Tessera
All work
Case · Wispy

English from shows and news — with an AI conversation partner

An English app where AI generates almost all the learning content on LangGraph — words in context, stories, exercises — plus a voice partner.

EdTechAIMobileBackendAIiOSAndroid
App Store
English from shows and news — with an AI conversation partner
5,000+
Words · Oxford 5000
A1–C1
CEFR levels
AI voice
Speaking practice
01
Challenge

The context

Words from shows and videos are forgotten by the next day, and textbooks are cut off from real speech. Ordinary apps drill through lists with no context — dull, and nothing sticks.

02
Approach

The idea

We built an app where AI is the engine: almost all the learning content is generated on LangGraph for the user’s level. Words are shown in two contexts — “in the show” and “in life” — daily stories and exercises assemble automatically, and an ElevenLabs voice partner lets you speak out loud. Spaced repetition brings words back in time so they stick.

03 · What we built

What we built

01

A multi-agent graph on LangGraph

Content is built not by a single prompt but by a graph of agents: an architect plans, a generator writes, a critic checks the content, a translator translates, a second critic proofreads the translation, and if phrases are missing from the dictionary, a separate step fills it in.

02

A dictionary from content

Save a word and AI explains it in two contexts — “in the show” and “in life.” Alongside are ready Oxford 5000 lists by level, A1–C1.

03

AI conversation partner

Speak English out loud: a LangGraph agent leads the dialogue and ElevenLabs voices it — real speaking practice without a teacher.

04

Smart practice

Spaced repetition and reminders bring words back at the right moment so your vocabulary doesn’t fade.

05

Daily Drops

AI generates short stories from fresh news at the user’s level — daily reading practice you actually want to open.

04 · Gallery

Screens

Voice AI partner: live practice and speech feedback
Voice AI partner: live practice and speech feedback
Exercises: AI finds and explains mistakes right in the text
Exercises: AI finds and explains mistakes right in the text
Profile and progress: comprehension, speaking, writing
Profile and progress: comprehension, speaking, writing
05 · Stack
React NativePythonLangGraphClaude APIElevenLabsPostgreSQL
Tessera didn’t build “yet another vocabulary app” — they built an engine that explains a word through content and lets you talk to AI. That’s what keeps people in the app.
Founder, Wispy
06 · Result

In Wispy, AI is the engine: almost all the learning content is generated on LangGraph for the user’s level, and an ElevenLabs voice partner gives real speaking practice. English is built into what people already watch and read.

Next case
BS Control

School automation: from the library to safety

School automation: from the library to safety

Need a similar project?