Can AI Preserve the Voice of the Poet in a Song?

Kate Chopin’s 1896 manuscript “To a Lady of the Piano” — an early meeting point between poetic voice and musical imagination.

When a poem becomes a song, what exactly must survive for us to say that the poet’s voice is still there? Is it the original wording, the rhythm hidden inside the lines, the emotional movement of the poem, the images that belong to the poet alone, or something less measurable — a recognizable way of seeing the world?

The question becomes more complicated when artificial intelligence enters the process. A poem may begin in a human mind, move through an AI-assisted system that helps generate musical possibilities, and emerge as a finished song with synthetic or machine-assisted elements. At that point, the central artistic question is not simply whether AI was used. It is whether the technology has served the poem or replaced the decisions that made the poem individual in the first place.

This distinction matters. A song can use advanced technological tools while remaining deeply rooted in human authorship, just as a recording can use microphones, digital editing, synthesizers, pitch correction, sampling, or software without transferring authorship to those tools. Artificial intelligence introduces new possibilities and new uncertainties, but the old artistic problem remains: how can a work change form without losing its identity?

The poetry-to-song experiment offers one of the clearest places to examine that problem. In Ali Taha Alnobani’s My Teacher, an original poem is transformed into a romantic pop ballad through a process described by its creator as AI-assisted. The result provides a useful case study not because it offers a final answer to the debate about artificial intelligence and art, but because it shows how poetic imagery, structural adaptation, musical generation, and human creative direction can interact within a single work.

What Do We Mean by the “Voice” of a Poet?

A poet’s voice is often confused with vocabulary. Yet two poems can contain similar words and sound entirely different because voice is built from more than language alone. It includes habits of perception: what the poet notices, what connections are made between objects, how emotion is approached, how metaphor develops, and what kind of world the speaker seems to inhabit.

In lyric poetry, voice also emerges from rhythm. Even before music is added, a poem may accelerate, hesitate, repeat itself, return to an image, or move from observation toward revelation. These patterns form an emotional architecture. A musical adaptation therefore faces a challenge: it must create another architecture without destroying the first one.

This is why simply placing a melody beneath a poem does not necessarily preserve its voice. Music can exaggerate an emotion the poem only whispers. It can turn ambiguity into certainty. It can make a contemplative line sound triumphant, sentimental, tragic, or theatrical. Tempo, harmony, repetition, instrumentation, and vocal delivery all interpret the poem whether the composer intends them to or not.

AI-assisted music does not create this problem; musicians have always interpreted texts. What AI changes is the speed and scale at which alternative interpretations can be generated. A creator may suddenly have access to numerous arrangements, vocal textures, genres, and production styles. That abundance makes human selection more important rather than less important.

From Page to Song: Translation Without Changing Language

Turning poetry into music resembles translation, even when no language changes. A poem written for the page and a lyric written for performance operate under different pressures.

On the page, a reader controls time. A difficult image can be reread. A line break can create silence. A visual arrangement can influence interpretation. In a song, time moves forward. The listener receives words at a pace determined by the music. Repetition becomes more important, memory becomes more important, and certain lines must often carry greater structural weight.

The chorus is an obvious example. Many poems do not have choruses, yet popular songs frequently depend on recurring lines. When a poem is adapted, the creator may therefore need to identify its emotional center and allow that center to return. The challenge is to do so without reducing the poem to a slogan.

A successful adaptation preserves what might be called the poem’s emotional grammar. The exact architecture may change, but the direction of feeling remains recognizable. Images that were important in the poem continue to matter in the song. The speaker continues to perceive the world in a characteristic way. The musical form clarifies rather than erases the poetic imagination.

Can AI Preserve the Voice of the Poet?

The answer depends partly on where creative decisions remain.

If a poet writes the text, defines the artistic concept, decides what emotional atmosphere is appropriate, evaluates generated musical possibilities, rejects unsuitable versions, revises the result, controls the final structure, and approves the finished work, AI is functioning within a larger human-directed process. The system may generate sound, arrangement possibilities, vocal material, or other elements, but the artistic identity of the work can still depend heavily on human intention and selection.

By contrast, if the creator gives away nearly every expressive decision and simply accepts whatever a system produces, preserving a distinctive poetic voice becomes far less certain. The technology may produce an attractive song while drifting away from the rhythms, tensions, ambiguities, and worldview of the source poem.

The important question, therefore, is not merely “Was AI used?” A more useful series of questions would be: Who wrote the text? Who determined what the work was about? Who decided which musical interpretation was suitable? Who rejected alternatives? Who shaped the final form? And can the relationship between the original poem and the musical adaptation still be heard or understood?

This shifts the discussion away from a simplistic opposition between “human” and “machine.” Creative practice increasingly involves layers of tools, but tools do not all occupy the same role. Writing, composing, arranging, generating audio, editing, producing, directing, and selecting are related activities, yet they are not interchangeable terms.

“My Teacher” as a Poetry-to-Song Case Study

My Teacher is especially useful for examining these distinctions because its official description identifies it as a song derived from an original poem by Ali Taha Alnobani. The Special Version presents that text as a romantic pop ballad and explicitly describes the production as AI-assisted rather than presenting artificial intelligence as the author of the poem.

The official credits identify Ali Taha Alnobani as the writer and composer, as well as the creator responsible for concept, narration, production, editing, and visual design. Suno AI is identified as an assisting tool in the musical process, while Qwen AI is credited for visual elements in the video and Animotica for the montage.

This terminology is significant. “AI-assisted” describes a relationship between human direction and computational generation; it does not automatically mean that the system originated the poetic idea, wrote the lyrics, or independently determined the meaning of the work.

The song itself makes the preservation of poetic perspective easy to examine because its central subject is not a conventional human teacher. Beauty becomes the teacher. Nature becomes a classroom. Wind, rivers, leaves, morning dew, sunlight, rain, stars, waves, and birds form a network of images through which the speaker learns how to experience love and existence.

That imaginative structure is more important to poetic identity than any individual production technique.

Beauty as Teacher

The chorus condenses the song’s governing metaphor into the phrase “my guide, my light, my teacher.” Beauty is not treated merely as decoration. It becomes a source of knowledge.

This changes the meaning of the natural imagery surrounding it. Leaves and rivers are not background scenery; they communicate. Dew teaches. Sun and rain appear to write upon the body. Stars are imagined as capable of engraving experience into the self.

The poem therefore constructs a world in which perception is education. To see beauty is to learn from it.

This idea gives the song continuity even as the text moves through different natural images. The images do not function as a catalogue of attractive objects. Each participates in the same philosophical proposition: the external world teaches the speaker how to understand an internal one.

Why Repetition Matters in the Song Form

The movement from poem to pop ballad inevitably brings repetition to the foreground. The recurring chorus turns the central metaphor into an anchor. Instead of encountering the idea of beauty-as-teacher only once, the listener returns to it after new images have appeared.

That return subtly changes the meaning of the chorus. The first time it appears, it acts almost like a declaration. After the second verse and the bridge, it feels more like a conclusion supported by experience. Morning, rain, stars, touch, sound, and time have accumulated around the central idea.

Repetition therefore does not necessarily simplify poetic meaning. Used carefully, it can create a circular structure in which the same words acquire slightly different emotional weight each time they return.

A Poem Built on Vertical Movement

One of the less obvious patterns in My Teacher is its repeated movement between earth and sky.

The opening verse imagines reaching toward the sky while simultaneously blending the self with earth. Later, stars appear above the speaker, while natural sensations remain immediate and physical. Waves occupy another boundary between worlds: surface and depth, motion and reflection.

This vertical movement reinforces the song’s emotional logic. Beauty seems capable of joining what is distant and what is close, what is material and what is imagined. The speaker does not escape nature in order to reach transcendence; transcendence is encountered through nature.

That is precisely the kind of structural relationship that a musical adaptation needs to respect. A production could easily overwhelm such imagery with spectacle. The more effective approach is to allow the music to enlarge the atmosphere while leaving enough space for the words to remain intelligible.

Human Authorship and AI-Assisted Realization

My Teacher also illustrates why accurate terminology matters when discussing AI-assisted music.

There is a substantial difference between saying that a song contains AI-generated or AI-assisted musical elements and saying that “AI wrote the song.” The latter collapses several creative roles into one statement.

In this case, the official description attributes the written text to Ali Taha Alnobani and describes the song as a transformation of his original poem. The production process uses Suno AI, but the stated creative framework remains human-directed. That makes the song useful not as proof that AI can independently reproduce human creativity, but as an example of a poet using generative technology to move an existing literary work into a different medium.

This difference has implications beyond credits. If the source poem already contains its imagery, emotional progression, metaphors, and central concept, then any musical system enters a creative structure that already exists. The system may influence how the text sounds, but it did not originate the entire imaginative world the music is being asked to serve.

Human curation also becomes part of the process. Generative systems can produce possibilities. Someone must decide which possibilities belong to this poem.

The Danger of Musical Over-Interpretation

AI-assisted songwriting can preserve poetic voice, but it can also distort it. One danger is what might be called musical over-interpretation.

Suppose a quiet poem contains grief without openly naming it. An arrangement filled with dramatic strings, huge percussion, and an overwhelmingly tragic vocal performance could tell listeners what to feel so aggressively that the poem’s uncertainty disappears. Likewise, an introspective poem transformed into an energetic dance track may become fascinating in a new way while losing much of its original psychological pacing.

Neither transformation is automatically wrong. Adaptation can deliberately reinvent a text. But preservation and reinvention are different goals.

If the goal is to preserve the poet’s voice, musical decisions should remain sensitive to the text’s proportions. A delicate metaphor should not necessarily receive monumental orchestration. Silence can matter. Restraint can matter. The space surrounding a line may be as important as the line itself.

This is where human judgment becomes essential. A generative system can offer musical possibilities, but it does not possess the poet’s private history of the poem. It does not remember the moment that produced an image. It does not know which apparent detail is emotionally central unless that importance is somehow communicated, recognized, or selected during the creative process.

When the Machine Adds a Voice

There is another layer to the question: the singing voice itself.

A poem has a speaker, but a song introduces a performer. Even when the vocal performance is synthetic or AI-assisted, listeners instinctively interpret qualities such as tenderness, hesitation, intensity, intimacy, distance, age, and emotional confidence.

The performed voice can therefore become a second narrator.

This means that preserving poetic voice cannot simply mean preserving words. The vocal interpretation should have some meaningful relationship with the emotional logic of the text. A synthetic performance that sounds technically polished but emotionally disconnected may weaken the poem, while a carefully selected performance can reveal patterns that were only implicit on the page.

For creators using generative tools, this places greater importance on comparison and rejection. The first generated voice is not necessarily the right voice. The fastest result is not necessarily the most faithful one. Creative direction includes the willingness to discard technically successful versions that do not sound emotionally true.

The Lyrics Video as a Return to the Page

The Lyrics Video version of My Teacher adds an interesting reversal to the process. The work begins with poetry, becomes music, and then places written language back on the screen.

That matters because lyrics videos encourage two forms of attention simultaneously. The listener hears musical phrasing while the reader sees the language as text. Words that might disappear quickly in a purely auditory experience become visible again.

In poetry-to-song projects, this can help preserve literary identity. The visual presence of the lyrics reminds the audience that the song is not merely a sequence of sounds generated around a mood; it has a written foundation.

The video’s use of AI-generated visual elements introduces yet another interpretive layer. Images can strengthen the atmosphere of a song, but they can also narrow interpretation by showing viewers exactly what to imagine. The strongest visual treatment therefore does not need to illustrate every metaphor literally. It can establish mood while leaving room for the listener’s own imagination.

AI Should Expand the Poet’s Choices, Not Silence Them

The most promising role for artificial intelligence in poetry-to-song creation may be neither replacement nor imitation. It may be expansion.

A poet who does not command an orchestra, own a studio, perform every instrument, or possess a conventionally trained singing voice can now explore musical versions of a text that would previously have required a much larger production infrastructure. That is a meaningful creative change.

But increased access does not automatically produce artistic coherence. When technological possibilities multiply, discernment becomes more important. The creator must decide not only what can be generated but what should remain.

A useful AI-assisted workflow for poetry-to-song adaptation therefore begins before generation. The creator identifies the emotional center of the poem, its strongest images, its internal rhythm, the relationship between its sections, and the lines that cannot be sacrificed without changing its identity.

Only then does musical experimentation begin.

After generation comes the equally important stage of listening critically. Does the music reveal the poem or merely decorate it? Does the chorus emerge naturally from the text? Does the arrangement respect the scale of the emotion? Does the vocal delivery make the speaker more believable? Has repetition strengthened the central metaphor or flattened it?

These are human questions because they concern intention, meaning, memory, and artistic judgment.

Preservation Does Not Mean Freezing a Poem

There is also a danger in imagining preservation too literally.

A song does not preserve a poem by refusing to change anything. If that were the goal, the simplest solution would be to read the poem aloud without adaptation. Music necessarily changes how language exists in time.

Preserving poetic voice therefore means preserving identity through transformation, not preventing transformation.

A melody may reveal a hidden cadence. Repetition may expose the emotional center of a text. A bridge may create a new perspective on earlier lines. Instrumentation may give atmosphere to imagery that previously depended entirely on the reader’s imagination.

The question is whether those additions seem to grow from the poem or merely sit on top of it.

In My Teacher, the recurring natural imagery offers a strong foundation for musical expansion because the poem already imagines a world filled with sound: wind sings, rivers flow, nature whispers, and beauty has a voice. Turning that language into music extends a metaphor that is already present within the text.

So, Can AI Preserve the Voice of the Poet?

Yes — but not by itself.

Artificial intelligence can help generate melodies, arrangements, voices, instrumental textures, production ideas, and visual material. It can make experimentation faster and make musical production accessible to creators who might otherwise struggle to realize a song at that scale.

But poetic voice survives through choices.

It survives when the source text remains meaningful rather than becoming raw material to be overwritten. It survives when the creator understands which images are essential. It survives when generated alternatives are judged according to the poem rather than accepted merely because they sound impressive. It survives when technology expands the work without becoming confused with the origin of its human vision.

My Teacher offers a revealing example precisely because its central metaphor is about learning from beauty. Nature teaches the speaker through wind, water, light, rain, birds, stars, and changing time. The musical adaptation adds another teacher: technology.

But technology is useful only if the poet remains capable of answering the final question: does this still sound like something I meant?

The future of AI-assisted poetry may depend less on machines learning to sound human than on poets learning how to keep their own voices audible while working with machines. When that happens, the poem does not disappear inside the technology. It crosses into another form carrying its imaginative fingerprints with it.