Universal Talking Lab

Lots of languages.
Each taught its own way.

What is easy in one language is a wall in another. Talkverse designs the method for each pair — the language you speak and the one you are learning — then levels it by where real conversation actually gets harder.

Browse languages How leveling works
21language apps
17live today
17corpora compared
4measured levels, R1–R4

Languages

Pick a language. Get a method built for it.

Numbers under each app are its adopted cliff boundaries — the word rank at which conversation coverage reaches 55, 60, 70 and 80 percent.

РусскийLive

Russian

Close the loop on Russian cases and verbs with about 300 words — minimum vocabulary, maximum sentences.

R1156R2251R3685R41985
中文Live

Chinese

209 characters unlock 89% of spoken Chinese. Built around the character cliff and Sino-Korean readings.

R1433R2616R31238R42500
日本語Live

Japanese

Kana, kanji and particles mapped one-to-one onto Korean, backed by a kanji, reading and word database.

R1294R2437R3998R42436
한국어Live

Korean

For English speakers: 19 particles and endings plus 2,500 frequent words, straight into Seoul conversation.

R1100R2500R31500R43000
ไทยLive

Thai

Bangkok speech organized around 44 consonants, five tones and the polite particles ครับ / ค่ะ.

R199R2137R3286R4653
МонголLive

Mongolian

Same word order as Korean, cases that map to particles, no articles — from travel talk to a first date.

R1120R2210R3230R4250
Bahasa IndonesiaLive

Indonesian

Fourteen affixes turn one root into thousands of words. Toggle to Malay in the same app.

R1120R220R3—R4—
EspañolLive

Spanish

Conjugation, agreement and pronunciation — the three axes of Spanish, taught in order of frequency.

R1190R2200R3—R4—
FrançaisLive

French

Conversation, grammar and speaking practice in one app.

Cliff not yet measured
ItalianoLive

Italian

Conversation, grammar and speaking practice in one app.

Cliff not yet measured
العربيةLive

Arabic

Right to left: Egyptian colloquial conversation and the 5,000 most frequent words.

Cliff not yet measured
हिन्दीLive

Hindi

Conversation-first Hindi, starting from a 40-turn first dialogue with Hangul pronunciation.

R1229R2365R3944R42500
TürkçeLive

Turkish

Conversation-first Turkish, starting from a 40-turn first dialogue with Hangul pronunciation.

R1520R2722R31345R42500
MagyarLive

Hungarian

Conversation-first Hungarian, starting from a 40-turn first dialogue with Hangul pronunciation.

Cliff not yet measured
PortuguêsLive · draft

Portuguese

Conversation, grammar and speaking practice in one app.

Cliff not yet measured
DeutschLive · draft

German

Conversation, grammar and speaking practice in one app.

Cliff not yet measured
فارسیLive · draft

Persian

An Indo-European language in Arabic script — five dialogues, ten patterns and forty words to begin.

R1415R2603R31216R42500
Tiếng ViệtComing soon

Vietnamese

Southern (Saigon) Vietnamese built on forms of address and mood particles.

R189R2115R3199R4362
EnglishComing soon

English

Five minutes a day on the clips you choose — hear the hinges first so long sentences never fall apart.

R1154R2236R3639R43000
မြန်မာComing soon

Burmese

App scaffold and first Android build — content being written from scratch.

R1267R2366R3699R41425
PolskiPlanned

Polish

Slavic cases mapped onto Korean particles, on the same backbone as the Russian app. Planned.

Cliff not yet measured

Method

Levels drawn from real speech, not textbook grades.

Count every word in a large conversation corpus, rank by frequency, and accumulate. Coverage climbs steeply, then flattens — the bends are where learning gets hard. Talkverse cuts its levels at those bends, and grades each sentence by its rarest word.

R1 · 55%R2 · 60%R3 · 70%R4 · 80%101001,000
English · top-N words vs. share of spoken tokens
Frosted glass staircase that rises gently, then steeply, then levels off — the shape of a frequency cliff

Four levels

Gentle, then steep, then flat.

A few hundred words carry most of a conversation. Then each new word buys less and less. The four levels follow that shape.

R1 · 55%

Conversation skeleton

Function words, pronouns, particles, core verbs — cases for Russian, particles for Japanese.

R2 · 60%

Everyday core

Negation, tense, connectives and the verbs you need every day.

R3 · 70%

The real cliff

The bulk of everyday vocabulary — emotions, relationships, routines.

R4 · 80%

Broad vocabulary

Common nouns and adjectives. Sentences with rarer words are left out.

Design

Built for the pair, not just the language.

For a Korean speaker, Japanese particles map one-to-one — but Russian cases and Thai tones are walls. Each app starts from the learner's language and attacks the target language's own difficulties.

Four tiles in different materials — frosted glass, ceramic, acrylic, brushed metal — each carrying a letter: あ, 中, 한, Я

Script

Hangul pronunciation guides, kana and kanji mapped to Sino-Korean readings, right-to-left layouts for Arabic and Persian.

Sound

Five tones for Thai, stress for Russian, Hangul pronunciation for Hindi and Turkish, and short daily listening built from clips you choose.

Grammar

Case systems, affixes and verb conjugation introduced only when the frequency data says you will meet them.

Start with the words that matter most.

Learn offline. Apps update themselves in place.

Browse languages