mobile phone screen is better than the other. If you don't show 1080P and 2K Blu-ray HD, the users will turn their heads and go to the competition. The higher the definition, the larger the file, and the bandwidth and storage will have to soar."
This is almost all video business development to a certain stage will hit the "growing pains":
Image quality
with
Cost
It's like a fish and a bear's paw.
In order to break this strange circle, tengxunyun handed over a fairly hard-core answer sheet--
Media Processing Service (MPS)
. The name sounds academic, but if translated in plain English, it is actually a "super compressor" and "plastic surgery master" focusing on video files"
. Its best housekeeping skill is to be able
Without sacrificing any picture quality (even making the picture look clearer), cut down the volume of video files by 30% to 50%.
Today, we will use a down-to-earth real-life perspective to take you down in depth the black technology behind the efficient transcoding of Tengxunyun MPS to see how it helps enterprises find the perfect balance between saving money and preserving picture quality.
1. the hard wounds of traditional transcoding: "one size fits all" blind compression
Before talking about MPS, we need to know how traditional video transcoding works.
If a user uploads an original video taken by a mobile phone, the volume is 1GB. Due to different formats and extremely high code rate, the network cannot be pulled at all if it is directly shown to other users. At this time, transcoding servers (such as open source FFmpeg) are required to convert it into the H.264/MP4 format with the best compatibility and compress the code rate (Bitrate).
The traditional transcoding method is very "simple and crude" and belongs
One-size-fits-all model
: No matter what content is put in your video, as long as you specify 1080P resolution, the system will assign a fixed code rate (such as 4Mbps).
This brings two huge pain points:
When the picture is simple, bandwidth is wasted: if the video is a PPT recording screen for online education, the picture is almost static. This kind of content is actually as long as 500Kbps code rate is very clear. If the non-fortress gives it 4Mbps, it is purely burning the company's bandwidth fee.
When the picture is complex, it is directly pasted into a mosaic: if the video is a fierce football game, or an action movie with full screen special effects, the picture changes drastically from frame to frame. At this time, a mere 4Mbps is not enough, and serious smearing and mosaic will appear in an instant.
Two
, Tengxun Cloud MPS's Core Ace: Speed HD (Bright Eyes Transcoding)
In order to solve the problem of "blind man touching the elephant" in traditional transcoding, Tencent Cloud MPS has introduced a set of self-research programs for dimension reduction and strike--
Speed HD transcoding (also called Bright Eyes transcoding)
.
This technology no longer uses fixed parameters to knock all videos, but puts a "AI brain" on the transcoding engine ". Its workflow is extremely clever and can be divided into three steps:
1. "In-depth physical examination" of video content"
When a video is sent to MPS, the AI engine first scans it in milliseconds in all directions. Like human vision, it can understand at a glance what scene the video belongs to (whether it is an action movie, a little sister dancing, a live game, or a press conference), and can keenly identify where the main vision (the face of the character, the props in the foreground) is in the picture and where the irrelevant background (solid color walls, blurred lawns) is.
2. Dynamic code rate allocation (adapted to local conditions)
After reading the content, MPS began to "watch the dishes and eat".
When the picture is still or only local micro-motion, the AI will press the code rate to very low;
When the picture appears high-speed racing, explosion effects, the AI will instantly increase the code rate, giving sufficient space to ensure that the picture does not collapse. Through this scene adaptation (CAE) technology, the overall average bit rate of the video is greatly reduced.
3. Quality restoration and visual enhancement (cosmetic magic)
It's not enough to save bandwidth, MPS can also make the video look better. It has a number of Tencent's top self-developed algorithms built in:
Intelligent denoising: a lot of old movies or short videos shot at night, there are a lot of noise on the screen. MPS can clean up these noises without blurring the details. Less noise, the encoder is better compression, the file size will become smaller.
Color enhancement and super-resolution: it can make dimly colored pictures brighter and have higher contrast, and even use AI to amplify an originally fuzzy 720P old video, "brain supplement", into a clear 1080P.
[Image concept: AI video processing where standard encoding leaves visual artifacts, while Tencent Cloud MPS intelligently enhances visual priority training and reduces unnecessary video block bitrate.]
Final result: the video transcoded by MPS speed HD is as good as or even better than the original in subjective vision, but the file volume and playback bandwidth are reduced by 35% + on average. This means that you originally had to pay 1 million CDN fees, but now you only have to pay more than 600,000 yuan.
3. Combat Advanced:
With the speed HD, what other money-saving measures does MPS have?
As a one-stop media processing platform, tengxunyun MPS helped the development team to contract all the trivial needs on the video link. The following functions are also artifacts of "saving money and improving efficiency" in actual operation and maintenance:
1. Embrace the new generation format: efficient H.265 / AV1 transcoding
Many people still use the old H.264 code. Tencent Cloud MPS Pair
H.265(HEVC)
and a new generation
AV1
Encoding provides top-of-the-line hardware acceleration support.
The compression rate of AV1 encoding is surprisingly high, but the disadvantage is that it is extremely power-consuming, and self-built server transcoding can burn the CPU to smoke. MPS relies on Tencent Cloud's huge computing resource pool, which can complete AV1 transcoding with extremely fast speed and extremely low computing cost, allowing your enterprise to enjoy the latest industry dividend perfectly.
2. Audio and video enhancement and copyright protection (watermarking and digital watermarking)
Anti-piracy is the first priority for teams that go to sea for short plays or pay courses.
MPS supports adding a dynamic image watermark to a video with one click while transcoding. Even better, it supports
invisible digital watermark (blind watermark)
. There seems to be no trace on the outside, but if someone dares to record the screen or pirate your video, you only need to upload the video to Tengxun Cloud for analysis, and you can extract the user ID hidden in the pixel and catch the inside ghost.
3. Intelligent video slicing and micro-drama workflow
MPS provides an out-of-the-box automated workflow for micro-skits and short video streaming that are currently on fire. The video is transmitted, and the background is automatically sliced (M3U8 format is generated), a few seconds of moving picture (GIF) cover is automatically generated, and the first 10 seconds of automatic extraction are generated to generate free trial clips. All the logic is configured in the console several times, and the developer does not need to write a line of code.
4. Pit Avoidance Guide: How to Configure MPS Access Is the Most Cost-effective?
In the actual selection of enterprises, many novice students often open products but do not use their ideas. Here are two real-life experiences of avoiding pits:
According to the traffic scale, decide whether to turn on "speed HD": speed HD (bright-eyed transcoding) consumes a AI of computing power, and its transcoding unit price is higher than that of ordinary standard transcoding. Accounting logic: if your video is a small long tail video and only a few people watch it after uploading, the saved bandwidth fee may not be enough to cover the high transcoding fee. On the other hand, if your video is a hot video, explosive short play, or live broadcast of a big class, tens of thousands or tens of millions of people will watch it repeatedly, and the high transcoding fee will be in front of massive CDN bandwidth savings, totally negligible. In a word, the more videos are played, the more brainless it is necessary to turn on the speed HD!
Make good use of intranet transmission: when configuring MPS, the input source video and the output transcoded video should be placed in tengxun cloud object storage (COS) in the same region as MPS as far as possible. In this way, the video is saved
The transmission between the storage and transcoding servers is tengxun cloud intranet, which is not only full of speed, but also free of charge for external network traffic.
5. Summary: "Fine Throttling" in the Digital Age"
In today's era when the Internet dividend is peaking and all walks of life are shouting "reducing costs and increasing efficiency", the model of barbaric growth and extensive operation is no longer feasible.
The core value of tengxun cloud media in dealing with MPS is that it turns the originally vague and subjective "video quality" into a hard science that can be accurately calculated and optimized through AI technology. It makes us understand that providing users with a high-definition experience does not necessarily mean paying expensive "bandwidth protection fees".
By handing over the audio and video compression, repair and compliance work at the bottom to Tengxun Cloud MPS, a professional cloud expert, enterprises can not only save a large amount of heavy bandwidth bills, but also free up valuable manpower to knock down business logic. This is the most elegant and efficient survival rule in the cloud computing era.
