---
language:
- en
- la
---
# Voynich Manuscript Transliterations
This dataset provides **8 scholarly transliterations** of the **Voynich
Manuscript**, compiled by **René Zandbergen** and made available at
[voynich.nu](https://www.voynich.nu/).
## About the Voynich Manuscript
The Voynich Manuscript (Beinecke MS 408) is a 15th-century codex, carbon-dated
to 1403-1438, written in an unknown script by an unknown author. Often called
"the world's most mysterious book," it contains approximately 240 vellum pages
filled with text in an undeciphered writing system, accompanied by illustrations
of unknown plants, astronomical diagrams, and human figures.
The manuscript was rediscovered in 1912 by rare book dealer Wilfrid Voynich and
is currently housed at Yale University's Beinecke Rare Book & Manuscript
Library.
### Included Transliterations
See Zandbergen's [README](data/README.md) (from
https://www.voynich.nu/data/000_README.txt) for more info.
| Dataset ID | Name | Alphabet | Description | Loci Count |
|-----------|------|----------|-------------|------------|
| **CD2a-n** | Currier-D'Imperio | Currier | Original Currier-D'Imperio transliteration | ~4,000+ |
| **FG2a-n** | FSG | FSG | Original FSG transliteration | ~4,000+ |
| **GC2a-n** | GC v101 | v101 | GC's v101 transliteration with IVTFF locus conventions | ~4,000+ |
| **IT2a-n** | Takahashi (Interlinear) | Basic EVA | Takeshi Takahashi transliteration from Stolfi interlinear (1999) | ~4,000+ |
| **RF1b-e** | Reference (EVA) | Basic EVA | Reference transliteration (GC+ZL combination) in EVA | ~4,200 |
| **RF1b-er** | Reference (EVA-r) | Basic EVA | Reference transliteration with revised locus definitions | ~4,200 |
| **VT0e-n** | Takahashi (Voynichese.com) | Basic EVA | Takeshi Takahashi transliteration from voynichese.com | ~4,000+ |
| **ZL3b-n** | Zandbergen-Landini | EVA (extended) | Complete Zandbergen-Landini transliteration (1999) with high-ascii EVA extensions | ~5,400 |
## Usage
### Downloading the HuggingFace Dataset
```python
from datasets import load_dataset
dataset = load_dataset("AncientLanguages/Voynich")
texts = dataset['train'].to_pandas()
# Set index
texts.set_index(['source', 'source_name', 'alphabet', 'description', 'page', 'page_comments', 'locus_num', 'text'], inplace=True)
texts
```
Output
|
|
|
|
|
|
|
|
B |
C |
F |
H |
I |
L |
P |
Q |
X |
locus_id |
locator |
locus_type |
transcriber |
| source |
source_name |
alphabet |
description |
page |
page_comments |
locus_num |
text |
|
|
|
|
|
|
|
|
|
|
|
|
|
| CD2a-n |
Currier-D'Imperio |
Currier |
Original Currier-D'Imperio transliteration |
f1r |
NaN |
1 |
VAS92 9FAE AR APAM ZOE ZOR9 QOR92 9 FOR ZOE89 |
1 |
1 |
a |
1 |
T |
A |
A |
A |
V |
<f1r.1,@P0> |
@ |
P0 |
None |
| 2 |
2OR9 XAR O R 9 FAN ZPAM ZAR AR QAR QAR 8AD |
1 |
1 |
a |
1 |
T |
A |
A |
A |
V |
<f1r.2,+P0> |
+ |
P0 |
None |
| 3 |
29AU ZCF9 OR 9FAM ZO8 QOAR9 QR 8ARAM 29 |
1 |
1 |
a |
1 |
T |
A |
A |
A |
V |
<f1r.3,+P0> |
+ |
P0 |
None |
| 4 |
2OM OPCC9 OPCOR 2OEOP9 QAR 8AM OFAM OE OFAD |
1 |
1 |
a |
1 |
T |
A |
A |
A |
V |
<f1r.4,+P0> |
+ |
P0 |
None |
| 5 |
2AT 9 SCAR QAM WAR YAM |
1 |
1 |
a |
1 |
T |
A |
A |
A |
V |
<f1r.5,+P0> |
+ |
P0 |
None |
| ... |
... |
... |
... |
... |
... |
... |
... |
... |
... |
... |
... |
... |
... |
... |
... |
... |
... |
... |
... |
... |
| ZL3b-n |
Zandbergen-Landini |
EVA (extended) |
Complete Zandbergen-Landini transliteration (1999) with EVA alphabet |
f116r |
page 234\nCurrier's language B\ntext only / stars\nLight star, 7 points |
47 |
osain shky qorain chckhey qokey lkechy okeey okal chedkaly |
1 |
None |
z |
3 |
S |
B |
W |
T |
None |
<f116r.47,+P0> |
+ |
P0 |
None |
| 48 |
sykar ain olkeey dainchey qokar chey dain y otan otain oly |
1 |
None |
z |
3 |
S |
B |
W |
T |
None |
<f116r.48,+P0> |
+ |
P0 |
None |
| 49 |
sysor shey qokey okeolan chey qol or cheey qor arom ol lkan |
1 |
None |
z |
3 |
S |
B |
W |
T |
None |
<f116r.49,+P0> |
+ |
P0 |
None |
| 50 |
sodal ch al chcthy chckhy qol ain ary |
1 |
None |
z |
3 |
S |
B |
W |
T |
None |
<f116r.50,+P0> |
+ |
P0 |
None |
| f116v |
page 235\nhas "key-like" sequence\nhas extraneous writing mixed with voynich script\neva in < >, approximately:\np*clabc* nomon nucor*v\n+ nuchicon oladab<yd> + multo<d> + tc + c<h>vc + porta<d> + n +\nfix + marix + mocix + vix + ahra + maria + |
1 |
oror sheey |
1 |
None |
z |
3 |
T |
None |
X |
T |
V |
<f116v.1,@Lx> |
@ |
Lx |
None |
31045 rows × 13 columns
### Getting a specific transliteration
```python
zl = texts.loc['ZL3b-n']
zl
```
Output
|
|
|
|
|
|
|
B |
C |
F |
H |
I |
L |
P |
Q |
X |
locus_id |
locator |
locus_type |
transcriber |
| source_name |
alphabet |
description |
page |
page_comments |
locus_num |
text |
|
|
|
|
|
|
|
|
|
|
|
|
|
| Zandbergen-Landini |
EVA (extended) |
Complete Zandbergen-Landini transliteration (1999) with EVA alphabet |
f1r |
page 1\ntext only\nCurrier's Language A, hand 1\nThere is a key-like sequence in the right margin apparently added after\nthe main text was written. Unclear in Petersen copies, but seems to be\nthe roman alphabet. It starts at the end of the second line and it\nis not well aligned with the rest of the lines: a,b,c,d,e,f,g,h,\ni, (j is missing) k, (l,m,n are not readable),o,p,q,r\nThere is non-Voynich text at bottom of folio: the signature\nof Jacobus de Tepenecz\nAt the end of the paragraphs there are further words (right justified)\nas if they were the name of the author in a quotation.\nNext to the end of the 1st line there is a thick non-Voynich symbol\nresembling an underlined 29 (commented as @254; )\nThe 2nd and 3rd paragraphs start with thick non-Voynich symbols. |
1 |
fachys ykal ar ataiin shol shory cthres y kor sholdy |
1 |
1 |
a |
1 |
T |
A |
A |
A |
V |
<f1r.1,@P0> |
@ |
P0 |
None |
| 2 |
sory ckhar or y kair chtaiin shar ase cthar cthar dan |
1 |
1 |
a |
1 |
T |
A |
A |
A |
V |
<f1r.2,+P0> |
+ |
P0 |
None |
| 3 |
syaiir sheky or ykaiin shod cthoary cthes daraiin sy |
1 |
1 |
a |
1 |
T |
A |
A |
A |
V |
<f1r.3,+P0> |
+ |
P0 |
None |
| 4 |
soiin oteey oteos roloty cthiar daiin okaiin or okan |
1 |
1 |
a |
1 |
T |
A |
A |
A |
V |
<f1r.4,+P0> |
+ |
P0 |
None |
| 5 |
sair y chear cthaiin cphar cfhaiin |
1 |
1 |
a |
1 |
T |
A |
A |
A |
V |
<f1r.5,+P0> |
+ |
P0 |
None |
| ... |
... |
... |
... |
... |
... |
... |
... |
... |
... |
... |
... |
... |
... |
... |
... |
... |
| f116r |
page 234\nCurrier's language B\ntext only / stars\nLight star, 7 points |
47 |
osain shky qorain chckhey qokey lkechy okeey okal chedkaly |
1 |
None |
z |
3 |
S |
B |
W |
T |
None |
<f116r.47,+P0> |
+ |
P0 |
None |
| 48 |
sykar ain olkeey dainchey qokar chey dain y otan otain oly |
1 |
None |
z |
3 |
S |
B |
W |
T |
None |
<f116r.48,+P0> |
+ |
P0 |
None |
| 49 |
sysor shey qokey okeolan chey qol or cheey qor arom ol lkan |
1 |
None |
z |
3 |
S |
B |
W |
T |
None |
<f116r.49,+P0> |
+ |
P0 |
None |
| 50 |
sodal ch al chcthy chckhy qol ain ary |
1 |
None |
z |
3 |
S |
B |
W |
T |
None |
<f116r.50,+P0> |
+ |
P0 |
None |
| f116v |
page 235\nhas "key-like" sequence\nhas extraneous writing mixed with voynich script\neva in < >, approximately:\np*clabc* nomon nucor*v\n+ nuchicon oladab<yd> + multo<d> + tc + c<h>vc + porta<d> + n +\nfix + marix + mocix + vix + ahra + maria + |
1 |
oror sheey |
1 |
None |
z |
3 |
T |
None |
X |
T |
V |
<f116v.1,@Lx> |
@ |
Lx |
None |
4197 rows × 13 columns
### Getting a specific page
```python
# Get list of all pages
pages = zl.index.get_level_values('page').unique()
# Access Nth page
f1r = zl.xs(pages[0], level='page')
# Access named page
f1r = zl.xs('f1r', level='page')
# Get all text from a page as a single string
print("\n".join(f1r['text']))
```
```
fachys ykal ar ataiin shol shory cthres y kor sholdy
sory ckhar or y kair chtaiin shar ase cthar cthar dan
syaiir sheky or ykaiin shod cthoary cthes daraiin sy
soiin oteey oteos roloty cthiar daiin okaiin or okan
sair y chear cthaiin cphar cfhaiin
ydaraishy
odar c'y shol cphoy oydar sh s cfhoaiin shodary
yshey shody okchoy otchol chocthy os chy dain chor kos
daiin shos cfhol shody
...
```
### IVTFF Format
The **IVTFF (Intermediate Voynich Transliteration File Format)** is a
standardized format designed by René Zandbergen to represent all publicly
available Voynich manuscript transliterations in a consistent manner.
Full format specification: [IVTFF Format Documentation
(PDF)](http://www.voynich.nu/software/ivtt/IVTFF_format.pdf)
## Citation & Attribution
### Data Source
All transliteration files are sourced from the comprehensive Voynich manuscript
research repository maintained by **René Zandbergen**:
**Website**: [https://www.voynich.nu/](https://www.voynich.nu/) **Data
Repository**: [https://www.voynich.nu/data/](https://www.voynich.nu/data/)
### Recommended Citation
When using this dataset, please cite:
```
Zandbergen, René (2023). Voynich Manuscript Transliterations in IVTFF Format.
Retrieved from https://www.voynich.nu/data/
Individual transliterations:
- Zandbergen-Landini (ZL): Zandbergen, R. & Landini, G. (1999)
- Currier-D'Imperio (CD): Currier, P. & D'Imperio, M.E.
- FSG (FG): First Study Group
- GC: Gabriel Clausi (v101)
- Takahashi (IT, VT): Takeshi Takahashi
- Reference (RF): Automated combination of GC and ZL transliterations
```
### Original Contributors
- **René Zandbergen**: IVTFF format design, Zandbergen-Landini transliteration,
format standardization
- **Gabriel Landini**: Zandbergen-Landini transliteration
- **Prescott Currier & Mary D'Imperio**: Currier-D'Imperio transliteration
- **First Study Group (FSG)**: FSG transliteration
- **Gabriel Clausi**: GC v101 transliteration
- **Takeshi Takahashi**: IT and VT transliterations
- **Jorge Stolfi**: Interlinear file compilation (1999)
- **Jeremie Bornais**: Original Python IVTFF parser
([ivtff-py](https://github.com/jere-mie/ivtff-py))
### Related Resources
- [IVTFF Format
Specification](http://www.voynich.nu/software/ivtt/IVTFF_format.pdf)
- [Voynich MS Transcription Overview](https://www.voynich.nu/transcr.html)
- [Beinecke Digital Library - Voynich
Manuscript](https://beinecke.library.yale.edu/collections/highlights/voynich-manuscript)
- [Wikipedia: Voynich
Manuscript](https://en.wikipedia.org/wiki/Voynich_manuscript)
## License
The Voynich Manuscript itself is in the public domain. The transliteration data
has been made publicly available by René Zandbergen and the original
contributors for research purposes.
**Note**: Please respect the work of the transliterators by providing
appropriate attribution when publishing research using this data.
## Updates & Versions
- **IVTFF Format**: Version 2.0 (stable since 2023)
- **Last Data Update**: Files range from 2023-11-03 to 2025-06-25
- **Dataset Compilation**: 2025
## Contact & Contributions
For questions about the IVTFF format or the original transliteration data, visit
[voynich.nu](https://www.voynich.nu/).
For issues with this dataset compilation or the Python parser, please open an
issue on the repository.
---
**Keywords**: Voynich Manuscript, cryptography, medieval manuscripts,
undeciphered scripts, historical linguistics, IVTFF, EVA alphabet,
transliteration