Inspiration

When I first understood what Qloo does, one thought hit me immediately: culture is built mostly from movies and the people in them. Actors inspire how we dress. Films shape the music we love, what we want to watch, and what makes us feel we belong. Taste in one thing says something about taste in everything else, and Qloo is built to understand exactly that.

Two ideas followed that I could not stop thinking about.

For fans: if Qloo knows the cultural links between the films, artists and brands you love, it can recommend your next favourite film in a way that feels personal, not generic.

For studios: many great films fail to stand out in other countries, and the problem is rarely the story. It is how the film is introduced. What excites an audience in Tokyo is not what excites one in Dubai or London. So I asked: what if the story stays exactly the same, and only the way it is presented to each culture changes? That became PLOT TWIST: where culture changes the scene, not the story.

I love movies, and I love the idea that the right film can reach the right people anywhere in the world. That is why I built this.

What it does

PLOT TWIST has two sides.

For fans: find your next obsession

  • Pick up to five films, artists or brands you love.
  • Qloo returns what is culturally connected to them, with a source label, a timestamp and an affinity score.
  • An animated culture map shows your picks and what Qloo returned for them together.

For studios

  • Enter a film brief: logline, genre, what must stay unchanged, cast, and the story moments you are happy to show in a trailer.
  • Choose up to three markets, for example Tokyo, Dubai and London.
  • For each market, PLOT TWIST retrieves real Qloo evidence across films, music, brands, TV shows and places.
  • A cross-market comparison, computed directly from the Qloo results, shows what is shared across markets and what is unique to each.
  • For each market, the AI writes a separate campaign direction: tagline, music and styling, headlines, and a trailer plan built only from the story moments the studio provided. It recommends whether to lead with theatres or OTT in that culture.
  • A cast promotion plan uses only the actors the studio entered and suggests where, how and when to promote in each market.
  • The story, characters and ending never change, and spoilers stay protected.

How we built it

  • Backend: Python and FastAPI. All API keys stay on the server and never reach the browser.
  • Frontend: a single-page app in plain HTML, CSS and JavaScript with a light, animated design.
  • Cultural intelligence: the Qloo Hackathon API.
  • AI: an open model (gpt-oss-120b, served by Groq) that only interprets the evidence Qloo returned.
  • Deployment: Docker, so it can run on a free host.

How Qloo is used. Qloo is the core of the product.

  • /search resolves what the user types into real Qloo entities.
  • /v2/insights with taste signals (the fan's picks) powers personalised recommendations.
  • /v2/insights with location signals powers the studio's per-market evidence across films, artists, brands, TV shows and places.
  • The cross-market comparison is computed directly from those results.

Honesty by design.

  • If Qloo or the AI fails, the app says so. It never shows fake data.
  • Every result is labelled as either "retrieved from Qloo" or "AI-generated hypothesis".
  • The AI is told to cite only names that Qloo actually returned, and the server removes any citation that is not real evidence.
  • The film brief is treated as untrusted text, never as instructions.
  • PLOT TWIST does not claim to predict box office. Affinity is a signal, not a guarantee, and cast and venue ideas are leads to investigate, not confirmed partnerships.

Challenges I ran into

  • Keeping the AI creative without letting it invent evidence or change the story. I solved this with strict output rules and server-side filtering of every citation.
  • Building on free tiers. Model names, rate limits and image-generation access kept changing while I built, so I redesigned the AI step to run on free services.
  • Poster tailoring. I wanted to adapt a film's real poster to each culture while keeping the actors and title untouched. Free image-editing services were not available, and I did not want to ship anything that could alter an actor's face, so I chose not to include it. It is on the roadmap.
  • Keeping a feature-rich idea simple. I focused on two clear modes with one main action each.

Accomplishments that I am proud of

  • The whole flow works end to end on real Qloo data, from a film brief to market-specific campaign, trailer and cast plans.
  • The cross-market comparison shows how different the results are for each city, which proves the core idea: one story needs different introductions in different cultures.
  • Trust is built in: nothing is faked, everything is labelled, and the AI cannot cite evidence that Qloo did not return.
  • It runs entirely on free services.

What i learned

Cultural data is most powerful when it is used to adapt how a story is presented, not to change the story itself. I also learned how much trust depends on being open about what is real data and what is AI interpretation.

What's next for Plot Twist

  • Accounts and a private Film Vault for studio projects.
  • Market-tailored poster artwork that keeps the actors and film title unchanged.
  • PDF campaign reports.
  • Privacy-protected audience insights from fans who opt in.
  • More Qloo signal types, such as demographics, for sharper audience targeting.

Built With

Share this project:

Updates

Submission history