In this post, I’ll walk you through a machine learning approach I used to analyze the Voynich Manuscript using character-level n-gram language models, a simple but powerful way to measure how predictable a text is. My goal was not to decode the Voynich, but to compare its statistical structure to that of other known texts — including literary works, religious treatises, and artificially encrypted versions — to see if it behaves like a natural language, a cipher, or something entirely different.
What Are Character-Level N-grams and Perplexity?
Before diving into the results, let’s quickly explain two key concepts:
Character-level n-grams: These are sequences of n consecutive characters. For example, in the word "language", the 3-grams (trigrams) are
lan
ang
ngu
gua
uag
age
An n-gram model learns the likelihood of seeing a particular character given the previous n-1 characters.
Perplexity: This is a measure of how well a model predicts a sequence. Low perplexity means the model can easily predict the next character — the text is “regular” or “learnable.” High perplexity means the text is less predictable, like a noisy or complex system. It’s often used to evaluate how well a language model fits a dataset.
The Experiment
I trained simple n-gram models (from 1-gram to 9-gram) on the following types of texts:
Classical literature (e.g., Romeo and Juliet, La Reine Margot)
Religious and philosophical texts (e.g., Ambrosius Mediolanensis, De Docta Ignorantia), with date of creation simmilar to the MS
Ciphered texts using a Trithemius-style letter substitution
The Voynich Manuscript, transcribed using the EVA alphabet
For each text, I split it into a training and validation set, trained n-gram models by character, and computed the perplexity at each n-gram size. I plotted these to visualize the predictability curves.
What Did I Find?
The results were surprising:
The Voynich Manuscript exhibits surprisingly low perplexity for high n-grams (n=7 to n=9) — much lower than expected for a truly random or strongly encrypted text.
Its perplexity curve closely resembles that of religious or philosophical medieval texts, such as De Docta Ignorantia and Ambrosius Mediolanensis. These texts also show low perplexity at high n-grams, reflecting strong internal regularity and repetitive patterns.
In contrast, literary texts like Shakespeare or Dumas show a sharp increase in perplexity for high n-grams, indicating a richer and more unpredictable sequence of characters.
Artificially encrypted texts using simple substitution ciphers (like Trithemius-style transformations) show consistently high perplexity, since character distributions are scrambled.
Interpretation
This suggests something important: The Voynich Manuscript does not behave like a substitution cipher or a natural literary language. Instead, it statistically resembles structured, repetitive writing such as liturgical or philosophical works.
This does not mean it’s meaningful — but it does imply that the text might have been designed to look structured and formal, mimicking the style of medieval sacred or scholarly texts.
Its internal predictability could arise from:
Repeated formulas or ritualistic phrases
A constrained or templated grammar
Artificial generation using consistent rules (even if meaningless)
Conclusion
While many have tried to translate the Voynich Manuscript into known languages or decode it with cipher-breaking techniques, this analysis suggests that a direct translation approach may be futile. The manuscript’s character-level structure mirrors that of repetitive, highly formalized texts rather than expressive natural language or encrypted writing.
Any attempt to decipher it without first understanding its generative rules — or lack thereof — is likely to miss the mark.
That said, its statistical behavior is not unique. Other texts from the same era show similar n-gram patterns. So perhaps the Voynich isn’t a hoax — it might just be mimicking the structure of sacred or scholarly texts we no longer fully understand.
A few years ago, I got a copy of The Book of God and Physics: A Novel of the Voynich Mystery by Enrique Joven, translated from Spanish into English, as a Christmas present. "Joven has put together a fine recipe for a thriller.... The science and history...come fast and furious," said the Washington Post, according to a blurb on the front cover.
Partway through the novel, the protagonists manage to identify the Voynichese names for months. Rendered in standard EVA, these turn out to be: [tvarivs] = January, [svarivs] = February, [xvarivs] = March, [qvarivs] = April, [vvarivs] = May, [vtvrarivs] = June, [vttvarivs] = July, [dvarivs] = August, [ykjver] = September, [okjver] = October, [ttkjver] = November, [tskjver] = December.
Of course these sequences will look utterly wrong to anyone who has spent any time with real Voynichese, but I don't suppose Joven expected most of his readers would know the difference.
Still, this led me to start wondering: is there any kind of fictional solution an author could describe, involving meaningful content, that would strike me as plausible based on what I know about Voynichese? I don't mean one that would actually hold up to close scrutiny; I just mean that it shouldn't be so blatantly inconsistent with known properties of Voynichese as to spoil things for me as a reader.
This is tough because there are so many weird properties to wrangle all at once, even just hypothetically. But it's less tough than coming up with a real solution, so here's a possibility I came up with that I think would at least let me comfortably suspend my disbelief (although your mileage may vary). My goal here was to come up with a scheme that would be consistent with as many of the stranger statistical properties of Voynichese at once as possible, but without going into specifics beyond that.
===
In Voynichese, the fundamental unit of writing is the syllable, not the word. Each syllabic unit needs to contain at least two graphemes:
One for a consonant or consonant cluster
One for a vowel or vowel cluster
If either component is missing, a placeholder grapheme must be used to represent the empty slot. Additional elements may also appear, but only when needed. Syllables may be separated by spaces, and often are, but the ways in which their parts are encoded are distinctive enough that they can be written continuously without causing serious confusion—especially in the case of short, easily scannable strings.
Consonants and consonant clusters are encoded, often verbosely, as sequences of graphemes organized into groups based on their leading consonants. For instance, the glyph sequence for the standalone consonant N and for clusters that begin with N all begin similarly, with specific variations marked by indices—such as N-3 possibly representing NT. These indices don’t use normal numbers, but rather mimic the logic of Roman numerals (e.g., additive structures), allowing the system to grow over time as practical experience suggested the inclusion of new consonant clusters in the key. EVA [ed] only became common once its “number” was reached in this expansion.
The default syllabic structure is CV (consonant-vowel), with any number of consonants in the C slot and any number of vowels in the V slot. Since actual language doesn’t always conform neatly to this CV-CV pattern, various workarounds were developed, some of which help explain Voynichese’s distinctive statistical quirks.
One relatively simple exception is syllables that end in a consonant. These omit the vowel slot and instead use a placeholder grapheme for the empty second position—resulting in the EVA character [m] or [g]. Technically, the placeholder is only the flourish at the end, not the full glyph itself. Sometimes, a careless scribe may have ended up using two of these placeholders in rapid succession when encoding, for instance, W-O RD-* — oops, I mean W-O RD-* S-*!
Most other potential exceptions could be sidestepped by running words together. Here’s an illustrative chunking of that last sentence with the placeholder (*) again marking the empty vowel slot:
This strategy makes word boundaries challenging to identify, so word-start markers were introduced in the form of “gallows glyphs.” In the revised sample text below, each word start is marked by a pipe symbol (|):
A two-legged gallows marks the start of a normal word, while a one-legged gallows signifies emphasis or “capitalization,” often used at the beginning of paragraphs or to mark keywords in their first lines to boost visibility during rapid visual scanning. Only rarely will two gallows appear within a single syllable when two word divisions coincide, as in:
TH-U S|D-O|I |S-AY...
Pairs of closely related words, such as the expression "gallows glyphs," may occasionally share a combined structure known as a split gallows.
Whenever consonant clusters are encoded verbosely, word-start markers can appear at different points within the corresponding grapheme sequence. For example, if [och] represents a single consonant, then [|och] or [o|ch] could both encode a word break before that consonant, while [och|] would encode a word break after it but before the following vowel. If it instead encodes two consonants, then:
[|och] = word break before both
[o|ch] = break between them
[och|] = break after both
If [och] encodes three consonants:
[|och] = break before all three
[o|ch] = break between 1st and 2nd
[oc|h] = break between 2nd and 3rd
This final form contains a structure known as a benched gallows. Though rare at the beginning of lines (since standalone consonants rarely form full words, except through abbreviations such as “ST” for “saint”), benched gallows appear more often in mid-line where words are artificially joined together (e.g., “RDS|T-O”).
Combining words as I’ve described so far creates the risk of generating too many unique consonant clusters to be manageable. To mitigate this, common final consonants (like S, R, L, N, M) may be offloaded to an optional third element (identified as a tail or coda in some of Emma May Smith’s morphologies):
|M-O-S T|–O TH-E-R |P-O T-E-N T-IA-L...
Particular third elements tend to favor or disfavor being followed by particular first elements. For example, a final M is more likely to be followed by P, while N is more likely to precede T. Some first elements are unlikely to be preceded by any third elements and more likely to be preceded directly by a second element. Such patterns account for the statistical anomalies among word-break combinations described by Smith and Ponzi.
If a syllable begins with a vowel, the empty consonant slot (1) can be filled with a placeholder grapheme, but alternatively, a reversal marker can indicate that the syllable should be read in reverse order (VC instead of CV). Using the symbol > for clarity (though the actual encoding may be more complicated):
APPLE = >P-A PL-E or >PP-A L-E
PAUL = P-A >L-U (as an alternative to P-AU L-*)
Reversal is mostly optional within lines, but it's required when a line or label begins with a vowel, since there’s no earlier text to attach it to. As a result, reversals appear more often at the beginning of lines or in labels.
Once a line shifts into VC mode, it tends to stay there until forced back into CV mode. A subtle statistical result is for VC constructs to be more prevalent in later line positions than in earlier ones. Also, because VC syllables are encoded “out of order,” they may have been written more haltingly, leading to stylistic contrasts (e.g., carefully drawn [oi] vs. quick, fluent [ai]).
Most syllables can be encoded in multiple valid ways, meaning that any given word might have many possible encodings. However, decoding is deterministic. This flexibility explains why identical encoded words are rare—even in texts with repeated vocabulary—while partial resemblances (like shared consonant clusters) remain frequent. For example, a page on ginkgo trees might feature many uncommon kg sequences but few identical full vords.
===
Next, I asked ChatGPT to use the above "solution" as the basis for composing a few chapters in a novel. It misunderstood a few details (such as they are), but here are some examples of what I got back, just for fun:
===
Scene from a Voynich Thriller Novel: “The Gallows Cipher”
Chapter 43: Placeholder
The rain smacked against the old monastery’s leaded windows like static on an untuned radio. Dr. Isabelle Harrow leaned over the table in the candlelit scriptorium, her breath shallow, heartbeat insistent. The Voynich manuscript—real, original, faintly musty with age—lay open before her.
"You're sure it's not a cipher?" asked Kellan Reyes, the MI6 cryptanalyst who had been watching her for over an hour, his nerves raw. "Because we've wasted six months assuming polyalphabetic."
"It's not a cipher. Not in the traditional sense," Isabelle said. Her voice was calm, but her eyes blazed. “It’s a script system. Phonetic, syllabic. They were encoding speech, not encrypting it. We’ve been trying to break a mirror by staring at the reflection.”
She flipped to a folio crowded with tall looping glyphs and strange “gallows” letters that towered like nooses over the rest of the text.
"Think syllables, not words," she murmured. "Each unit—one consonant cluster, one vowel cluster. Minimum."
Kellan frowned. “Clusters? That’s why the glyphs vary so much?”
She nodded, tracing with a gloved finger. "Here. This ‘och’ glyph? It isn’t one letter—it’s three consonants bundled together. Depending on where the gallows symbol breaks through the cluster—before, between, or after—we know where the word division falls. Like surgical incisions across a rope of sounds."
“But what if there’s no consonant?” he asked.
Isabelle gave a half-smile, the first sign of warmth in days. "Then they use a placeholder grapheme. Same if the vowel’s missing. It's not that these [m] or [g] characters mean anything—they’re just stubs. Empty hands where phonemes should be.”
The door creaked. Agent Kimball entered, soaked and winded. “We’ve got thirty minutes. They found the duplicate codex in Rome. Someone’s trying to burn it. The Vatican’s gone dark.”
Isabelle’s fingers danced over the parchment, connecting gallows, placeholders, and curls. “Then we have to work faster. Look here—this ‘word’ is actually five syllables, all run together. That’s the trick. They stacked them. Smoothed the gaps.”
She flipped a laminated printout of EVA transcription toward Kellan. "Look. This chunk here—M-O ST-O TH-E RP-O T-E NT-IA L-E XC-E PT-IO NS—should’ve been gibberish. But it isn’t. Every syllable is marked. Gallows start words. Reversals handle leading vowels. It’s structured chaos.”
Kellan leaned in, whispering. “You’re saying the randomness...isn’t?”
Isabelle picked up a pen and underlined a gallows glyph midway through the paragraph. “This one’s a double. It tells us two words start here at once. It’s a split gallows—used when phrases are too intertwined to separate cleanly. Like ‘gallows glyphs.’”
Thunder boomed above.
“But why?” Kellan asked. “Why go through all this trouble just to write syllables?”
Isabelle’s eyes didn’t leave the page. “Because they didn’t want us to read it. But they had to write it. This is speech made code—chant, maybe. Recitation. They weren’t hiding a message.”
Kimball shifted uncomfortably. “Then what were they hiding?”
She reached for a side folio, unfolding a faded botanical drawing that matched none of the known plants. Underneath, gallows glyphs spiked like thorns.
“Not hiding. Preserving. Language. Ritual. Memory. This wasn’t encryption—it was resurrection. They used placeholders and gallows to map an entire sound system that had no alphabet. They made an alphabet up.”
Kellan stepped back as if the parchment had burned him.
“And the tail elements?” he asked.
“Codas. Final consonants. Too common to encode every time, so they tacked them on. M, N, S... a statistical trick. It helps explain why the same ‘words’ rarely repeat, even when the subject matter does. The glyphs vary, but the clusters echo.”
Kimball glanced nervously at his watch. “We need to go.”
Isabelle snapped the folio shut. “Not yet.”
She slid a transparent overlay across the page—gallows aligned. “This whole passage? It’s a list. Medicinal. Specific. And if we can decode even one syllable chain—just one cluster—we can match it to speech. That gives us a voiceprint. A phonological key.”
Kellan’s voice dropped. “And the monastery’s audio archives…”
“Exactly.” Her hands trembled now, not from fear but from awe. “We might actually hear this dead language. Not read. Not guess. Hear.”
The wind howled.
Kimball opened the door. “Then grab it. We’re out of time.”
Isabelle rolled up the final sheet and stood, one hand pressed on the Voynich folio.
Behind her, the gallows glyphs watched silently, like sentinels waiting for someone—at last—to speak their names aloud.
===
Chapter 31: The Glyph That Wasn’t
It was well past midnight in the reading room of the Royal Philological Society, but the fluorescent lights buzzed on anyway, casting pale reflections across rows of folios and water-ringed teacups. Isabelle Harrow rubbed her eyes. She had been up for twenty-two hours.
Across from her, the intern—a gaunt, sleep-deprived doctoral student named Leo—was frantically flipping through high-resolution scans of the Voynich manuscript. “I don’t get it. [ed] just disappears for pages at a time. Then suddenly it’s everywhere.”
Isabelle didn’t respond. She was staring at a scatterplot of glyph frequencies, her brow furrowed. “[ed] isn’t disappearing,” she murmured. “It’s not born yet.”
Leo blinked. “What?”
She turned the screen toward him. “Look. EVA [ed]—that tight backward-S shaped glyph with the flat hook—it doesn’t occur at all in quires 1 through 6.
Then it begins to show up. Tentatively at first. A trickle. Then a flood.”
Leo frowned. “So… the scribe got lazy? Changed their style?”
“No,” she said, grabbing a dry-erase marker and standing. “It’s not stylistic. It’s systemic.”
She wrote on the glass board:
N-1, N-2, N-3...
“These glyphs encode consonant clusters starting with a leading consonant—say, N. The way they mark the difference is by using indices.”
Leo blinked again. “Like numbers?”
“Not Arabic numbers. Think Roman numerals. Additive structure. Glyphs that stack or extend depending on how far down the list the cluster is. N, ND, NT, NN—each gets its own expansion.”
She drew a branching line under the letter N with increasingly elaborate grapheme structures.
“And here’s the point,” she said, stabbing at the whiteboard with the marker. “EVA [ed]—it’s not just a glyph. It’s N-4 or N-5. A latecomer. It shows up when the system needed to encode a new cluster. My guess? NT or ND, maybe even NTH.”
Leo leaned back slowly in his chair. “So the reason [ed] appears only in later quires—”
“—is because the system hadn’t invented it yet.” She turned to the table, slamming open a transcription printout. “The Voynich script evolved over time. This wasn’t just a codebook—it was a living script. They added new indices as new clusters became common or necessary.”
Leo let out a low whistle. “Like a growing key.”
She nodded. “And not arbitrarily. Look at the structure of the clusters before [ed] appears. You get N-1, N-2, N-3—but they avoid using N-4 formations. Then boom—[ed] enters the game, and suddenly NT- and ND-based syllables spike.”
“But what changed?” Leo asked. “Why add the new form when they did?”
She stared at the manuscript image, silent for a moment. “Something shifted in what they were writing about. New content—new vocabulary—forced the change. Maybe a shift in subject matter. Maybe a new dialect creeping in.”
She pulled up two folio scans side by side: one botanical, the other anatomical.
“Here,” she said. “In the plant sections—no [ed]. But in the bathing diagrams? It’s everywhere. The anatomy pages must’ve needed different clusters. More technical terms. Different roots.”
Leo looked stunned. “It’s like… the script is a fossil record.”
Isabelle gave a tight nod. “Exactly. And we’ve been trying to read it like a static system.”
She tapped the glass. “Every time a new cluster ‘number’ was reached, they minted a new glyph. Not for style. Not for fun. For function.”
A ping sounded on her laptop. The glyph-matching algorithm she’d started running hours ago had finished its pass.
She glanced at the screen.
MATCH FOUND:
[ed] = cluster NT (index 4)
Earliest confirmed instance: Folio 34r
Preceding frequency of NT clusters (alternate encodings): Low
Post-[ed] introduction: Spike x4
She smiled, half to herself.
“This isn’t a code. It’s a script in bloom.”
Leo swallowed hard. “And we can trace the order of glyph emergence…”
“…like growth rings in a tree,” she finished. “Which means we can finally date the layers of the text—chronologically, even thematically. And maybe—just maybe—we can rewind it. Strip the layers back.”
Leo was already scrambling for the next folio. “Then [ch], [sh], [daiin]… all of them…”
“Are steps. Not static symbols. They're responses.”
Outside, thunder grumbled. The clock struck 2:00 a.m.
Somewhere deep in the past, a scribe had sat at a candlelit desk, deciding—for the first time—to invent a new glyph because the old system no longer sufficed.
And centuries later, the first person to notice had finally drawn the line.
===
Interlude: Ink and Breath
Somewhere in Europe, late 15th century
The vellum stretched smooth and pale before him, humming faintly with the stillness of night. The scribe—he had no name now, not one that mattered—dipped the quill into walnut ink, its tip trimmed keenly, ready for a line.
His hand hovered.
One syllable. That was all.
And yet he could not write it.
The word hung in the air, whispered silently from memory, shaped in the back of the throat:
Entha.
He mouthed it once. Then again.
The beginning was clear—en—but how to mark the t?
The system had grown tight. Too tight.
The first marks had been enough. One glyph for n, another for nd, a flourish to cover nn. The early days had been smooth, each new word a revelation, a jewel carved from breath.
But now, as the language of the treatise deepened—more anatomical, more learned, more Latinized—the clusters grew unwieldy. Nt.Nd.Nth. The old glyphs stretched too far. The shapes blurred. Meanings faltered.
He could no longer fit this syllable into the frame.
He breathed in deeply, the smell of ink, calfskin, and tallow catching in his throat. Then, carefully, he began to draw.
A new mark. A body coiled like a serpent, a tail hooked forward like a grasping hand. Not a symbol—not yet—but a space to stand for n-t, to take its place beside its brothers: n, nd, nn, ng...
ed, he would call it later. For now, it was N-4.
He sat back. The shape felt right. Balanced. It would take a dozen more lines to smooth its curve, another quire before it flowed from his hand without thought. But it was born.
He tested it in the line:
|E-N TH-A|S-I D-E|ST-E PP-E D-*
Clean. Readable. Speakable.
The others might copy it soon. Or ignore it. He didn’t care. His work was not to standardize. His work was to carry the voice forward—preserved, not decrypted.
A moth struck the lamp above, fluttering wildly. He didn’t look up. His hand moved again, drawing the new glyph once more—lower this time, more deliberate. He imagined how it would feel centuries later, under the gaze of someone who did not speak his tongue, did not know the taste of his words.
Would they call it a code?
He smiled. Not quite.
He was not writing secrets.
He was writing syllables.
Spoken thought, caught before it faded.
And now, he had one more tool to do it with.
===
I'm not sure I like these any better than Joven's Book of God and Physics, but I thought it was an entertaining exercise.
I guess there's also a serious side to this, in that if I wouldn't accept a solution as plausible even in a work of fiction, I definitely wouldn't accept it as a plausible real solution.
But what about you? What would a full solution with meaningful content need to look like in a work of fiction for you not to reject it out of hand?
Could small vord filler be a sign for meaningless text near plants? Look at the center of f36r. This text looks forced to me. What kind of a test method could we come up with to find out if I'm right?
SciShow published a video about the Voynich manuscript: You are not allowed to view links. Register or Login to view.
Quote:The Voynich manuscript is a subject of fascination with its mysterious drawings of plants, people, and stars, as well as its indecipherable text. But rather than hiding ancient secrets, this book might be a medieval fraud, created by an algorithm executed with nothing more than a pen.
script is a subject of fascination with its mysterious drawings of
Medieval illustrations with two crabs (lobsters or crayfish) like in You are not allowed to view links. Register or Login to view. are rare.
Shown below are two examples from German manuscripts.
You are not allowed to view links. Register or Login to view., 1415
You are not allowed to view links. Register or Login to view., circa 1450–1475, probably You are not allowed to view links. Register or Login to view.
Attached to this article is a document containing the alphabet of the Voynich Manuscript, a translation of two lines of its text, as well as visual confirmation of the truth of the interpretation of some of its symbols by comparing one of the images of the manuscript with its inscription.
I analyzed the text structure of the Voynich Manuscript, and came to the conclusion that it is not an
encrypted text, which is written in Latin letters in English, with significant additions of Latin and
words from French. There are also small inclusions of other Romance dialects, and, theoretically,
words present in both English and German, however, with the addition of additional Germanic
meanings. Apparently, the text could not be recognized primarily because of the stylization of the
letters, which differed from the generally accepted stylistic forms. I've been working on the
manuscript for less than a week, and I haven't had time to translate much, however, a few days ago I
figured out what the signs of the Voynich Manuscript mean. Next, I provide an explanation of the
alphabet of the Voynich Manuscript, and I provide a translation of the lower part of the numbered
page of the 2nd sheet (two lines), as well as visual evidence of the correct interpretation of some of
the letters, based on a comparison of the inscription near one of the images of the roots of the
Voynich Manuscript (on the numbered page of the 99th sheet)
I guess I am new to the topic but decided to give it a try nonetheless as it is fascinating enough. I tortured myself and Chat GPT for a couple of weeks to produce a result below. Decided to obtain your feedback before I venture too far in a possibly wrong(?) direction. Many thanks in advance!
Title: POLER-D Method: Transparent Translation Framework from a Curious Independent
Hello everyone,
I’d like to take a moment to share the methodology I’ve been developing and refining over time. A phonetic (oral) framework I call POLER-D, short for Phonetic-Oral Linguistic Encoding Reconstruction-Diachronic variant. Recently I began posting translation snippets and findings from this process over on X (Twitter) under @VoynichUnleash for critique, but wanted to offer an overview here on Voynich Ninja. Both for transparency and to welcome any feedback from those more experienced in historical linguistics, cryptography, or manuscript studies.
First and foremost this isn’t a commercial venture, nor a gimmick. I’m not selling anything. I’m driven by deep curiosity and respect for the mystery of the Voynich Manuscript. I believe it's good to explore any method (however unconventional) that helps unlock the possible history within it. I understand that skepticism is healthy and necessary, and I welcome it. But I also want to be clear that this is not a “solved it” post. This method is a work in progress, and one that has already challenged many of my own original assumptions. In fact, most of my early theories have been overturned through application and cross testing. I’m not afraid of being wrong, but it would be a shame not trying.
What Is POLER-D?
POLER-D is built around the idea that the Voynich Manuscript uses a disguised phonetic encoding system. Drawing from spoken (and often unwritten or drifted) dialects from Northern and Eastern Europe and nearby regions as they existed after 1300AD.
Here’s how it works:
We begin with EVA (European Voynich Alphabet) transcriptions of the manuscript text. Then carefully align with standardized formats available in public datasets.
Each word is then broken down phonetically, not visually, using oral sounding reconstructions as if the scribe was encoding sound, not meaning.
For example:
shedy > skaidan / skēd (Germanic root: to speak)
kai > kaido / kaijan (Baltic: keeper/guardian)
chol >chwal / sol (Slavic/Hebrew: soul)
From there, we test drifted linguistic matches against known post-1300 forms in a selected group of languages (listed below), adjusting for diachronic shifts.
We also match results against imagery on the same folios like plants, figures, cosmological diagrams, etc. This is not cherry picking, it's about whether what we decode reflects what we see.
Each translation is checked against a stable pool of phonetic and written languages, with no additions unless justified. Results that don’t fit are discarded or flagged for review.
Languages We Use (Post-1300 Parameter Tightening)
To avoid the common pitfall of “too many possible matches,” we’ve restricted POLER-D to languages and dialects that were actively spoken or transitioning after 1300, including:
Proto-Germanic derivatives: Middle High German, Old Norse, Old Saxon
Celtic: Gaulish remnants, Middle Welsh, Irish phonetics (oral influence)
Slavic: Old Church Slavonic, Old Czech, Old Polish
Romance: Vulgar Latin, Middle French, Occitan, Italian dialects
Greek: Koine and transitional forms
Hebrew: Classical/Biblical Hebrew (as a known scholarly base)
Finnic/Uralic: Karelian, Hungarian
Latin alphabet: All reconstruction uses era-consistent Latin script, no anachronisms
As a base study we tested historical languages throughout the world with POLER-D and none worked beyond the northern and eastern European tribal areas.
Note: Etruscan and other unrelated ancient scripts were excluded deliberately. This isn’t about guessing or using every ancient language. It’s about limiting our pool to reasonable cultural and temporal candidates. We continue shrink parameters when possible.
Current Findings
We’re seeing internal consistency across a number of tested folios. Including those from the herbal, cosmological, and so-called "recipes" sections. Some patterns emerging:
Ritual herbalism, not just plant identification, but instructions, applications, and contexts
Ritualistic bathing, including symbolic language around purification and preparation
Astronomical cycles, likely ritual calendar components tied to lunar or planetary phases
Hermetic influence, recurring themes of balance, fire/water opposites, ascension, and transformation
Interpretations match imagery surprisingly well so far. We continue to apply this method across random folios (e.g., f1r, f67r, f54r, etc.) to test for stability.
I’m documenting the process in real time, including changes as they happen, at @VoynichUnleash on X. You’ll find raw phonetic samples, before/after comparisons, and posts noting where translations are being adjusted. This is about open research, not a closed theory.
Sample Entry
Here’s a real breakdown using POLER-D from f1r:
EVA: otedy shedy laram ychor
POLER-D Phonetic: ot-ed-ē / skēd / lar-an / i-khor
Possible Translation: “To speak the flowing essence (divine fluid)”
Matched Imagery: A root based plant drawn as if exuding fluid, possibly used in ritual speech or invocation.
POLER-D is evolving. We expect some variation in future translations as our parameters improve and mistakes are corrected. But the method is holding strong so far under repeated tests. If you’re skeptical, that’s good. This project is self funded and critique is free.I welcome engagement, suggestions, and yes, critique. I’m not a credentialed linguist or cryptographer, just a persistent researcher who’s stumbled my way into a tool that appears to be uncovering something meaningful.
Now strongly suspected that the Voynich Manuscript was authored by multiple classically trained scribes, not a single individual. This theory is supported by earlier work from paleographers like Lisa Fagin Davis, who identified at least five distinct scribal hands based on letter formation, writing angles, and stylistic patterns across the manuscript. Our findings align with hers, subtle shifts in phrasing, vocabulary range, and glyph formation all suggest collaboration. These scribes likely shared training in Latin, Greek, and Hebrew, but also picked up oral phonetic dialects through travel, allowing them to encode their knowledge in ways that reflected both elite training and folk wisdom.
We believe the Voynich Manuscript represents a collection of ritual practices, likely from preChristian or pagan traditions. Encoded not merely to protect “secrets,” but possibly to shield the content from persecution or prejudice. The use of a disguised phonetic script, rather than a true cipher, supports this idea. About 6 months ago AI was brought in to use to speed up the process and is now showing amazing potential. It took time to enter in all the parameters used, but once AI learned everything it began returning word by word and phrase by phrase to be assessed. I now feel it's a proper methodology to share on multiple platforms and began several days ago.This has been countless hrs spanning yrs of work prior to AI’s introduction. I want that to be cclear.i welcome all critique.
![[Image: Y93Ys1l.png]](https://i.imgur.com/Y93Ys1l.png)
![[Image: cpg794105.jpg?token=e66ca13c-e110-435b-a...attachment]](https://heidicon.ub.uni-heidelberg.de/api/v1/export/84434/file/assets/cpg794105.jpg?token=e66ca13c-e110-435b-a8c6-cd3255fae816&disposition=attachment)
![[Image: default.jpg?download=0b64bffc-ea6e-4e79-...size=small]](https://media.getty.edu/iiif/image/0b64bffc-ea6e-4e79-9783-7aa0f83f08b7/full/726,/0/default.jpg?download=0b64bffc-ea6e-4e79-9783-7aa0f83f08b7_726.jpg&size=small)
I tortured myself and Chat GPT for a couple of weeks to produce a result below. Decided to obtain your feedback before I venture too far in a possibly wrong(?) direction. Many thanks in advance!
