Smalltown Boy AI Dance Trend: The Screaming Duo Explained

A person grabs their head and appears to scream into the camera. Behind them, a second performer is already moving to Bronski Beat. The roles switch, the drama gives way to dancing, and then you see the same routine performed by two cats or an unlikely pair of fictional characters.

That is the recent Smalltown Boy AI dance trend: AI remakes of a recognizable two-person screaming-and-dancing performance, including cat and fictional-character versions. This article is about those AI remakes—not the 2024 TikTok challenge where parents showed their 1980s dance moves. The song is old. The human choreography came before the AI versions. What AI changes is who gets cast in the scene.

If you have been searching for the “Smalltown Boy screaming dance,” the defining feature is the exchange between two performers: dramatic distress in front, contrasting movement behind, and a transition into dance.

What is the recent Smalltown Boy AI dance trend?

The Smalltown Boy AI trend uses generated or AI-modified video to place new subjects into the two-person screaming-and-dancing format associated with Bronski Beat’s “Smalltown Boy.” The subjects might be people, pets, illustrated characters, or deliberately absurd pairings.

The recognizable elements are the performance rather than a particular filter: one subject takes the foreground with an exaggerated emotional reaction, the other provides a contrasting movement or response, and the sequence changes into a shared dance. Some versions preserve the original staging closely. Others adapt the setting, character design, or transitions.

“Smalltown Boy AI” is a descriptive name for those remakes, not the name of one universal model or an official Bronski Beat product. A dedicated template, motion-transfer workflow, and custom-generated sequence can all produce videos that use the format. They do not necessarily use the same technology or include the same audio.

The simplest way to recognize it is to ask: Does the clip preserve the two-person dramatic handoff, or is it just a character dancing to the song? The latter may use the same soundtrack without reproducing this particular meme.

Where did the screaming dance come from?

A traceable human reference is the “Don’t cry. Dance!” performance posted by elvira.mua.music and mr_unvrs. In the clip, the performers use exaggerated expressions, foreground-and-background staging, and a transition into dance. That human performance is important context for the AI versions: the concept did not originate as a text prompt.

The contrast is immediate: an apparent emotional breakdown in front, movement behind, and then a release into the dance.

An AI fashion-character remake by thedungerzone explicitly credits the performers and describes using image generation and motion transfer. It is a useful example of what changes in the AI adaptation: the cast becomes fictional, while the underlying performance remains recognizable.

That gives us a documented connection between a human performance and an AI remake. It does not establish a single first upload across every platform, or tell us which repost made the format spread. Those are different claims, and a repost’s popularity is not proof that it invented the dance.

What song is used—and is this the 2024 TikTok trend?

“Smalltown Boy” was released as Bronski Beat’s debut single in May 1984. Jimmy Somerville’s striking falsetto and the song’s synth-pop arrangement make it instantly recognizable, but its meaning is more specific than “a nostalgic dance track.” It tells the story of a young gay man leaving a hostile hometown in search of acceptance.

The current AI dance format is not the 2024 parents’ challenge. As BBC Culture’s account of the song’s history and earlier TikTok revival explains, that earlier #80sdancechallenge invited parents to demonstrate their 1980s dance moves. The recent AI remakes instead recast the dramatic two-person performance. Same song, different format—and the 2024 coverage is background, not evidence for the newer AI trend.

The cultural context matters, too. You can enjoy the absurdity of the dance without reducing the song to a generic “screaming sound.” Its emotional force comes from an actual story of exclusion and escape.

Why the Smalltown Boy screaming dance works as an AI meme

The performance has a clear visual contrast. A close, distressed face demands attention, while the second subject creates a competing mood in the same scene. When the focus changes or the pair moves into dance, the viewer gets a quick payoff rather than an expression that simply repeats.

This structure gives AI remakes something concrete to preserve. A cat does not need a complicated backstory to make the scene funny. Once it takes the foreground role and performs the head-clutching reaction, the mismatch between subject and behavior is already the joke.

The two-person setup adds another layer. A single dancing character is a spectacle; a pair can suggest a relationship. Two pets become household rivals. Two original characters can play the serious one and the theatrical one. Familiar, consenting adults can turn an existing group-chat joke into a performance.

There is also a useful limit built into the format. The background, clothing, and cast can change, but if the remake loses the dramatic handoff, it loses much of what makes the routine distinctive. Better resolution cannot compensate for missing the joke’s timing.

That is why the most useful creative question is not “How realistic can I make this?” It is “Which parts of the performance need to survive for viewers to recognize it?”

How AI versions are made—and what a prompt cannot guarantee

There is no single production method behind every Smalltown Boy dance trend video. Three routes are worth distinguishing.

A dedicated two-subject template

A template packages the performance so the creator mainly chooses the cast. With Smalltown Boy AI, upload two separate portraits or one photo of both subjects to create a 21-second vertical scream-and-swap dance video. No prompt writing or video settings are needed.

The advantage is a narrower task: you are not asking the model to invent the whole sequence. The limitation is equally important. A preset decides much of the staging, and its result still needs to be checked for identity changes, awkward anatomy, and timing.

Do not assume that every tool using the trend’s name takes two separate photos or lets you choose which subject leads. Check the actual input controls and example output.

Motion transfer or a reference-guided performance

Motion transfer uses an existing performance to guide a new subject’s movement. The reference contributes posture, gestures, pacing, and sometimes expression; the target image contributes appearance.

For this meme, that can be more relevant than a long prompt because the order of the actions matters. “Look distressed, then dance” describes an idea. A performance reference can show where the subject stands, when a gesture begins, and how the transition happens.

Support varies by tool. A single-character motion workflow should not be assumed to handle two subjects reliably, and a preset-based tool should not be assumed to accept an arbitrary uploaded video. Use reference footage you have permission to use; a public post is not automatically a reusable production asset.

Free-form image-to-video generation

A general image-to-video model can animate a still image from text directions. This is useful for creating an original scene inspired by the contrast between distress and dance, but it is not a guarantee of the exact routine.

If you choose this route, describe visible actions rather than relying only on the trend’s name: which subject is closer to the camera, what each subject does, how the focus changes, and how the dance begins. Keep camera movement simple while you test the interaction.

Expect differences between attempts. Two stable identities, hands near faces, overlapping bodies, and timed choreography are several demands at once. A convincing opening frame does not prove that the entire sequence works.

How to make a version that keeps the joke intact

Start by watching the human performance rather than only an AI repost. Notice the relationship between the close subject and the person behind them, then decide whether you want a recognizable remake or an original variation.

Choose a pairing with a clear premise. Two of your pets, two characters you created, or two consenting adult friends give the video a reason to exist beyond testing an effect. One short caption should be enough to explain the relationship.

Match your images to the workflow. A two-portrait preset needs distinct, readable identity references. A full-body dance workflow needs visible limbs and room for movement. Avoid asking a tightly cropped headshot to supply an entire convincing body without expecting the model to invent those missing details.

Preserve the action sequence before adding decoration. Check the foreground reaction, the other subject’s response, and the change into dance. If the subjects merely sway together, the result may be pleasant but it is no longer doing the same joke. Extra props and dramatic camera moves can wait.

Review the whole clip, not just the first second. Pay particular attention when hands cross the face, one subject passes in front of the other, or the focus changes. Those are the moments where a face can drift, a limb can merge, or the two identities can become confused. Change one input at a time so you can tell what improved the result.

The soundtrack is a separate part of the workflow

A clip can resemble the dance visually and still fail to land if the sound is wrong. The apparent screaming is part of the performance; it does not mean the subjects are genuinely distressed, or that every version contains a separately recorded scream.

Check whether your generator exports silence, generated audio, or a soundtrack. Do not assume that typing the song title produces the original Bronski Beat recording, or gives you permission to use it.

When adding music, align the dramatic gesture and the release into dance with the audio excerpt you actually choose. Different edits can start at different points, so a fixed timestamp copied from someone else’s tutorial may not fit your version.

Use music available for your account and intended use, or a recording you are licensed to use. Availability in one platform’s music library does not automatically establish permission for an advertisement or a repost elsewhere.

Label realistic AI remakes and respect the people involved

The lighthearted premise does not remove the need for consent. Use your own likeness or images of adults who agreed to the remake, and do not present an invented performance as footage of something they actually did.

TikTok’s AI-generated content guidance explicitly gives portraying someone dancing when they did not dance as an example of significant AI editing. It requires labels for realistic AI-generated images, audio, and video. Its guidance also prohibits certain likeness uses, including adult private figures without permission and the likeness of people under 18.

An AI label explains how the remake was made. Crediting the human choreography explains where the idea came from. Neither substitutes for permission where permission is needed.

Smalltown Boy AI trend FAQ

Is the original Smalltown Boy screaming dance AI-generated?

There is a traceable human performance by elvira.mua.music and mr_unvrs captioned “Don’t cry. Dance!” AI remakes adapt that performance with different subjects. Do not assume every video using the sound is synthetic.

Why are there so many cat versions?

Cats make the emotional overreaction especially absurd, while the two-subject staging gives each animal a clear role. The joke remains readable even when the performers change species.

Is this the same as the 80s parents’ dance challenge?

No. The 2024 challenge asked parents to demonstrate their 1980s dance moves. The screaming-duo format uses a staged emotional contrast and a transition into dance; AI versions replace or reimagine the performers.

Can I recreate it with only a text prompt?

You can generate a dance-inspired scene, but a prompt alone does not guarantee the exact choreography, two consistent identities, or synchronization to the original recording. A compatible preset or permission-cleared performance reference gives the movement more specific guidance.

The Smalltown Boy AI trend is easiest to understand as a human performance with a changeable cast. Keep the dramatic exchange readable, acknowledge the original performers, and check the movement and sound separately. The surprise comes from who appears in the scene—not from how many effects you add to it.