There’s so many films to watch these days from a myriad of sources, but have you ever stopped to think how many you will actually watch in your lifetime? As an overall fraction, the answer is not many. Do you then just watch what is on offer or would you rather watch only the best films for you based on all films ever created! You know the answer. You need a personalised film recommendation engine; this is where Film Fan (see site filmfan.ai) comes in.

What is Film Fan?

Film Fan uses AI to recommend films based on your preferences. The key word here is ‘fan’; the target audience being the avid film fan who truly desires to see only those films they were meant to see. This means one can’t settle for whatever your streaming services or cinemas offer you, you need to seek the best films; often referred to as your ‘life-altering films’. The premise here is simple: you tell the AI from the list of films, which films you prefer (whether you’ve seen the film or not). There’s currently 18k films (for all time) focused on the English language; some foreign films also exist where English is spoken.

Reviewing films

Here’s how it works, see example below (one of my personal favourite films):

In this case, I’ve watched it and believe that it is double thumbs-up (Awesome). But for you it’s possible you may give it a lower score (and we can’t be friends) or perhaps you haven’t even seen it. For the latter case, based on the artwork and details, would you watch it, if so, how likely is that? The more films you review, the more data the AI has to recommend films, and hence power to find you your life altering films. Simply click the button corresponding to your review, then the next film appears. You will notice newer and more highly ranked films will be served in the app first.

Inspiration and comparison to existing solutions

I’ve been a lover of films and data since I was a young child, so this project was a natural evolution for me after learning these skills over my career and starting Data Karate. AI film recommendations have been around for a while and exist in two different forms with various trade-offs:

1. Streaming services and film marketing
   1. Pros:
      1. Often best in class ML and engineering due to big budgets.
   2. Cons:
      1. Will only recommend films in the set of films they sell (think hundreds of films, not 10s of thousands).
2. Film or content recommendation specialists
   1. Pros
      1. Will recommend films across a wider base.
      2. Often have a research focus.
   2. Cons
      1. Data collection is just one of many areas and often not given enough attention in its own right.
      2. A common technique is to use similar users’ reviews for recommendations, but this is difficult given (i).

Film Fan is more like category 2, with a film universe of 18k films (and growing), however at its heart it’s all about data collection. The central idea which underpins the best AI is that all else being equal, the system with the best data has the best AI. AI models are commodities, however collecting and maintaining data takes ongoing effort from all angles. Don’t get me wrong about AI modelling though, this is an extremely important aspect, which is a specialty in its own right, but you can’t make a brilliant model effective with bad data. Your AI model needs to be state of the art, that’s just table-stakes, but your point of difference is the data you have.

Watched and unwatched films

As you review films, it’s handy to know what films you’ve watched or not yet watched, in these film lists below. At the time of writing, I have personally reviewed 12609 films in my Film Fan’s database, having watched 1354 and not watched 11255 as depicted below. (It doesn’t take as long as you’d think and it’s a fun game – mobile friendly too!). These lists are paginated and sorted by score to see the best films at the top.

Also, they are hyperlinked to the specific film pages (as shown above with “I Am Sam”) to view or change your review score, and see all of the details. Changing the review score is another data collection point because the AI will know both what you reviewed now and how you have reviewed this film in the past. New review scores trigger new film recommendations to be produced as can be seen on the next film list…

Recommended films

Once you’ve reviewed some films, you should expect to receive your first recommendations in 15 minutes on average. And it will be updated every half an hour for which you add more reviews. So keep checking for more and better recommendations as you give more reviews! Also keep checking as new films are added daily which you can review as perhaps they may be included in your set of recommendations.

Searching films

After displaying each of the: watched, unwatched and recommended films above, you can then search each list to find something you’re looking for. E.g. here are some of the “romance” films recommended for me:

Or perhaps you want to search all films, not just the current list. For example here’s some of the films with Tom Hanks:

Watch films

Now that you know your awesome films to watch, where do you watch them? Fortunately there are sites that have done this work for you so that you don’t need to manually check all your streaming services one by one. Simply click the film you want to see and it will open the film details page, from there click “Watch!” (where the back/forward buttons are when reviewing films) 

This will trigger a search to be automatically performed on JustWatch website, showing you how you can watch this film (it’s also region based for: au, us, nz, ca & uk). Please ensure to accept pop ups from filmfan.ai because a new tab is opened for convenience. Always check the “Free” tab because often many critically acclaimed or indie films can be watched for free on services like: Kanopy and Plex. Like this: 

Future

So there you have it, your personalised film recommendations await, let us know how we’re doing! To continue to enhance the experience here are some features to look forward to for the future:

1. Joint recommendations – what recommendations do you have in common with your friends and family? This should make movie night a breeze! (NB: this feature already exists as a hidden easter egg, reach out if you want to learn more).
2. Notifications to new recommendations – get notified when your new recommendations are ready for you
3. Enhanced AI from more data sources – lets add more data and enhance our ML models to improve your recommendations