Which genres dominates when movies blend styles?

I studied the subtitles of over 60,000 films and found that some genres almost always shape the tone, even in hybrids.

Last week’s look at which genres share the same fans led to yet more reader questions.

I love this as it gives me permission to go even deeper into the topic, and hopefully help everyone understand what ‘genre’ really means.

Some people asked about how subgenres work. It’s a good question because the answer is rarely clear from marketing or loglines. The way a film is described often suggests a neat hybrid, but the reality on screen can be quite different.

I analysed the dialogue from 65,000 movies that were listed on IMDb under more than one genre (i.e. Action and Adventure, or Horror and Comedy).

We have to start by getting a baseline for how movies of each major genre sound, and after that we can look at which has the greatest influence on what the characters are saying on-screen.

Dialogue styles differ between genres

Every genre has its own style of dialogue. Some are built around events and spectacle, others lean towards character and emotion. The language of a sci-fi film typically sounds nothing like that of a family comedy or a war drama.

To do this, I measured how similar the language of each film was to other films in its genre. By looking at the dialogue across thousands of movies, I could see clear patterns in how each genre tends to sound.

Using these results we can visualise how close the dialogue of different genres are from each other, on two separate axes.

Note: The text labels on the axes are my own, based on interpreting the data. The maths can tell you what’s significant, but not why or how to refer to it.

When we consider the distance between genres (i.e. how similar they are to one another) some patterns emerge:

• Genres like romance and musical lean towards personal, everyday conversations.

• Action, sci-fi and war pull the language towards event-driven, genre-specific territory.

• Comedy, drama and biography sit closer to the centre, often bridging the gap between styles.

To make the patterns even clearer, I have put another copy of the same chart below, overlaying some blobs to represent clusters of genres.

Here’s my thinking when determining what the clusters were:

• Tense. This cluster contains horror, thriller, crime and mystery. The language is event-driven, sharp and tightly focused on tension, danger or investigation. The dialogue tends to be functional, clipped and less personal.

• Spectacle. Here we find sci-fi, action, adventure and western. The focus is on big events, high stakes and genre-specific language and so the speech is shaped by plot mechanics and technical terms.

• Fantastical. Including animation, fantasy and family, this cluster uses genre-specific and imaginative language. Dialogue is shaped by world-building, magic and playful or didactic exchanges, often reaching beyond everyday speech.

• Emotional. This includes drama, romance, musical, music and comedy. The dialogue centres on personal relationships, emotion and everyday speech rather than genre tropes or technical terms. This group leans into character-driven storytelling.

• Truth. This perhaps the most contrived of the groups, as I am using my understanding of the plot settings, not just the semantic data (i.e. war, history, biography and sport). That’s why is covers a large area and has a stretched shape. I considered leaving these uncluttered but given that their topics do overlap, I felt it was worth putting them together. It is interesting to note that ‘truth life stories’ can have very different styles of dialogue, as illustrated by the distance between sports and war, a greater distance than that between sci-fi and musicals.

The genres that dominate when mixed

Now that we have our baseline for each major genre, we can see what happens when a film is a hybrid subgenre, such as ‘action comedy’, ‘sci-fi romance, or ‘sports western’ (ok that last one was a stretch, but they do exist, unsurprisingly around horses, such as Hidalgo and 8 Seconds).

So when a movie is two genres, which takes the lead?

Most two-genre movies lean heavily in one direction, with genres like sci-fi, horror, action and western often overpower their pairings.

This is perhaps why movies of this ilk are often literally referred to as ‘Genre Films’ by professionals in film sales. Also, the fact that in only one subgenre did drama dominate, ‘drama mystery’, adds further proof to those same folks oft-repeated mantra that “Drama is not a genre” (for the purposes of film sales).

Let’s get under the skin of a couple of them.

Sci-fi acts like a black hole in subgenres, pulling the film’s language and tone into its orbit. Whether paired with comedy, drama, fantasy or family, sci-fi usually takes over the dialogue style. The rhythm, word choice and focus become recognisably sci-fi, even if the other genre remains part of the plot.

The ‘sci-fi-ness’ is overpowering in all sci-fi subgenres, other than ‘sci-fi fantasy’. Unsurprisingly, this covers most of the Disney tentpoles from within the Star Wars and Marvel universes, which can’t financially afford to be seen as fringe genre films.

Within horror it’s a more complicated story. When mixed with crime or sci-fi elements, the horror influence is a lot weaker, but it rules when mixed with fantasy, mystery, comedy, romance and drama.

Subgenres in detail

Finally, for you completionists out there, here are all the subgenres for which I tracked at least 50 movies.

Notes

Today’s research started with around 65,000 subtitle files provided by OpenSubtitles.org and IMDb genres from the OMDb.

I sorted and cleaned the subtitle files, embedding them with a sentence-transformer model, producing a vector that reflects the overall style and content of the dialogue. I calculated the typical dialogue style by averaging all films in that category. For each two-genre film, I compared its dialogue to both genres’ averages to see which it was closer to.

Pairs where more than 55% of films leaned towards one genre were marked as dominated by that genre. Pairs within 5% of a split were listed as balanced. All dominance findings used only pairings with at least 10 films. To show subgenre splits across the board, I limited detailed visualisations to pairs with 50 or more movies.