Can a computer write a screenplay? Yes!

I'm not sure cinema can get more digital than this short film. Yes, it was shot on film and yes, it utilizes a little CGI, but the deeply digital quality stems from the writing: it was written by a neural network, an algorithm.

The film begins with some confusing yet expositive words: 

Just above your smartphone keyboard lives an artificial intelligence. It was trained on lots of texts and emails. And tries to guess what you'll type next. We were curious what would happen if we trained this kind of software on something else; science fiction screenplays. So we fed a LSTM Recurrent Neural Network with these: [long list of sci fi screenplay txt files]. Then gathered a cast & crew for one day. Then we fed in random seeds from a sci-fi filmmaking contest...[contest prompts]....and turned it on. This was the screenplay it wrote:

In other words, the screenplay was sourced from three ingredients: neural network software, sci-fi screenplays, and solicited user contributions. The creation of this film is innovative in its use of software, databases, and collective intelligence. Now, that's one complex digital "author."

The outcome is a 9 minute short film that is just as incoherent as you might expect. What's surprising though, is that the combined sci-fi source material, user input, and smart direction creates something that does indeed look and feel like a sci-fi movie. The chaos of so many authors (seriously, consider the authorship that went into all of the screenplays) churned through an impressive predictive-text-AI-software-package does indeed produce a genre film. Sci-fi is a generous one in that it's very core is about innovation, boundary-pushing, and exploring the unexplorable. This would have been less successful of a test if the genre of choice had been, say, a romantic comedy.

Most of this blog focuses on the use of digital technologies in production and post-production. What's so valuable about this film is that it is a rare case in which the digital dominates in pre-pre-production. I, for one, look forward to seeing further experiments in digital film writing.

Click here to watch Sunspring.


The art of statistics reveals fascinating things about the art of cinema. Seriously, these number crunching cinephiles do amazing things to track the changes of the art form. The best resource, IMHO, is Cinemetrics. Check out the site to see info on shot duration averages for individual films and for the industry as well.

People tend to theorize that shot duration is shrinking as the medium changes over time. That's true, but the details of how and when those changes take place are interesting. For example, some believe that shots shortened up with the early advent of VFX. This was, apparently, due to the poor quality of the effects to the filmmakers would not want to linger on imperfect elements. This only applies to effects films, so you can see that the theory doesn't or shouldn't hold for all movies of that period.

The site provides a ton of data, and great instructions on how to interpret it from acronyms to graphs and more. You can also conduct your own analyses by using the site's free software and  following instructions provided.

Finally, don't miss the Measurement Theory page. It provides an impressive list of scholarly works that take statistical analysis into account within film analysis.

Click here to have fun with movie stats.

And Chill: A bot, not an innuendo.

No, this post is not about the symbolic meaning of "Netflix and Chill." The clever designers of this bot have named it as such in order to conjure the implication, but that part is not that interesting to me. This bot's functionality is what has my attention.

One of the greatest challenges of information glut is organizing it. We've all had that vertiginous sensation of knowing the there's SO MUCH available, but digging through multiple collections is stultifying. The big and small media companies are hard at work to find solutions to this problem, and while they scramble we'll probably continue to see projects like this one.

And Chill is a bot that utilizes a little input from you, and a lot of input from whoknowswhere, to recommend movie choices. You text the app (213-297-3673) asking for a suggestion, tell it something that you've liked, and presto: you get a recommendation. According to an article in Engadget, the app "uses a few different frameworks to detect patterns, attributes, and other factors" to produce suggested films.

The site tells you to either send a text or utilize Facebook Messenger. Hmm. Something's up with that. The developers are Sense Technologies, which does not appear to be a Facebook affiliate, but something tells me that it is. Well, Engadget straight up makes the connection, but I don't see factual evidence. But, such a link would answer my main question: how do you do it? It's entirely possible that this app is collecting enough user data to be able to tailor responses, but it doesn't appear have the massive audience it would need. If it's culling data from Facebook, in collusion with the site of course, *that* would provide an answer. If this bot is reading user data for media references and preferences (they ask you what you like when you sign up!), the bot would certainly have a lot of opinion to work with. If we are to believe Engadget's implication, Facebook is hoping to offer more bot's for media consumption in the future, and this is either one of them or an example of what's to come.

Deadpool and VFX

As a person who pretty much can't stand superhero movies, I think I've found my entry into the genre: Deadpool. My usual sense of agitation that develops after too many action sequences simply never happened when watching this particular movie. The intelligence of the writing, and the surprisingly strong performance by Ryan Reynolds (I wasn't a fan until now), kept me interested and delighted during and between action-heavy sequences. Take note, superhero filmmakers: make them smarter, like Deadpool!

While reveling in smart, dark ideas and words, I was also dazzled by the effects. This video compiled by Visual Effects: Behind the Scenes showcases just a few of the truly astonishing achievements by RodeoFX. You'll see examples of all kinds of VFX techniques in the video. The demonstration of additive/combined layers to create Colossus is particularly illuminating. 

Click here to watch view the a series of VFX befores, durings, and afters. One warning: for some reason people often like to overlay lame rock musical scores over compilations. This one has a track that's particularly grating and repetitive, so hit mute before pressing play.

Analog Aesthetics in the Digital Era

I was recently watching an episode of The Goldberg's when I got this feeling that something wasn't right. It took me a while, but then it donned on me: the image quality is too digital, too perfect for a show about the 1980s. It's clear that painstaking detail goes into recreating the 80s for authenticity but this one glaring but hard to notice (yes, I just said that) signifier of the present remains. For those of us who lived through the low-def era, we know the texture and quality of analog video intimately. And, although I noticed this contradiction in the program, I'm not at all suggesting that they revert to the previous paradigm. However, I do think a one-off episode that looked, sounded, and felt like VHS is not a bad idea (take note, showrunners!).

To meet the nostalgic needs of we older folks, developers have created a ton of "downgrading" filters and apps to make pictures and videos look like analog video and/or VHS. The same is, obviously, available to filmmakers. It happens all the time that a movie will show footage from the past that is clearly brand spanking new, but it perfectly resembles that old, hazy, muffled, scan-lined, footage.

Some clever pranksters have taken these tools to contemporary programs that we're used to seeing in HD/UHD. Check out the videos linked from the Boy Genius Report article on this phenomenon. Game of Thrones as an 80s/90s TV show is the most precious and hilarious thing ever. They not only downgraded the footage, but they also applied the formal properties of 80s/90s TV to the current credit sequence. This is a lesson not only in the flexibility of digital media, but also the dramatic changes televisual formats over time.

Click here to travel back in time with Game of Thrones.

From Shaping Minds to Shaping Faces: Disney's FaceDirector

We don't normally think of Disney as a major player in the photographic plate software business, but they are clearly invested. FaceDirector synthesizes two facial expressions into one. Engadget explains how it works in greater detail and focuses on using it to lower production costs. Didn't get that expression quite right? No need to reshoot: just make it happen by morphing existing footage. While there are many implications to be drawn, I'm most interested in the fact that a studio identified mostly for animation has invested resources into something specifically for photographic filmmaking. One thing is for sure, FaceDirector is yet another indication that actors really don't have to fear being replaced by VFX programmers.

Read Engadget's article about Disney's Face Director