DEMO / RECORDING
Overview & result
Masahiro Chaen built an app that listens to overseas sales meetings every 4 seconds, using Whisper for transcription and Jev for sub-50ms tactical objection classification and response card triggering.
Show the original post (Japanese)
話題の「Jev」で商談中に「次の一手」をリアルタイムで分析するアプリを作成。
圧倒的速度がウリなので、即時性が求められる分野で大活躍。
海外の商談動画から、
・英語がリアルタイムで文字になる
・日本語の翻訳と議事録が流れる
・図解カードが次々に増える https://t.co/jtoyVRCUjH
How Jev fits in the loop
- Ingest live audio chunks continuously without blocking audio capture.
- Transcribe and translate dialogue into structured candidate debate points.
- Jev evaluates client intent, purchase friction, and next objection strategy in 45ms.
- Display contextual battle cards instantly on the representative's screen.
How to reproduce
- Review video showing live English negotiation transcribed and analyzed every 4 seconds.
- Pass extracted dialogue turns to Jev Choice to select objection counter-strategies.
- Surface matching visual response cards to the sales representative on screen.
Why this build matters
Brings real-time AI assistance into live interpersonal conversation where multi-second LLM generation is completely unusable.
Limitations
- Requires low-latency speech-to-text pipeline (Whisper / Deepgram).
- Limited to predefined tactical cards; novel negotiations need human judgment.
- Author reported community project.