A new GamesBeat event is around the corner! Learn more about what comes next.
Making movies and commercials isn’t for the faint of heart. It’s backbreaking work translating a script to the silver screen, and by the time all is said and done, it’s often months before filming kicks off on a greenlit screenplay.
But Debajyoti Ray, Sadaf Amouzegar, and Nathan Crockett think they can speed things up a tad. They’re the cofounders of startup RivetAI, which develops machine learning-infused moviemaking tools designed to streamline preproduction.
“It helps people to create content much faster,” Ray told VentureBeat in a video conference, “by using AI to augment creativity. Everything starts with data.”
Script breakdowns, storyboards and shot list generation, optimizing schedules, and creating budgets are some of the most tedious parts of preproduction, Ray explained. Once a script is finalized, it’s annotated liberally: Every character, visual or audio effect, set, prop, and location is highlighted in a different color for accounting and scheduling purposes. It’s not unusual for 90-minute movie scripts to span hundreds of pages, making the process a major time sink.
RivetAI’s platform automates things. With the help of natural language understanding and algorithms trained on hundreds of thousands of publicly available movie scripts, it analyzes ingested PDF or Adobe Final Draft files and quickly distinguishes between the props, sets, and other entities in the text with a high degree of accuracy.
“It takes a few minutes to do a process that normally takes three to four weeks,” Ray said.
It’s not perfect, Ray said — there’s a drop-down menu that lets members of the production team correct errors. But the system uses each miscategorization to improve its language processing models, effectively learning from its mistakes.
Automated annotation just scratches the surface of what RivetAI’s suite can do. Its other feat of movie magic is budget prediction: With no more than a script, it can estimate the required number of shoot days and prep days (down to the length of scenes), predict a project’s total budget, and spit out a line-item, department-by-department bill of materials.
That feature is not always spot on, either. In a demo, it pegged Pixar’s animated blockbuster Toy Story 2‘s costs at around $60 million, about $30 million short of its actual production budget. But Ray said it’s meant to give a ballpark estimate rather than a dead-on projection, and that it’ll get better over time.
RivetAI’s tools go beyond highlighting and budget projection. Its generative AI editor can extract a screenplay’s plot and map it visually in a diagram, representing scenes with nodes and the characters within those scenes as concentric circles. Clicking on one of the nodes brings up a window showing the list of characters in the scene along with their emotional state. It gets pretty granular: Using a combination of natural language processing and sentiment analysis, RivetAI’s algorithms assign percentages for traits like “agreeableness,” “conscientiousness,” “openness,” “extraversion,” and “neuroticism.”
Ray said that can help scriptwriters to spot continuity problems. “[The system] can look at a scene in isolation and ask, ‘Is this necessary? Does it help with audience engagement?’ The production team can then drill down to see the projected cost of producing the scene and make a decision about cutting it.”
The ‘human element’
Ray, Amouzegar, and Crockett don’t seem like the filmmaking type. Ray, a Microsoft Research alum, studied artificial intelligence at Caltech, where he went on to earn his Ph.D. Amouzegar, a data scientist, worked at SpaceX before joining RivetAI. And Crockett holds a doctorate in astrophysics from the University of Michigan.
But the trio’s enthusiasm for machine learning is equaled only by their passion for film. One of RivetAI’s first generative AI projects was a riff on the iconic TV series Mystery Science Theater 3000, in which a cast of characters humorously lampoon B movies as they play.
“There are very few other mediums that drive behavior, adoption, and culture,” Ray said.
To that end, they’re adamant that RivetAI’s suite of tools only aid, rather than interfere with, the creative process.
“If you just have AI generate words, they might be statistically sound [and] even profound, but they’re not capturing the human experience,” Ray said. “Our solutions can help amateur writers get close to an archetype that will resonate with an audience [and] enable people to focus more on creative content.”
RivetAI got its start as a part of End Cue, a Culver City, California-based production company founded by film producers Andrew Kortschak, a former Pixar Animation Studios producer, and his father Walter Kortschak, a venture capitalist formerly at Summit Partners. Ray, the former CTO, spearheaded the spinoff that became RivetAI, training and testing the AI on End Cue’s productions.
Ray wouldn’t disclose the names of RivetAI’s initial clients, but said they range from a Fortune 500 company with a corporate video department to production companies submitting budgets for film bids.
VentureBeatVentureBeat's mission is to be a digital town square for technical decision-makers to gain knowledge about transformative technology and transact. Our site delivers essential information on data technologies and strategies to guide you as you lead your organizations. We invite you to become a member of our community, to access:
- up-to-date information on the subjects of interest to you
- our newsletters
- gated thought-leader content and discounted access to our prized events, such as Transform 2021: Learn More
- networking features, and more