A character in an Indian film says: "Yaar, this is not done." Three words of English in a Hindi sentence. A grammar checker flags it as inconsistent. A naive language model suggests "correction" to either full Hindi or full English. Both responses are wrong, because the sentence is not an error. It is a choice that carries specific social meaning that would evaporate in either corrected version.
Hinglish dialogue, which is the dominant mode in most mainstream Hindi OTT content today, is not a failure of code discipline. It is a register system, and screenwriters who work in it are making choices at a level of granularity that is invisible to tools trained on monolingual data.
Register, Not Random Switching
The term code-switching describes the alternation between languages within a conversation or sentence. But in Indian urban and semi-urban social contexts, the switching between Hindi and English is not random. It tracks class position, aspiration, irony, generational identity, and the social distance between characters in a scene.
Consider a common pattern in contemporary Hindi OTT drama: a character from a small-town background who has moved to a metropolitan city for work. In scenes with family members who have remained home, the character's dialogue will typically carry more Hindi. In scenes with urban professional peers, the ratio shifts toward English. The ratio itself is character data. A writer who understands this uses the mixing deliberately to show the character's code navigation across social contexts. A writer who does not understand this produces dialogue that sounds flat because the mixing looks careless.
The English insertions in Hindi dialogue also carry specific functional loads. English technical vocabulary appears in professional settings without irony. English emotional vocabulary ("upset," "sorted," "chilled") appears in informal peer contexts with a generational warmth. Formal or elevated English ("I would prefer," "this matter requires attention") appears in contexts of social performance, often with a slight ironic distance. These patterns are consistent enough to constitute conventions, not choices made fresh every time.
What Dialogue AI Tools Get Wrong
Most dialogue analysis tools available to screenwriters were trained on English-language corpora and apply analysis frameworks derived from English-language dialogue conventions. When they encounter Hinglish text, they typically do one of two things: treat the Hindi portions as noise (since they are unparseable by the model) or flag the mixing itself as a consistency problem.
Neither response reflects what the writer is actually doing. A dialogue tool that cannot recognize that "yaar" is a term of address signaling peer intimacy, that "bhai" in a formal context is either a deliberate downward social gesture or an ironic distance marker, or that switching to formal Hindi in an English-dominant conversation signals emotional retreat rather than grammatical inconsistency, is not reading dialogue. It is reading symbols.
This is not a critique of those tools' general capabilities. It is a recognition that they were built for different material. The problem arises when Indian writers use them and receive feedback that treats their choices as errors, which either leads the writer to second-guess valid choices or leads them to ignore the feedback entirely, in which case the tool has no value.
Aspiration, Irony, and the Voice That Does Not Match the Face
One of the more subtle uses of English in Hindi dialogue is the aspirational register: characters who use English in situations where it is not strictly necessary, as a signal of social positioning or desired identity. This is not mockery; it is a real social behavior that Indian writers observe and render with great accuracy. A character who has recently received a promotion and begins inserting business English into casual conversations with family members is showing you something about their psychology and about the social dynamics of the household.
Ironic English operates differently. Characters in contemporary Hindi drama sometimes deploy English phrases precisely because the formality of the English carries a deadpan irony in context: the character is playing at sophistication to comic or dramatic effect. The register is the joke. A grammar model that normalizes the sentence removes the register and removes the point.
There is also the question of dialogue that deliberately mismatches register and character. A young woman from a traditional household who code-switches fluently into corporate English when speaking with her employers, then drops all of it when she returns home, is showing you her double life more efficiently than any expository dialogue could. The switch is the story.
How We Think About Dialogue Analysis at Mugafi
The dialogue feedback in Mugafi is designed with these register functions in mind. We do not flag Hindi-English mixing as an inconsistency to be corrected. We try to identify whether the mixing within a scene is consistent with the character's established voice and the social dynamics of the scene, and whether shifts in the mixing ratio across scenes are intentional or are drifting without apparent purpose.
We built the dialogue analysis this way because the most common dialogue feedback problem we observed with our early-access cohort was not "this dialogue sounds wrong." It was "this character sounds different in this scene and I cannot figure out why." The why, in a significant proportion of cases, was a shift in the Hindi-English ratio that the writer had not intended or had not noticed. The character was moving in or out of a social context and their code was shifting, but the shift was not calibrated to what the scene required.
This is the kind of pattern that is genuinely difficult to catch without someone reading the entire screenplay with the character's voice profile in mind. It is the kind of feedback that a good script editor will catch on a second pass, once they know the character well enough to notice the drift. We are trying to surface it earlier, before the second pass, when the writer still has the whole draft in their head and revision is less costly.
The Limits of What We Can Analyze
We are not yet able to fully analyze the prosodic and rhythmic qualities of Hinglish dialogue, the way that the cadence of a sentence changes when Hindi and English elements combine in specific syntactic patterns. Some of the most distinctive dialogue writing in current Indian OTT content works at exactly this level, and it requires a depth of linguistic analysis that is beyond what we currently do.
We also recognize that our training data has better coverage of urban Hindi OTT dialogue conventions than of regional language dialogue conventions, even in scripts that are primarily in Hindi but set in specific regional contexts. The dialogue of a character from rural Bihar sounds different from the dialogue of a character from South Delhi, and those differences matter dramatically. We are working on this, but we want to be honest that our current coverage is uneven.
What we can say is that a tool designed for Indian screenwriters should start from the position that Hinglish dialogue is not a compromise between two pure forms. It is a distinct medium with its own rules, conventions, and expressive range. The writers using it know those rules better than any grammar system does. The tool's job is to help them apply those rules more consciously, not to correct them toward a purity the dialogue was never trying to achieve.