Introduction

What is exactly Per-Title video encoding? The title says everything, it is the process where video is encoded based on every type of video content. Before we dive into this article further, let's spend a few minutes to talk about the basis of a video. Video quality depends primarily on two factors: resolution and bitrate. This concept has gradually gained popularity and reached many people, with the term SD, HD, FullHD, for those who use phones, computers, tablets, TVs, screen photos, and for those who regularly use video from Netflix, Youtube, Facebook, and media channels using the Internet.

For videos playing on Netflix, Youtube or Facebook, etc, if people notice that there will be a lot of quality selected video quality (profiles), from low to high, equivalent to standard SD with low bitrate to medium HD bitrate and FullHD bitrate, and UltraHD bitrate has been very high in recent years. With said profiles, viewers will have the best video viewing experience, but the transmission bandwidth and infrastructure for storing video content will be an enormous trade off.

In other words, the quality of the video depends on its bitrate and its size. For example, a 10-minute HD quality video with a bitrate of 2Mbps will yield a capacity of approximately 150 MB, such as FullHD 10 minutes with a bitrate of 4Mbps, with a capacity of 300 MB. It can therefore be seen that bitrate plays an important role in the display of video content quality.

Over the past decade, technology solutions companies have found solutions as well as technological methods in order to try reducing the bitrate, yet still maintain the image quality in many ways, including how to carry out standard compression research. The most optimal (see also Overview AV1), in order to reduce bitrate more optimally, one can think of a method based on video content that is Per-Title Video Encoding, the appropriate bitrate must be selected instead.

The basis of Per-Title Video Encoding

Video Encoding process requires many processes and many factors to implement Per-title video encoding. Essential elements include: Collecting genres, classification of categories, genre analysis, collecting videos by category, analyzing videos by category, and testing methods for testing bitrate to match each type, measure comparison, and finally store the results of each category of video analysis.

First, let us discuss about collecting genres: for videos, there will be many genres such as animation (animation), action, life (style), landscape (landscape), darkness, music videos, presentations, performances, etc. Taking this into account, we can realize that countless genres count. But should each of these categories be carefully considered? Response: There is no need for such categories because this will be very difficult to handle for the previous and post-processing stages and quite expensive. Instead, analyze and include the categories in a category that we have defined.

Encoding process in metaphor illustration

Taking the first example of early animation concept to analyze and categorize genres. For audience, animation is assumed to be simply animated, not a character, a real scene, but a drawing or computer sketch. So how to properly understand animation? We'll go into many kinds of animation to say people imagine what I'll think about the first animation? Doraemon, Dragon Ball, Tom & Jerry or Pokemon? It's animated, but here animation is 2D sketch animation, which means simple animation. So animation isn't going to be like that? Give a thought about it. Even if it's just a video game, a 3D animation or a variety of 3D animation like Frozen, how to train your dragon, find Nemo, etc. What we are talking about is list in 3D video.

The more details of an effect, with very complex animations such as video games, as a result, people tend to imagine their difference in a simple animation genre and a more complex animation genre. For example: explosions, movement effects, vehicle landscaping details such as PUBG, etc At the end of the day, we only classify 2 categories for easy access. Simply put, with simple animation, will we now return to bitrate, do we need to select a high bitrate for it? Choose suddenly if you want, but please say that no matter how high you set it can't be more beautiful because it has reached the pixel threshold.

(480p at 1750 Kbps vs 1080p at 1540 Kbps - Source Netflix)

Sometimes we assume the higher the bitrate, the better it looks. That's true if you want to store it in the best quality, but it has a good bitrate threshold, which is enough even if you try to select a high bitrate while exchanging it again and lose the video. Speaking of which, how much is it difficult to select bitrate? Even when we are writing a post, we still don't know how much. By testing it, we will figure out a way to know how much is appropriate. we will test the rules and measure the bitrate and compare the results to obtain the final result.

Credit to: Wrong (hungtc@uiza.io)