Quoted from FrankM So I gave this a whirl by uploading Who Wants to Be a Princess? and got the feedback. The AI classified it as "adventure" instead of "family," which may mean I need to strengthen the theme. The script is probably not structured quite the way the AI expects it to be, thinking Act II started about six scenes later than it really did. Oh, and it tagged the romantic interest as the antagonist. That may mean I need to find a way to introduce the real antagonist earlier. It did mark the real antagonist as the least likable character, so at least that's working I used the feedback form to let them know what they got wrong about the script, trying to sound as constructive as possible. For one thing, it looks like there are zero G-rated scripts in their sample set, and that may be skewing the AI to read things as if written for an older audience. I'd say this system can find issues with your script, especially ones you're likely to miss when reading your own work. But not sure I'd trust it to make a go/no go decision on a real project, though. Edit: The AI says this film (as it understands it) would be more popular with critics than real audiences. I'm getting an Oscar!

Still, dunno what it does/means for any of the metrics.  Even the basic stuff like emotions.

Let's say Dave is having a chat with the wife and his wife asks him if he loves her. Dave responds:

DAVE
You fucking bet I do.

What is the emotion in this dialogue that is going to be labeled by the machine? Is it going to be love?? I'm guessing it is going to be anger.

Okay - so I put in a second script

Brother's keeper - a western - which has not done as well as 7 days in La Suerte in contest or with producers.

First my scores for 7 days
IMDB WAS 5.9 (just under average - I can live with that from a computer program because most will judge on a curve.

Here's my problem - they said it was 97% crime - it's a dream story that's contained in a crime situation

The character analysis -
My hero. who comes back to save the day, was less likable than the town sheriff who sold out he town. He was only as likable as the villaness who kills at a whim - even her gophers.

They also had a character listed as Cindy - I checked and checked - no cindy in the script.

Emotions - in the scene where the hero's wife leaves him - they had happiness as the top emotion.

ratings it said that the audience would be a gegative 43% (yes-43%) with their opinions to friends - two producers who read this said it was a very audience centric script with a happy leave everybody smiling ending but wasn't enough of a financial reward.

In the last scene where the hero (back from the dream) realizes his life long dream - quits the force and buys a restaurant they said the hero emotion was 53% happy.

2nd script Brothers keeper
Better scores 6.1 in IMDB and higher across board.

Its a dreamed and they had 54% for comedy and 34% for drama - that's okay but then they compared it to very dark serious crime dramas. (none are even in the same genre.

Okay my problems
Key words - Dismount? Hub? It's a western they mount and dismount.

missed genre equity - only two females - but they are much bigger parts then 7 days - major roles.

Had a small PG rating - 7 days had none - this one had family violence - hangings and mass killings. 7 days because it;'s a dramedy I went out of my way to hide the actual violence.

They got most of the likability right (Maybe I got it right this time)

The comparisons - one western - the others were all over the place.  

My final thoughts
I coached for 7 for 40+ years and worked with my analytics programs. The problem with analytics is that they're only as good as what's put in.

They fed it a lot of scripts and the results - financially and view wise. If it's not in those script modes youre not going to get good rating. The more you follow those norms - you're gonna get a good score. (i.e if the critics raved a film that yours mimics high you get a high critic score because that's the data that's in the program - yet your script may suck - not the same writing quality of the first one)
Computers can't rate quality - coaching example - the analytics I used in coaching gave me great tendencies but couldn't tell me how good the opposing players were.

Brothers keeper and 7 days pave my point - Brithers keeps more tradition structure higher rating - 7 days dramedy dream sequence - lower rating.

Interesting but I'll pass and I think producers will.

Two final words -
I agree with dave - I don't get the emotion thing
How the the avengers on that last graph appeal to a very wide auduiance (it was for marvel junkies) and have a low audience liking - the sequels made millions - somebody must've loved the first one. (brothers keeper a western got a wider viewership in the graph than 7 days an dramedy. when did westerns come back!) AND ...

Did I write this much

As already pointed out, it's an AI driven analytics piece that has learnt from the 000s of scripts that it's gone through.

But it's still AI, so fallible.

I've been thinking about the service since I did the interview and my feelings is that the Financials are the key piece of the puzzle for producers, because the ability for a project to turn a healthy profit is what they care about.

So the emotions stuff, likeability etc are probably more interesting to us, as a point of debate, maybe stuff we can learn about in our own scripts and act as a re-write aid.

It's still in Beta too, so be interesting to see where it goes and how it develops.

Quoted Text Here's my problem - they said it was 97% crime - it's a dream story that's contained in a crime situation

Really emblematic of the problem. I submitted four comedy scripts that were classified as dramas.  Of these, one is really a dramedy, two are Rom-coms and one is a straight comedy. So, if the algorithms can't even detect the correct genre, that should tell you there are other macro-level problems. It would have been interesting if they let the writers input some of the information - like genre - so at least they could compare their results with what the writer determined the genre to be - that would allow them to detect errors.

Quoted Text First my scores for 7 days IMDB WAS 5.9 (just under average - I can live with that from a computer program because most will judge on a curve.

I've looked at a ton of scores from different scripts - there are outliers - but they all cluster around those numbers with some exceptions.

Quoted Text The character analysis - My hero. who comes back to save the day, was less likable than the town sheriff who sold out he town. He was only as likable as the villaness who kills at a whim - even her gophers.

Again - this is going to be a context thing which I don't think the AI can handle.  I have submitted In Search of Hysterical Jesus (analysis pending) and I'm dying to see the likeabity scores on Jesus.

Quoted Text Emotions - in the scene where the hero's wife leaves him - they had happiness as the top emotion.

Maybe he was happier than he let on

Quoted Text How the the avengers on that last graph appeal to a very wide auduiance (it was for marvel junkies) and have a low audience liking - the sequels made millions - somebody must've loved the first one. (brothers keeper a western got a wider viewership in the graph than 7 days an dramedy. when did westerns come back!) AND ...

It's impossible to tell if the audience appeal x - y axis is meaningful since the comparable measurable points are after the fact. e.g.,  A Star is Born - a script they placed on based on their analysis is in the same spot that Meet Joe Black is (a script that is just on the chart for reference using actual numbers.) The fact is that A Star is Born did more than 4 times more box office than Meet Joe Black.  

Quoted from AnthonyCawood As already pointed out, it's an AI driven analytics piece that has learnt from the 000s of scripts that it's gone through. But it's still AI, so fallible. I've been thinking about the service since I did the interview and my feelings is that the Financials are the key piece of the puzzle for producers, because the ability for a project to turn a healthy profit is what they care about. So the emotions stuff, likeability etc are probably more interesting to us, as a point of debate, maybe stuff we can learn about in our own scripts and act as a re-write aid. It's still in Beta too, so be interesting to see where it goes and how it develops.

Agree that the financials in terms of production costs are probably going to be the most reliable data - in terms of box office - don't think so.  

Interesting article from back in 2017

https://nofilmschool.com/2017/04/scriptbook-black-list-screenwriting-ai-algorithm

Three things I noticed from the comments

1) financing - still not accurate
   a. If I have 10 scenes but they can be done in one studio how does the computer know that (I had a producer once tell me he can do brothers keeper much cheaper because he had a western town on his set.) Also, if i use subheadings - i get less scenes that somebody who doesn't a computer can't see this.)
   b. Quality of production will change cost.

2. As with your comedies my western came out 92% western - what was the other 8% - and my villain who beats his sons - kills violently - lives without any woman. Beats up the town prostitute got a 47% for love - lol.

Dave - did you notice in the article they charged $100 back then. Now free - think nobody was interested?

Quoted from JohnI Three things I noticed from the comments 1) financing - still not accurate    a. If I have 10 scenes but they can be done in one studio how does the computer know that (I had a producer once tell me he can do brothers keeper much cheaper because he had a western town on his set.) Also, if i use subheadings - i get less scenes that somebody who doesn't a computer can't see this.)    b. Quality of production will change cost. 2. As with your comedies my western came out 92% western - what was the other 8% - and my villain who beats his sons - kills violently - lives without any woman. Beats up the town prostitute got a 47% for love - lol. Dave - did you notice in the article they charged $100 back then. Now free - think nobody was interested?

Yes - my guess is that writers were not interested in shelling out $100 for this service. I'm assuming they flipped their business model under the assumption that if they can get enough scripts they'll end up with something of value that they can sell to producers.

I don't know when, but eventually, there will be a consolidation of these sites. Right now they are striving for different niches. IMO - eventually there will be a consolidation along these kinds of lines. Just as an example:

- SS is great for script hosting, learning writing and writer interaction.
- Script Revolution is great for hosting and for searching by type of script given all the data elements that can be captured when you submit.
- Coverfly is great for competitions
- Story Data has AI

At some point - whether next week or 10 years from now there will be more of an all of the above approach. A writer will submit a script and have the option on that same site of having it hosted, AI assessed, registered for comps,  open for discussion, etc.

I did make a mistake - I forgot there was a Cindy early in 7 days - bit part = a couple of lines

One other interesting thing I found was - I have a character in 7 days called the “Chinaman” - she’s actually a woman. Yet she was listed as a male character.

Quoted from JohnI One other interesting thing I found was - I have a character in 7 days called the “Chinaman” - she’s actually a woman. Yet she was listed as a male character.

If you read the interview with the founder, they added the Bechdel test after it was requested by their professional users, so it’s kind of an afterthought. I’ve had some limited back and forth with them myself and it appears they go by the character name (not, for example, what pronouns you use to refer to the character), and my guess is functional titles like BANK TELLER and LAB TECHNICIAN get assumed male.

For Who Want to Be a Princess?, it classified 80% of the characters as male (spoiler alert: they’re not). That’s almost certainly because names like TALL GLENWOOD CONTESTANT were assumed male. Despite that, the comp films were appropriate: all female leads, with plots about contests or royalty.

No system is perfect, but for giggles I’d like to see how the gender guesser handles the Oompa Loompas (males and females played by the same actor).

Dave

I noticed it's still analyzing "The search for the historical jesus." My scripts took a day on the site. You think he knows you threw him a ringer?

John

I also uploaded my script “Sheean” I was surprised by how it went. I was expecting a much poorer scores. Just a side note, I would love to know what is their algorithm looking for? How would it account for different spellings of the same word? E.g realise vs realize. What about slang or other other terms in the writers  own lexicon.
A fascinating look into world of AI. I don’t think it would totally replace human interaction in the film/tv production industry. Well I certainly hope not.

Quoted from JohnI Dave I noticed it's still analyzing "The search for the historical jesus." My scripts took a day on the site. You think he knows you threw him a ringer? John

Funny, all my scripts were done in less than a day. Jesus has been in progress for more than two weeks - so, yeah - I'm beginning to think they don't know what to do with Jesus (e.g., what if he comes out as an unlikeable character

Then again Hannibal lecter ends up helping the fbi agaent find new killer does that make him like able?
It is interesting though