Next up in my series of interviews of interest to screenwriters is Nadira Azermai, CEO of ScriptBook the innovative new script platform recently launched.
ScriptBook is an AI driven script analysis and showcase site that is truly cutting edge in its approach… but I’ll let Nadira explain.
Q: First up, how about a little bit of background on Nadira Azermai, how did you get into the industry?
A: I studied Engineering and applied Economics many years ago. I wrote my master’s thesis on using AI to quantify story parameters into predictive models for motion pictures. I was about to start a PhD on the topic but my sister who then was finishing up her own PhD, compared a PhD to a trip to hell. So instead I decided to launch my own startup. Looking back now, running a startup is worse than hell 😊
Q: How would you describe ScriptBook to a new writer or someone interested?
A: ScriptBook’s mission is to democratize the business of storytelling through the art of AI. ScriptBook wants to level the playing field. Our state of the art technology has the power to disrupt the industry. You see, Hollywood hasn’t changed in 100 years. Why would they? The power is owned by Hollywood (the 1% ) while the other 99% (all of us) keep struggling. Only powerful technology will bring about change, will bring down walls & remove everyone standing in the way. I believe that a fully transparent & democratized marketspace with validated content is the future. This is the ecosystem that we should aspire to build. No middle man. Why does a writer need a talent agent to represent him (and give up 10-20% of their earnings)? Why does a writer always get the smallest piece of the pie?
This is my vision of the flow on the ScriptBook marketplace (the fully developed one): a writer uploads his script on the marketplace, the script gets an AI validation, an interested party makes a direct offer to the writer. The writer makes a counter offer because his script has a great financial forecast. He is now in a position to negotiate. Both sides accept the deal. Writer moves on to his next script. End of story.
All of sudden we create an ecosystem where we have a large percentage of working writers, instead of having a large percentage of non-working writers who spend years honing their craft and never get any recognition (not emotionally nor financially).
Q: And what prompted you to develop ScriptBook and when was it launched?
A: I founded the company in 2015 and raised early on $1.4M in seed capital from 2 VC funds. The reason why I developed ScriptBook was driven by the scientific research during my Masters thesis. I was astonished to find that there was so much research out there from top universities on predicting financial performance of movies. But they all had the same problem. All the predictive models were based on parameters that are known after a movie is produced or released (cast, director, release details, marketing, social media, etc.)
It seemed impossible to develop a predictive model that’s based on information prior to greenlighting a project. You see, the only information that you have at early stage is the script, meaning textual data. There was no decision support system that could assist with greenlighting in script-stage. It is incredibly hard to develop models based on data of textual nature. It is far easier to build an AI that works on numerical data. When you deal with textual data (scripts) you need a combination of different subfields of artificial intelligence (natural language processing, machine learning, deep learning). To give you an example: the feature “emotion analysis” for characters: in order to train an algorithm that does this fully automated we had to annotate emotions (we used Parrot’s emotion classification https://www.researchgate.net/figure/First-two-layers-of-Parrots-emotion-classification_fig3_258240889). We needed at least 1 million labels of the 8 primary human emotions (love, fear, anger, etc.) to have an algorithm that is accurate in predicting emotions. So we brought in film students who spent all summer annotating emotions on a sentence level for a diverse dataset of about 1000 scripts. We used these labels to train an algorithm that is capable of performing emotional analysis for characters in an automated way. Every feature that you see in our script technology took a lot of sweat, tears and effort from our team of brilliant data scientists. We’re the only company in the world that succeeded in building an automated script analysis system. A lot of people tend to forget this.
From 2015 to 2018, we were in R&D mode. We wanted to build a system that was unbiased, objective and didn’t need human intervention. Had we built a semi-automated system that still required human input, we would’ve never been able to process a script in 6 minutes. In 2018 we went to market with our AI script analysis system. We had to go through very rigorous technology audits with the major film studios. We validated our technology with the studios, talent agencies & production companies (does the AI work or not?).
We embarked on proof of concepts (which took 18 months) where we had to run our algorithms on scripts. These proof of concepts were two fold: part one was backtesting (running our AI on historical data). We analyzed thousands of scripts from released movies and presented our results (how did our accuracy compare with the accuracy from studio execs). Part two was running our AI on in-development scripts and film projects with a release date in the future (late 2018, 2019 & 2020). Again, we had to present our results to the C-level execs from the studios.
Both, in backtesting and future predictions, ScriptBook’s system outperformed every studio, agency & production company.
Our greenlight accuracy was 87% compared to the industry accuracy of 36%.
Q: There is a lot of really interesting data that you can see as a writer once the AI has been through the scripts, how did you decide on what the AI would focus on?
A: It’s not us who decide what the AI should focus on. AI is able to recognize patterns in our dataset of 30 000 scripts and it connect millions of datapoints (which humans simply can’t do because we’re human). The AI learns which features have successful scripts in common. It learns which features correlate with either commercial success or critical success.
Q: Did you work with Producers to work out what they’d like to see from the analysis reports?
A: Yes we did for some features. For example, the gender equality measure (Bechdel test) was something that we were not familiar with. Entertainment companies specifically asked for this feature. Also, the territoriality forecasts, the film positioning graph and the similar movies are very popular with producers and executives.
Q: A general consensus amongst screenwriters is that Producers are already inundated with scripts, why would they come to you to look for material?
A: That was my initial thought too. I often get invited as a keynote speaker at film festivals/industry markets and after my presentations loads of production companies, film funds, even sales agents come up to me asking if ScriptBook could send over scripts that have a good analysis. I was speaking last at the Berlin International Film Festival (Feb 2020) and it was unbelievable to see how many film funds are actively looking for scripts they could produce/finance. They want a validated pipeline of scripts. Most of these companies were complaining that they were getting tons of shitty scripts (their words, not mine) and they just want a pipeline of scripts that were considered ‘qualitative’ by our AI.
Q: Are prospective Producers vetted before they are allowed access and can you mention who you are working with so far?
A: At this point we consider our marketplace in beta phase. Because of the Corona crisis, we wanted to support all the writers out there by offering free script analysis, so we hurried up to launch this basic version of the marketplace. Our team is currently in development mode so that we can launch a full fledged marketplace in the future. The companies currently viewing the script analyses are our current client-base (2 film studios, film funds, production companies and a few film sales agents/distributors). I can’t mention any names because of NDA’s but I can tell you that Hollywood is deeply ashamed to admit to the use of AI. They still treat AI as a dirty little secret. I can say however that we analyzed script submissions for the Sarajevo Film Festival in 2018/2019. I can also say that Steven Knight (writer of Peaky Blinders) is a big fan of our script technology. Also on Linkedin someone who’s one of our clients came forward (one of our employees posted this on social media). See screenshot below:
Q: Can you clarify the pricing model for us? It appears that an individual screenwriter can currently list a script for free but with some restrictions on the data they then have access to, is that right?
A: We will keep the free offering for screenwriters until the end of 2020. We are currently discussing partnerships with other companies. The goal is to keep this service free for screenwriters. The free offering consists of script DNA features and Audience insights. The financial part is not free.
Q: So a writer can upload as many scripts as they like, just with the restriction on the financial analysis?
A: Yes, as many as they like.
Q: So with the key restrictions for the free service being the financial forecasts on potential performance is hidden for the writer, can this very interesting data be seen by potential Producers?
A: Yes our client base has access to the full analysis. From my experience, I would say it’s a 70/30 split of Producers going straight to the financial forecasts vs Producers going to the story features. Sales agents tend to focus on the territoriality forecast (can they sell distribution rights to different territories and how to price the rights)
Q: Have you had much interest from Producers/the industry?
A: ScriptBook is a Belgium based company. But for some reason almost every entertainment company in Hollywood knows about ScriptBook. We’re also very well known in the film festival scene. We received good press over the years from Variety, CNN, Vogue, etc. This helps in creating brand awareness.
Q: Once a Producer finds someone or a script through the site, what happens next?
A: As mentioned earlier we consider ourselves in beta phase. So for now we keep it simple. When a client is interested in a script, we give them the details + email address of the writer so they can connect. For now we get out of the way. I have to mention though that most companies are currently on a spending freeze due to Corona.
Q: Do you actively shop the high scoring scripts to Producers or is it strictly them coming to the site?
A: Our clients can view the full analyses of uploaded scripts. They have search filters which allow them to look for specific content according to their needs (whether it’s commercial success, films with a female lead, etc). Again, at this point we don’t shop scripts to producers (beta phase) but we intend (once development is over) to add metrics such as ‘most viewed script’, ‘scripts with high commercial value’, ‘scripts with critical success’ etc to the marketplace.
Q: What does the scoring represent to someone within the industry? Is it meaningful and/or valuable to Producers?
A: I can only speak from experience with our own client-base. In the beginning they were hesitant because it’s hard to grasp that an AI can do a far better & more accurate job than someone with 20 years experience in the industry. But the proof is in the pudding. When companies see that our AI keeps outperforming them, they become believers. For example ‘Knives Out’ is a film that so many companies & studios gave a pass. But ScriptBook gave it a massive greenlight early on.
Q: So if a screenwriter told a Producer that their script scored well on ScriptBook, will it hold weight like a normal coverage report?
A: I personally feel that companies follow our AI analysis rather than a normal coverage report. First, a coverage report has value to the writer because the feedback helps in changing certain things in a script. However, a coverage is also subjective and biased. It’s the opinion of one (or maybe 2, 3, 4). An AI report is the weighted average of millions of datapoints. Both, AI & normal coverage are complementary, they are not mutually exclusive.
Q: Have you had any success stories so far that you can share with us?
A: I can’t mention names of companies but we analyzed many films and predicted the correct outcome early in the process. Some examples are: I Tonya, Fences, The Wife, On the basis of sex, Suburbicon, etc.
Q: Of those you mention, was that before release and you were close to the numbers on box office?
A: When we speak about analyzing scripts for clients, it’s always long before the facts (either before production or before release)
Q: There are a lot of people competing for aspiring screenwriter’s attention (and money), from guru’s, through coverage services, and a plethora of competitions. What makes ScriptBook a good investment?
A: I’ve come to this realization too. It’s crazy the number of sites out there competing for screenwriters attention. I think a screenwriter shouldn’t be paying for anything.
We need to focus on a business model where the industry pays for services, not the writers.
The reason why ScriptBook is a good investment & how it differentiates itself:
- Free hosting.
- You get an objective, unbiased script assessment for free.
- Your work is showcased.
- You increase your chances of selling a script.
- You get free analytical insight.
- You can upload as many scripts as you like.
- You can upload as many rewrites you like.
- You get the validation you’re looking for.
To conclude, I think it’s a big win for screenwriters.
Q: When are you hoping to come out of Beta?
A: Good question, our team is currently working remotely due to corona, we have delays at the moment with development. I hope we’ll have a complete marketplace somewhere in the 4th quarter of 2020 (though I need to check with our CTO to be sure)
Q: And what about future developments for ScriptBook?
A: Too many to list. This is for a next interview…
Okay, now for some getting to know you questions…
Q: Fave movie?
A: I have very poor taste in movies. For some reason I just love Christmas movies. Doesn’t matter how horrible they are. Big fan of Jane Austen films too.
Q: Fave script?
A: I read and labeled many scripts when we were creating the ‘advanced sentiment’ algorithm (emotion analysis). I didn’t pay attention to the storyline as I was too focused on annotating emotions. Can’t really say I have a favourite script.
Q: Fave food?
A: Waffles, pancakes, fries.
Q: Fave drink?
Q: Fave sport and team if applicable?
A: I hate sport, especially football, I hate it with a passion.
Q: Fave thing to do outside of managing ScriptBook?
A: I love spending time with my family, I enjoy being out in nature with my dad. Long walks while having deep thoughts are my favourite. I love the sound of silence. I often have imaginary discussions with the greatest philosophers of all times. Planetary science is also a favourite of mine. I enjoy trying to find connections between planetary traits, life on earth and our ethics & morality.
Q: Any final words of advice to the aspiring writers out there?
A: Take your power back. I’ve witnessed how some talent agencies talk about screenwriters and how they handle submitted scripts. One would think why in God’s name would I ever want representation. Now that I know how the industry works from the inside, I would never ever give up my power to anyone. Use the AI to your benefit and chase opportunities.
Wow, thanks to Nadira for a truly fantastic interview. Now what are you waiting for, go checkout ScriptBook – http://www.scriptbook.io/