New Naqqaliشاهنامه
About the Poem

Translators’ Note

Farzad and Rhonda Dibachi on how this translation began with a gift, what it meant to live beside fifty thousand couplets, and how the work was done.

It Started with a Gift

It started with a gift. A friend, Soroush Kaboli, gave us an enormous, beautiful edition of the epic after we took Niku public in 2000. It was one of those objects you don’t forget. It was large, heavy, full of flowing Farsi calligraphy, framed with intricate borders and illustrated in several styles of the Persian miniature tradition. It felt special right away. We both love books, and this one had real presence: a large-format folio, approaching elephant folio scale.

At first, that’s mostly what it was to us: something to admire. Farzad would leaf through it, recognizing passages he remembered from his father and reading bits here and there. But Rhonda couldn’t read it, and as beautiful as it was, it didn’t give us the story.

Later we bought the Warners’ translation (Arthur George Warner and Edmond Warner, nine volumes, 1905–1925), the full Edwardian set. It did what it was supposed to do, but it had that stiff, formal tone. Rhonda would read, and then questions would come up: “Really? Caesarian?” or “Why don’t you ever take me to a Zoroastrian fire temple?” It became a running joke. The translation was impressive, but held up next to the Farsi original, something was clearly missing. You could follow the story, but you couldn’t feel it.

At some point Farzad translated a single couplet, just to show what it might sound like in English if it were done differently. That was all it took. The difference was obvious, and once we heard it, we couldn’t un-hear it.

The two of us came at it from opposite directions. Farzad knew the Shahnameh in fragments, learned episodically over a lifetime; this project was the first time he read it from beginning to end. Rhonda read it as one reads a novel, page by page, start to finish. Neither reading was the poorer for it.

Living with the Poem

No translation of the Shahnameh can be undertaken casually. Its scale alone resists ordinary measures of literary labor. We measured this effort in years. Fifty thousand couplets is not a project you finish; it is a thing you live beside.

The work advanced slowly, through repeated cycles of reading, interpretation, composition, revision and alignment. Earlier decisions are judged by later ones; the voice must stay continuous across thousands of transitions. The epic does not produce isolated moments of intensity so much as accumulated weight. Grief comes not once but in cycles. Loss recurs. Generations rise with hope and fall with depressing inevitability. The result is a sober gravity: this is a world where tragedy is natural.

Why This Translation

English readers have been served better than is sometimes said: the Warners rendered the whole epic a century ago, and many current English translations are good reading, but incomplete. What has not existed is a complete Shahnameh in modern English rhymed verse: heroic couplets, start to finish. We undertook this translation to fill that gap, and to honor the epic’s true nature: one continuous vision spanning mythic, heroic and historical time.

We were not trying to simplify or reinterpret the work. The goal was to rebuild it in a way that preserved the meaning and restored the structure: something that sounded right when read aloud and could carry the story the way it was meant to be carried.

We use the word translation loosely, and with deep respect for Ferdowsi and for everyone who has carried the Shahnameh forward. No translation can be a true representation of the power of the original; if it were, it would be a different work of art. This is a work in another language, meant to carry these stories in a particular place and at a particular time. If a term of art is wanted, it is an English naqqali translation: living, performable verse that carries the narrative substance along with the oral energy that keeps the poem alive. It is not a crib, an interlinear, a prose rendering, or a line-for-line mapping of Farsi couplets onto English ones.

We started with three governing priorities.

These priorities operate in tension. No single passage satisfies all three perfectly.

The Source Text

The base text was fixed before any translating began. We worked from a printed edition credited to Moscow but carrying the fuller K/VI recension, and treated it as authoritative for this edition. The Moscow edition was our North Star, but “conform to Moscow” needs one qualification. Our printed volume agrees with Mohl where Bertels differs, so our target text is Bertels’ main text together with the K/VI material Bertels prints in the molhaqāt, his appendices: the fullest attested text, not the main text alone. A reader checking this translation against Bertels’ main text will meet roughly eighteen hundred couplets he does not expect. Those couplets are in our source, and they are in the appendices of the very edition we followed.

As the translation took shape, we checked it against the Warners’ English, the Ganjoor online corpus, and two printed critical editions, Mohl and Moscow. These are not four independent witnesses, and it is worth saying so: Ganjoor’s paper source is the Moscow edition, so those two belong to one family. Read properly, the comparison gave us two real controls rather than four, and we used them to confirm that our structure, sequence and length were sound.

On length we stayed close to the source, and can be precise about it. This translation runs to 51,547 English couplets across 951 chapters, against the 49,610 couplets of the Ganjoor text — about 104 percent. Most of the surplus is the K/VI material, carried faithfully; a translation of this recension that did not run long against the vulgate would be the one with a problem. Chapter by chapter, our sections track their Moscow and Mohl counterparts closely. The printed editions and manuscripts behind all of this are described in Manuscripts, Editions and Translations.

A single example shows how these versions relate. Our section titled “1. The Greatness and Fall of Jamshid” appears in Ganjoor as:

بخش ۱ – پادشاهی جمشید هفتصد سال بود: گرانمایه جمشید فرزند او

In the Khaleghi-Motlagh edition, the same section is labeled more simply:

پادشاهی جمشید

Note the overlap (in bold):

بخش ۱ – پادشاهی جمشید هفتصد سال بود: گرانمایه جمشید فرزند او

In the Florence manuscript tradition, this material appears as the beginning of “The Reign of Jamshid,” often counted as a later section (for example, section 4). The Moscow and St. Petersburg traditions use comparable titles and the same sequence.

From Farsi to Prose to Verse

The narrative was first rendered into English prose. This allowed the structure of the poem (its actions, relationships, and narrative logic) to be clarified before any attempt was made to shape it into verse. The emphasis was on precision and transparency rather than style. We did not try to preserve line-by-line fidelity, because Farsi and English do not match grammatically and couplets did not correspond to complete sentences. Also, the original uses repetition and invokes common sayings with various flourishes that are multiple couplets long.

Once the prose layer was complete, we began converting it into metered, rhyming English verse. This became its own production cycle: draft, test, revise, test again. We took the document, built a rhyming dictionary and parser, wrote the couplets, scored all couplets on rhyme and meter, and iterated.

The couplets were then refined into a consistent heroic mode. This involved the careful adjustment of diction, rhythm, and tone so that the language remained elevated without becoming archaic, and formal without becoming rigid. Attention was given to rhyme and meter, which were evaluated and revised repeatedly. Each passage was worked until it reached a threshold of formal integrity, what might be called a “passing grade.”

Names, Places and Spelling

In a rhyming translation, name inconsistency is not just annoying; it breaks the work. The Shahnameh has a cast of hundreds: kings, warriors, courtiers, supernatural beings, places, and concepts. Many names repeat across generations. There are multiple Pashangs, sometimes even overlapping in the same timeframe, and at least seven Bahmans. The size of the work and the requirements on the part of the reader to keep everyone’s identity separate quickly became unmanageable.

This grew into a name register where every name was logged and tracked: characters, locations, dynasties, alternate spellings, pronunciation, genealogy, key deeds. This gave us a working system and allowed for creation of family trees. What started as a simple list became a core part of the project.

Spelling and pronunciation are tightly linked, and transliteration forced tradeoffs. A name had to meet three constraints: readability to an English speaker, reasonably close to the Farsi pronunciation, and usability in verse. Some names were fixed. Rostam’s horse, Rakhsh, is iconic, even though it barely rhymes with anything in English. Others gave us more flexibility.

Kioumars, for example, appears only a handful of times, which allowed us to make a more deliberate choice. That choice was not trivial. The name has multiple established forms: Kiumars, Kayumars, Gayomart. A phonetic rendering would be /kaj.uːˈmærs/, or “key-oo-MARS,” with the final sound an “s,” not a “z.” That distinction matters in verse. We considered several options:

KeyoomarsKeyoumars

…and rejected them for different reasons. Too awkward on the page, too ambiguous in pronunciation, too hard to read. We settled on Kioumars as the best balance.

We had similar debates about diacritics. Rhonda wanted none of them. Farzad wanted his native tongue to not be butchered. In the end, we chose readability. The goal was not to produce a linguistically correct document (requiring readers to adopt a non-English phonetic system): we wanted a document that sounded, when read aloud, like English with standard English sounds. The result is a naming system that is consistent throughout the work.

The location list similarly began as a simple “where is this place” and “why is this place important.” Like the name registry, we created and tabulated a location registry to answer these questions in a searchable and algorithm-friendly format.

That system also reveals the shape of the poem. Rostam’s name appears far more often than any other figure’s; even his father Zal comes only fourth, behind Giv and Afrasiab. The counts are in The Shahnameh in Numbers. The Glossary gives each name’s Farsi spelling, pronunciation and first appearance; this section explains why the choices were made, not what they were.

What We Chose Not to Do

Deliberate Constraints

Certain features of the original text cannot be carried directly into English without distortion. These were treated as constraints rather than failures.

We did not attempt to reproduce the original Farsi quantitative meter, which has no natural equivalent in English. Instead we use iambic pentameter in heroic couplets, which most closely approximates the dignity, drive, and self-contained couplet structure of the Shahnameh’s original verse.

We did not preserve line-by-line correspondence. Farsi couplets do not map cleanly to English syntax, and strict alignment would compromise clarity and flow.

We standardized names and references. Variations that exist across manuscripts or traditions were resolved into a consistent system to maintain continuity within the work.

These decisions are not editorial shortcuts. They are structural accommodations required to carry the epic into a different language while preserving its narrative and formal integrity.

The Couplet

The Heroic Couplet in English

The primary unit of this translation is the rhymed couplet: two lines forming a closed unit of sense and sound. Each couplet follows a simple scheme:

A successful couplet must:

Evaluation and Grading

Each couplet is evaluated across a small set of functional criteria:

These criteria were applied iteratively. Some aspects were performed computationally, such as rhyme detection and repetition analysis, but final judgment was always auditory and contextual.

The reference point is consistency at scale. A passage was considered acceptable when it maintained a high level of integrity across multiple readings, both silent and aloud. The cumulative effect matters more than isolated excellence.

Each couplet could have a:

Imperfect rhymes were allowed only where nothing better existed and the context carried them.

A couplet could look good on the page and still fail when spoken out loud. The standard was not “close enough.” Take this early draft:

One day upon his ivory throne he sat,A turquoise crown upon his head well set,

Sat and set are a near rhyme, and we could do better. Forcing the second line only made things worse:

One day upon his ivory throne he sat,A turquoise crown upon his head well set thereat.One day upon his ivory throne he sat,A turquoise crown upon his head and hat.

The first is archaic; the second is garbled. At that point the answer is not a better second line but a different pair, and the couplet stands in the poem today as:

One day he sat upon his ivory throne,With turquoise crown, those serpents fully grown.Zahhak on his throne, A5.5

Once again, the objective was to let the rhyme propel the story; we wanted the rhyme to be the beat of the work, just the way the beat of a song, or a naqqal’s verbal rhythm, could make a story go forward effortlessly.

How to carry a long rhymed poem into English is the $60,000 question. Translators of Dante’s Divine Comedy, with its interlocking three-line rhymes, have tried just about everything, and Spenser had to invent his own stanza just to write an English epic in The Faerie Queene. Along the way, translators of long poems have:

We’re not poets, and we’re not claiming a seat at that table. We picked the heroic couplet because Ferdowsi wrote in couplets, and it was the closest fit we could find. And we know the risk. When Alexander Pope translated the Iliad into heroic couplets, the scholar Richard Bentley supposedly told him, “A very pretty poem, Mr. Pope, but you must not call it Homer.” Fair enough.

Reading It Whole

Once a section was complete, it was read again at the level of the story. We examined the arc of each episode to ensure that nothing essential had been lost from Farsi to prose to verse. Where the narrative had weakened, the text was revised, sometimes substantially. In this way, the translation proceeded not only line by line, but in cycles, constantly returning to earlier stages to maintain coherence.

Next, we needed to make sure the translation was consistent. Consistent with tone? Are the key concepts referred to consistently? Is the drama apparent? Is it the rollicking good story that we know it to be?

Only after we had the full poem in English verse could we really start to understand it. Patterns became visible. We could track recurring behaviors, dynastic downfalls, hereditary moral failings, and repeated scenes.

Concepts that were passed over at first became a brick in a larger wall. For example, the concept of farr was more important than originally perceived. Many English translations use “grace,” which, to be fair, is a pretty good approximation. In our early versions, we hadn’t quite figured out how important farr was, so sometimes we used farr and sometimes “grace.” Later, when the concept rightly became as important to us as it needed to be, we changed all these references back to farr. That sometimes broke the rhyme. So, we would have to redo the rhyme. Another iteration.

We were able to use the rhyming system to search for excessive repetition, as well. Some repetition is part of the epic’s design, especially for performance. Some is simply exhausting in a modern textual version. The Iran–Turan standoff under Kay Khosrow, sometimes referred to as the Piran cycle, became our test case for how much repetition could be reduced without damaging the structure.

The Tools

This translation was made with substantial computational help: the rhyming dictionary and syllable parser; scripts that scored every couplet for rhyme, syllable count and stress; registries that held names, places, objects and family relationships and enforced their spellings across fifty thousand couplets; repetition and collocation analysis; and concordance tools that aligned our sections against Ganjoor, Mohl and the Moscow edition.

We used large language models (LLMs) extensively throughout this translation. We built and adapted custom LLMs specifically for the Shahnameh: for investigating historical allusions, comparing textual traditions, tracking characters and places, testing language, and helping maintain consistency across an immense narrative.

Their greatest contribution was the ability to work on a difficult passage and link it to the reference behind it. A name that seemed incidental opened up a disputed genealogy; a place can belong at once to geography, legend, and later historical memory; a king’s epithet may refer to an older religious tradition. We used our specialized models to assemble leads, explain unfamiliar references, identify relevant scholarship, and show where evidence was uncertain or where scholarly accounts disagreed.

They were especially useful in navigating differences among textual traditions. The surviving manuscripts and printed editions do not always divide the story in the same places or agree on wording, names, and episodes. We used dedicated LLM workflows to find corresponding passages, compare the organization of editions, flag possible interpolations or omissions, and bring contradictory readings to our attention. That helped us distinguish a textual difference from a difference in chapter numbering or spelling. It also prevented a choice made in one passage from quietly conflicting with a choice made several hundred pages later.

This was research assistance, not an abdication of judgment. LLMs can be fluent and wrong; they can conflate similar names (especially when working across different languages and alphabets), invent citations, and present a conjecture as absolute fact. We treated them as leads requiring corroboration. We decided what belonged in the translation, what required explanation in a note, and what remained uncertain. The tools broadened and accelerated our inquiry; the decisions and final text remained ours.

The LLMs changed the practical scope of what two translators could undertake. They found obscure references, contrasted competing explanations, compared passages between editions, exposed inconsistent naming, and proposed new rhyming words for a stubborn rhyme. We developed these specialized tools because the Shahnameh demanded unusually long-range memory and unusually wide-ranging research.

What the machinery did not do was decide. Every line was read aloud and kept, revised or thrown out by ear, and the arguments that shaped the translation (whether farr could be pronounced in English, how much repetition a modern reader will bear, whether Kioumars or Keyoomars) were had between the two of us, across a kitchen table, for years. The tools made a work of this scale possible for two people in a decade rather than impossible.

In outline the work ran: fixing and dividing the source text; Farsi into English prose, with the registers of names and places; prose into couplets; heroic refinement; grading for rhyme and meter, and iteration until each passage passed; a consistency pass across the whole poem; and validation of each story’s arc and of continuity between stories.

To give a sense of scale, the system supporting this work generated approximately 2.5 terabytes of working data (registries, intermediate representations, competing versions), while the text of the narrative itself is roughly 2.5 megabytes, on the order of a million times smaller.

AI can suggest a rhyme, but it can’t hear one; and it does not understand meter. Those parts were ours. But it is a great tool for translation. After all, this is not a new story. The story already exists, and has for a thousand years; we were just translating it.

One final help performed by the LLMs was editorial scoring. We created five different personas: a tenth grader, a New York Times editor, an Iranian diaspora reader, a native English speaker with no knowledge of the Shahnameh and a Shahnameh scholar. Each of the 951 chapters was scored by these five personas. We used these scores to make any final-final adjustments.

A Note to Readers

Near the end of the poem, Ferdowsi stops to take stock of what he has made:

Full sixty thousand couplets I have wrought,A lasting monument of verse and thought.No Persian book of verse on earth can showSuch vast endurance in its measured flow.Sift all these lines, and critics will agreeFive hundred flawed ones are the most they see!E7.1, The Prelude (Volume Three)

Ferdowsi had earned that confidence. We would not make the same claim for ourselves. A translation of more than fifty thousand couplets, made by two people, will have mistakes in it, however many times we have read it aloud.

If you find a couplet that is clearly wrong (a line that misreads the Farsi, a name in the wrong place, a passage that says something Ferdowsi did not), we would be grateful to hear from you at Editor@newnaqqali.com. We will correct it in the next printing, send you a copy of the updated volume, and thank you by name in the book and on our website.

Completion

Completion carries a rare kind of satisfaction. For so long this work has been a world we lived in, something we moved through daily, returning to its voices, its rhythms, its people. To have spent that much time inside it has been a privilege. There is real pride in bringing it to its end, and a quiet awareness that something sustained and absorbing has reached its natural close. It is a feeling the poem itself knows. Ferdowsi laid down his pen after thirty-five years with a claim that has turned out to be true:

Now this renowned and royal tale is done,The land will speak of me beneath the sun.I shall not die; alive in words I stand,For seeds of speech I scattered through the land.All prudent men of counsel, faith, and nameWill praise me when I leave this mortal frame.The completion of the Shahnameh, E11 (Volume Three)

This is an attempt to carry the naqqal’s torch into a new language. The epic has survived a millennium because each generation found new means of transmitting it: oral performance, manuscript illumination, courtly recitation, print. This translation is offered in that same spirit: not as a monument to the past, but as a living version of it.

We ended with a complete Shahnameh, carried from beginning to end, without cutting it down or picking favorites. We stayed with the whole story written in modern English verse, meant to be read comfortably or spoken out loud, with rhythm and rhyme doing the work they’re supposed to do.

When my father clapped his hands and began to recite, I did not understand that I was hearing the echo of a thousand years of culture. The Shahnameh has survived because generation after generation refused to let it disappear. This translation stands within that same chain.

Farzad Dibachi and Rhonda Dibachi