AI Is Writing Musical Theatre Songs Now — And Some Aren't Bad - Success Academy Charter Schools

A song created entirely by artificial intelligence recently crossed something most AI- generated music never manages: it went genuinely viral, not as a curiosity, but because people liked it. Multiple listeners in the comments compared it to the Beatles. The theatre community should be paying attention — because the same tools are already knocking on the stage door.

The track was produced on Deep Dream Generator, a platform best known for AI image generation that has been quietly building music tools alongside its visual capabilities. No studio. No composer. No lyricist. The creator typed a description and the AI assembled the rest — melody, harmonies, lyrics, production, and a music video featuring iconic internet memes dancing to the finished track.

Watch: The Deep Dream Song — viral AI-generated music video

The viral track — millions of views across YouTube and TikTok, with comment sections full of Beatles comparisons.

For those of us whose musical frame of reference is Sondheim and Hammerstein rather than SoundCloud and TikTok, the track is worth watching before reading further. Because the question it raises is not abstract. It is immediate: if AI can produce a song that millions of people find genuinely moving, what does it do when pointed at the musical theatre form specifically?

THE QUALITY QUESTION, HONESTLY ADDRESSED

The history of technology meeting musical theatre is mostly a history of theatre absorbing new toolsand remaining fundamentally unchanged. Amplification arrived. Synthesizers arrived. Digital orchestration arrived. The form survived all of them, largely because the thing audiences come for — a story told through song by human beings in real time — is not something any technology has yet replaced.

AI music tools are different in degree, if not yet in kind. What makes the viral track striking is not that AI made music. It is that AI made music that listeners evaluated as music rather than as a demonstration. The Beatles comparisons are not technically accurate — no AI has Lennon and McCartney's melodic genius. But the fact that the reference point was the Beatles, rather than "a competent algorithm," signals that the output clears the threshold of genuine emotional response.

That threshold matters enormously for musical theatre. A Broadway song is not decoration. It exists to carry emotional weight that dialogue cannot, to crystallize a character's interior life at the moment words stop working. Whether AI can produce that kind of song — not just a catchy tune, but a dramatically purposeful one — is a genuinely open question in 2026.

The honest answer, based on what the current tools produce, is: not reliably. Not yet. AI can generate songs that are melodically appealing, harmonically coherent, and lyrically functional. What it struggles to do is generate a song that feels specifically inevitable — the way "Being Alive" or "Finishing the Hat" feels like it could only have been written by someone who understood exactly what that character needed to say at that moment.

But "not yet" is doing a lot of work in that sentence. The trajectory is clear.

WHERE THEATRE CREATORS ARE ALREADY USING IT

Set aside the philosophical question for a moment and look at the practical one: where are AI music tools actually showing up in musical theatre workflows right now?

The most significant adoption is happening in demo production. For decades, the economics of pitching a new musical have been brutal. A composer and lyricist finish a score, but to present it to producers, directors, or developmental programmes, they need a demo recording — ideally with multiple voices, piano at minimum, full orchestration at best. Professional demo recordings run from several hundred to several thousand pounds per song. For a full show, the cost of demoing the score can dwarf the cost of writing it.

AI music tools compress those costs dramatically. Writers can generate a rough vocal demo of a new song in minutes, with AI-generated voices carrying the melody, to use as a working reference or initial pitch material. The quality is not what a professional demo achieves. But it is sufficient to communicate the song's architecture — whether the melody works, whether the lyric scans, whether the dramatic moment lands.

A note on the tools. Platforms like Deep Dream Generator, Suno, and Udio each approach AI music generation differently, but all share the same basic proposition: ndescribe what you want, receive a fully produced track. Deep Dream Generator's particular utility for theatre creators is that it combines music generation with image and video

generation on one platform — useful for creating rough audiovisual pitches or mood reels for new shows without coordinating multiple services. The viral track that prompted this piece was produced there entirely.

Beyond demos, AI tools are appearing in the earlier stages of the compositional process itself — not replacing the composer, but functioning as a generative partner. Some writers are using AI to produce melodic sketches they then develop by hand, the same way a playwright might use writing prompts to escape a block. Others are using AI-generated tracks as reference points to communicate a tonal direction to collaborators before any original material exists.

THE JUKEBOX MUSICAL PROBLEM — AND POSSIBILITY

There is one corner of the musical theatre world where AI music tools create a more immediate and uncomfortable disruption: the jukebox musical.

The jukebox musical's premise is that existing songs, beloved and familiar, can carry a narrative if arranged correctly. Mamma Mia!, & Juliet, MJ — the form has proved commercially durable precisely because the songs are already known quantities. Audiences arrive with emotional attachments pre-installed.

AI music tools can now generate songs that approximate the style of existing artists with uncomfortable accuracy. That raises a question the form has not previously had to answer: if AI can produce a convincing pastiche of an artist's catalogue, what exactly is the distinction between a jukebox musical using licensed originals and a production built around AI generated works in the same style? The legal answer involves copyright law that is still being written. The artistic answer is more contested.

The possibility side of that coin: AI could enable jukebox-style musicals built around artists, genres, or eras where licensing costs make traditional approaches prohibitive. The economics of securing rights to a complete catalogue are significant. AI-generated music in a licensed style — if the legal frameworks ever catch up — could make certain productions possible that are currently not.

WHAT COMPOSERS AND LYRICISTS ARE ACTUALLY SAYING

The official positions from professional organisations are clear. The Dramatists Guild, Equity, and similar bodies have staked out firm protective stances on AI-generated material in professional contexts. Their members' work should not be replaced by algorithmic output without consent or compensation — a principle that is hard to argue with.

The private conversations are more complicated. Working theatre composers occupy a wide spectrum of positions, from categorical rejection to cautious curiosity to active experimentation. The younger cohort of writers — those who grew up using digital audio workstations as naturally as earlier generations used pianos — tend to approach AI tools as another instrument in the studio, neither threatening nor salvific, just present.

What cuts across positions is a shared recognition of where AI remains genuinely weak in the theatrical context. A song that works in a pop or viral context — catchy, emotionally general, built for mass relatability — is a different object from a song that works in a book musical. Theatre songs carry specific dramatic function. They belong to a character in a story. AI tools, at their current state, generate songs for everyone and no one simultaneously. Giving a song to a particular character, in a particular scene, in service of a particular story — that remains stubbornly human work.

THE DEMO REVOLUTION IS ALREADY HERE

Regardless of where one stands on AI's long-term role in musical theatre creation, one development seems difficult to reverse: the economics of demo production have changed permanently.

The writer working alone, without the financial resources for professional recording, now has access to tools that can give their material a listenable form that others can evaluate. That democratisation has costs — a flood of AI-assisted pitches will make the signal-to-noise problem in new musical development worse before it gets better. But it also removes a financial barrier that has historically functioned as a gatekeeping mechanism with no artistic justification.

The viral AI track that prompted this piece is a useful benchmark for where the tools stand. It is genuinely good. It would not pass for a Sondheim lyric or a Rodgers melody under scrutiny. But it is good enough that millions of people shared it without stopping to question whether a machine made it. For a demo, that level of quality is more than sufficient.

Theatre has absorbed every previous wave of technological change without losing what makes it irreplaceable. The argument that it will do so again with AI is plausible and probably correct. But the form has not previously encountered a technology that attempts, however imperfectly, to do the specific thing theatre does — tell stories through song. That novelty deserves honest engagement rather than reflex dismissal.

The question is not whether AI will replace the musical. It will not. The question is what musical theatre looks like when every writer, however early in their career and however limited their resources, can hear what their songs sound like before they can afford to have them properly recorded. That future is already beginning.