Editor’s note: This article is from the micro-channel public number “Dear Data” (ID: deardata), Author: Tan Jing, 36 krypton release authorized.
Edit Shiba Inu, also contributed to this article
“You Young” has won many Hong Kong Gold Awards. If Zhou Dongyu and Yi Xuan Qianxi are replaced, will the outcome end?(Thousand Paper Cranes, please be angry)
“Ru Yi Biography” has a hot spot of “44-year-old Zhou Xun forced girl”, in addition to Zhou Xun, is there a better choice?
These direct torture of the soul makes the casting director
From (qiang) Gong (xing) to (bei) Wei (guo).
For decades, the legend of the movie’s achievements has been floating in the wind.Thousands of auditioners and casting directors find the most suitable actor for the film role, so that the audience is left with the impression that an actor is born for a role.
Even Ma Hao, the director of the variety show “Sons of Tomorrow”, said, “The system of casting is not done just because variety shows are needed, but continuously throughout the year.”
There is no need to talk about the importance of casting.
Cinelytic, a technology startup in Los Angeles, USA, is prepared for film and television investors. It uses powerful machine learning algorithms to discover and find “unique recipes” for movies.
As it happens, Warner Bros. and Sony Pictures are both customers of Cinelytic.With the cooperation of a powerful production company, Cinelytic’s R & D team developed an actor analysis platform.
Use artificial intelligence to answer: “The secret of producing blockbuster commercials?Casting
In other words, Hollywood’s unique recipe for “choosing actors” incorporates artificial intelligence technology to calculate the “box office influence” in the history of stars and to speculate and predict the “performance of new actors in actors”.
The charm of actors and characters is an incredible combination that achieves each other.A successful role will make people remember.And this charming “recipe” can also be analyzed by artificial intelligence technology.
Actors are the variables of a movie. Types, plots, climaxes, colorful passages, and dialogues are all 100% decided by people in the past. It is people’s understanding, experience, and talent of the film.
A film produced by Zhou Xingchi is more likely to make the film a “box office guarantee.”
Artificial intelligence thought: after “Star Girl”, I will have my own girl.
Film is a century-old art. During this period, technology has always accompanied the film.
Why don’t we find new ways to change angles and “understand” art?
What the hero sees is the same, and the film giants Warner Bros. and Sony Pictures think so:
The artificial intelligence software does not lay off the “director of casting” but provides a decision framework for understanding “how different elements have a huge impact on business performance.”
With artificial intelligence, it is possible that the “shadow casting” can also converge.
This kind of “unspoken rule” is similar to what we call “casting couch” in Hollywood. “Victims” are usually young actors who longed for fame.
Scarso, director of the world’s number one reputation management company Ingenious Group, agrees.Artificial intelligence guides movie investment and works best as a supplementary tool (it is a necessary medicine for home travel and refreshing).
Quantitative trading companies that predict stocks’ ups and downs have already made a lot of money. Why can’t movies predict daily profits?
The “viewing volume” determines the status of film lovers. If you want to mix artificial intelligence in this circle, you need to see enough “raw materials” (data).
Cinelytic classifies media metadata, a huge database (contains data from 8 million pieces of content) merges the ideas of data scientists and movie critics, and “invents” various methods that have never been thought of to analyze movies.
(The Douban film team cast a disdainful look.)
Artificial intelligence to understand movies and analyze movies is no longer a problem:
Analyze movie watching emotions (anger, sorrow) from the color and texture of text, voice, and video.
Find “faces” (face search + face comparison technology) from lengthy movies or simply replace “faces” (deepfake technology).
Read “understand” content from actors’ expressions and behaviors (expression recognition technology, behavior recognition technology).
At the product launch conference, the advertisements were bombed and bombarded. The audience was drowsy and playing a movie was a trick to wake up the audience and avoid embarrassment.The source of this information is “100 Tips for Waking Up the Audience at the Product Launch” (Hey).
Technology companies have long targeted the film market.
IBM and 20th Century Fox have created the first ever movie trailer produced by artificial intelligence, the science fiction thriller Morgan.
So far, artificial intelligence seems to have been dissatisfied with the previous credit.
The Belgian ScriptBook company was founded in 2015, claiming that its algorithm can predict the success or failure of a movie only by analyzing the script.
Vault, an Israeli startup founded in the same year, assures its customers that it can predict who will watch their movies by tracking and tracking the receipt of trailers online.
Another company called Pilot provided a similar analysis, promising to predict box office revenue with “unmatched accuracy” 18 months before the movie ’s release.
In the American drama “House of Cards”, Netflix’s success lies in the use of artificial intelligence as its “chief of staff” strategy.
Although conservative film industry executives may be reluctant to admit that “data-driven film development”, digital streaming companies have an inherent advantage, they have hit the tradition, and may “come later.”
The unique artistic charm and image appeal of the film determines that the film has a large audience and huge social coverage.
For decades, it has been the job of the Hollywood marketing team to gain an in-depth understanding of the movie watching community.Insights come from focus groups, questionnaire surveys and interview summary surveys.
Now, these researchers decided to use Google’s products (deep learning framework and machine learning algorithms) to analyze the similarities in the movie trailers of the studios, so that the producers can predict the composition of its most commercially viable audience.
Don’t think that the moon is only the circle of Hollywood.
Darling Meow (pen name), the project manager of Daguan Data Text Application Data, told “Dear Data” that artificial intelligence understands the voice of “people”.
“Different types of movies, such as science fiction, comedy, sports movies, etc., have different actors and actresses in different styles. For example, the film and television version of” Ghost Blowing Lamp “has three different styles of Shirley Yang, followed by Shu Qi.”Dragon Quest”, Chen Qiaoen’s “Fine Ancient City”, Zhang Yuqi’s “Longling Grottoes.” He talked.
There is a view that the character played by Shu Qi has the personality of Laura in the “Ghost Blowing Lamp” version, which is free and easy, but lacks a little calmness and generosity.Chen Qiaoen’s version is in shape, but the movement is not free and easy.
If you use artificial intelligence to choose the role in advance, to avoid mistakes in character selection, what technologies can be used?
His answer is: “Big data technology + NLP semantic analysis technology. (NLP, natural language processing technology)”
100 actors have been delineated, but it is still untimely to decide who will perform this role. You can collect media reports and comments and comments on pre-selected roles from the Internet, and then use NLP technology to reduce the noise of the collected data (Get relatively “clean” PGC and UGC data).
The evaluations of the public media and individual netizens reflect the impressions and evaluations of “most people” on actors, and they can get a relatively objective comprehensive evaluation of star tones.(Like in elementary school, the semester conduct comments written by the teacher every semester.)
Among them, the more critical step is to extract the evaluation / personality tags related to these stars (through NLP technology), and then use data reduction to integrate the stars and their corresponding tag information into the two-dimensional plane, so as to obtain the tonality of each starlabel.
Finally, compare the character sets of these tonal tags and screenwriters. If they are completely consistent, it is best to choose them. If they do not match, choose as close as possible.The following is a simple schematic diagram of the mining results, taking the simplest case as an example:
Three actresses (played by Shirley Yang) and ten less than tonality labels, lively, capable, knowledgeable, calm and so on.Using the point where they are as the center of the circle and a certain radius, you can anchor the tonality label of the star.
Another idea is to first analyze the social networks (Weibo, WeChat, Facebook, Twitter, etc.) of the pre-selected stars, and extract semantic tags that can reflect their deep emotional characteristics through semantic analysis.
Then perform cluster analysis to get its composite emotional motivation composition, compare some of the emotional motivation labels with the highest proportion with the preset tonality of the character, and delineate the best candidate of Shirley Yang.
Looking back to the “Ru Yi Zhuan”, a big heroine whose drama fell on the heroine.
“Actresses with high reputation in acting (over 65%) among actresses have been selected, including Zhou Dongyu and Yang Zi in their twenties, and Sun Li, 35-50 and so on.”
“Keywords for Ruyi include strong, persistent, stubborn, etc. The actress’ personality hot words are matched with Ruyi’s character tags, and the size of the ring represents the matching degree of the star’s personality tag and the role of Ruyi. The larger the ring,The higher the degree of fusion of representative word frequencies. ”
“Zhou Dongyu has a high degree of matching, but the story focuses on Ruyi between 35 and 50 years old, and the actual method is still selected from the 35-50 year old actors on the right half.”
Keywords such as Yi include strong, persistent, and stubborn.
Aiman Data is a leading entertainment big data service provider, and its staff confessed in an interview in Business Week Chinese Edition ©.
Aiman Data is also quite technically strong. Zhu Xiaoyan, the company’s director, is a professor in the Computer Department of Tsinghua University and the director of the State Key Laboratory of Intelligent Technology and Systems.
So far, artificial intelligence has not been 100% correct.To be fair, neither human beings.Dark horse movies are always coming and going, so why not let artificial intelligence and humans predict together?
A similar point of view also comes from Dr. Shen Cheng, a senior data scientist at Elsevier, a leading global information analysis company.
And, he believes that the relationship between artificial intelligence and the film industry seems to be more entangled.
He told “Dear Data”:
“Movies are an investment that naturally requires high returns and good box office. If there is a chance to discover new stars, the cost of actors will be reduced, and the question of” what are the characteristics of actors “will be predicted. Similar to the movie” Millions of Gold Arms “from nearly fourAmong the tens of thousands of people, two baseball pitchers have been discovered. The selection of such sports seedlings. That is, which subtle factors the actors meet can increase the box office. ”
Nor does it rule out some directors’ stable casting habits.For example, know the question: “Why does Jia Zhangke love to use his daughter-in-law Zhao Tao to play his film?”
In addition to casting, artificial intelligence, please answer quickly:
“How do you make commercial movies famous and have a good reputation?”
Waiting online is quite urgent.
He said: “Before the emergence of artificial intelligence technology, making choices among huge uncertainties is all based on human experience and artificial grasp. Now, for example, the selection of film release time and the creation of a film-making environment, Investment in film promotion, selection of online and offline distribution platforms …
There is a lot of data analysis and social network analysis technology in these works to support the fission-like spread of film reputation.”
Dr. Shen Cheng cleared his throat and continued to talk:
“It is also possible to predict whether a movie is suitable for a sequel, because there is a risk of breaking it. Artificial intelligence makes predictions by narrowing the scope of uncertainty.
The biggest feature of artificial intelligence is that through the analysis of multi-factor historical experience and data analysis, the original relatively large uncertainty is turned into a relatively small one.Or, turn the originally small probability of success into a relatively large probability of success.”
“The saddle does not leave the horse, and the armor does not leave the body.” It is said that artificial intelligence is used in many industries.
For 100 years, technology has been supporting the development of the film industry. If you let “artificial intelligence” predict itself, it will predict that it will be a rising star in the film industry.