Open Source
Meet Avremi
דער ערשטער אָפֿענער ייִדישער קול
The first open-source neural text-to-speech voice for Yiddish. Type any Yiddish text and hear it spoken aloud — no download required.
Hear Avremi in Action
Avremi and other voices are available inside Loqal — the AI language learning app with real-time avatar conversations, flashcards, writing practice, and structured lessons in 10 languages.
About the Voice
Avremi is a neural text-to-speech voice trained on the REYD corpus — a collection of studio-quality Yiddish recordings produced at the University of Edinburgh and the University of Regensburg. The dataset features readings from late 19th and early 20th century Yiddish fiction, spoken by native speakers.
The voice model uses Piper VITS, a fast non-autoregressive architecture that produces natural-sounding speech without hallucination. It was fine-tuned from a German checkpoint, leveraging the phonological similarities between Yiddish and German while learning Yiddish-specific sounds from the training data.
Avremi handles all Yiddish orthographies — YIVO standard, Hasidic, undiacriticized, Soviet-era, and mixed text. Numbers, currencies, percentages, and times are automatically expanded into Yiddish words. Loshn-koydesh (Hebrew and Aramaic-origin) words are phonetically respelled for correct pronunciation.
The Yiddish Language
Yiddish is a Germanic language written in Hebrew script, spoken for over a thousand years by Ashkenazi Jews across Central and Eastern Europe. It blends Middle High German with Hebrew, Aramaic, and Slavic elements into a language of remarkable expressiveness — equal parts humor, wisdom, and heartbreak.
At its peak before World War II, Yiddish was the daily language of roughly 11 million people. It had a flourishing literary tradition, a vibrant press with dozens of daily newspapers, a thriving theater scene, and was the lingua franca of Jewish communities from Vilna to Buenos Aires.
Yiddish was — and remains — more than a language. It carries within it a civilization: the jokes, the prayers, the lullabies, the arguments, the marketplace banter, and the intellectual discourse of a people who built a portable culture across borders and centuries.
Yiddish Dialects
Yiddish has three major dialect groups, each shaped by centuries of life in different regions of Europe. While speakers of all dialects can understand each other, the differences in vowels, stress, and vocabulary give each dialect its own unmistakable character.
The standard literary language (YIVO standard) is based primarily on the Lithuanian dialect, but all three dialects remain living traditions today.
Litvish (Lithuanian)
Lithuania, Latvia, Belarus
The prestige dialect of Yiddish scholarship and literature. Distinguished by its 'ey' and 'oy' diphthongs. The basis for YIVO standard orthography. Associated with the great yeshivas of Vilna, Volozhin, and Mir.
Poylish (Polish)
Central Poland, Galicia
Known for its distinctive vowel shifts and melodic intonation. The dialect of Warsaw, Lodz, and Krakow — the heartland of pre-war Yiddish culture, theater, and daily life. The most widely spoken dialect before the war.
Galitzianer (Ukrainian)
Western Ukraine, Romania, Hungary
Features unique vowel qualities and Slavic-influenced intonation. The dialect of Lemberg (Lviv), Czernowitz, and the Carpathian region. Rich in Slavic loanwords and associated with Hasidic traditions.
Avremi's voice is trained on recordings in the Polish dialect — the speech patterns of central Poland, the most widely spoken Yiddish dialect before the Holocaust. This gives Avremi the warm, melodic quality characteristic of everyday Yiddish as it was spoken by millions.
Before and After
Before 1939, Yiddish was one of the most widely spoken languages in Europe. Eleven million speakers. Schools, universities, newspapers, political parties, radio stations, film studios — all operating in Yiddish. In cities like Warsaw, Vilna, and Odessa, Yiddish was the sound of the street.
The Holocaust destroyed this world. Six million Jews were murdered, and with them the majority of the world's Yiddish speakers. Entire dialect regions were silenced. The language went from a thriving vernacular of millions to an endangered language in a single generation.
But Yiddish did not die. It survived in the Hasidic communities of Brooklyn, Antwerp, Jerusalem, and Montreal, where it remains a living first language for hundreds of thousands of people. It survived in academic programs at universities around the world. It survived in the songs, stories, and memories passed from grandparents to grandchildren.
Today, Yiddish is experiencing a quiet renaissance. New books are being published, new courses are being taught, and new technology — like this voice — is being built to ensure that the language not only survives but thrives in the digital age.
Where to Learn More
Resources for learning Yiddish and exploring its culture.
YIVO Institute for Jewish Research
The global center for Yiddish scholarship, founded in Vilna in 1925. Offers courses, archives, and the definitive Yiddish orthography standard.
REYD Speech Corpus
The Yiddish speech dataset that made Avremi possible. Studio recordings of native speakers reading classic Yiddish literature.
The Forward (Forverts)
Founded in 1897, the Forward is the oldest Yiddish-language publication still in print. Now publishes in both English and Yiddish.
Yiddish Book Center
A cultural institution dedicated to the recovery and celebration of Yiddish literature. Over 40,000 digitized titles available free online.
Open Source
Avremi is freely available under the CC BY-SA 4.0 license. The ONNX model runs on CPU — no GPU needed — and works with Piper, sherpa-onnx, and other VITS runtimes.
Built With
- REYD — Yiddish speech corpus (University of Edinburgh / University of Regensburg)
- Piper — Neural text-to-speech (Rhasspy project)
- yiddish — Yiddish text processing (Jochanan Stenzel)