Skip to contentEmail Sergio
Back to home
Available

Escucha Dual: hold a real conversation in a language you do not speak.

Demonstration

See it running.

A recorded demonstration of a live two-way conversation running through the tool.

The problem

Why I built it.

Translation tools give you sentences. They do not give you a conversation. You paste, you wait, you read something stiff aloud, and the other person is already three sentences ahead. The exchange never becomes a conversation because the tool sits in the middle of it instead of inside it.
Intended users. Anyone who has to talk with someone who shares no common language: a meeting with a counterpart who speaks only Chinese, Japanese, or Korean, or a friend you would otherwise never really talk to.

Design

Decisions I made and why.

01

Both directions, one flow

Listening to one side and translating for the other makes each person wait twice. Both directions run in the same loop so the conversation keeps its rhythm.

02

Say it, do not read it

Reading a transcript aloud in a language you cannot pronounce breaks the illusion instantly. Spoken output keeps the exchange feeling like speech.

03

Answers that follow the conversation

The point was never producing correct isolated sentences. Each turn has to make sense as a reply to the one before it, or the other person notices immediately.

Disclosure

What I built, what I would test, what it cannot claim.

What I built+
  • A dual listening setup that follows both sides of a live conversation
  • Spoken output in the target language rather than text to read aloud
  • Support for Chinese, Japanese, and Korean
  • A turn structure designed so both speakers can keep a normal rhythm
What I would test next+
  • Native speakers rating how natural each translated turn sounded, blind to the source
  • Measuring the delay between turns, since latency is what breaks the feeling of conversation
  • Trying it in a noisy room and in a meeting with technical vocabulary
Limitations+
  • The evidence here is one recorded demonstration and one real conversation with a Chinese friend, not a measured evaluation.
  • I speak almost no Chinese, so I could not independently verify nuance. My friend's reaction is my only check.
  • It has not been tested across accents, noisy rooms, or specialised vocabulary.
  • This is a communication aid, not a substitute for learning a language.