FEMI

A social network built on voice.

  • iOS

2021

A business project, led in 2021 with Tanel Petelot, met at CentraleSupélec, with whom I later co-founded Emobot. We wanted a social network built on the voice, on saying (“femi”, in Greek): a little like Clubhouse, people who talk rather than write, brought closer by what they feel.

Two apps were born. FEMI, on iPhone: friends to follow, voice messages to which the photos of one’s album attach, voice stories, and each message analysed for its keywords and emotions, to bring together those whose moods match. And a research app, where volunteers read texts in a given emotion, to build a dataset of voices and teach AIs to recognise the emotion a voice carries.

The app was launched on the CentraleSupélec campus, then the project ran out of steam. It gave Emobot its idea and its impetus.

The list of conversations: each friend with their avatar, and the keywords of their last voice message scrolling in a bubble. Below, the welcome tour: “Chat with your friends”.
A conversation: the voice messages on a timeline, each with its keywords. Below, the welcome tour: “Share vocals with your friend”.
The research app’s welcome, in French: the FEMI team asks people to read about ten texts, each in an emotion, or to improvise. Samuel Lerman and Tanel Petelot, students at CentraleSupélec, explain the project, and that the recordings will be neither shared nor used commercially. A “Let’s go” button.
An emotion to act: “Mood : Joyeux” (happy) on an orange ground, the microphone button in the middle, “Récite” or “Improvise”, and the sentence to read, about a surf session.

What was done

  1. 01

    The FEMI app

    Friends to follow, voice messages laid on a timeline with the photos of one’s album, voice stories for one’s followers; each message analysed for its keywords and emotions, the mood on every profile, and matches by mood.

  2. 02

    The research app

    About ten texts to read, each in an emotion, or to improvise, to build a dataset of voices and teach AIs to recognise the emotion they carry. The recordings serve research only: never shared, never used commercially.

Isometric drawing of two phones facing each other. By step: friends, as small discs, around the microphone; a voice wave going from one to the other; circles widening from one phone, a story; a link joining the two phones; finally, the voice going down into a dataset.

How it works

  1. Scroll: the drawing follows.

  2. 01 / 05

    Add friends

    You discover people, follow them, and your friends gather around the microphone, with their voices and their stories. Your profile holds your friends, your mood and your photo album.

  3. 02 / 05

    Talk rather than write

    You touch the microphone and speak: each voice message lands on a timeline, in the conversation. The photos added to your album attach themselves to the message you send, and a long press on a message shows the ones that go with it.

  4. 03 / 05

    Tell a voice story

    A story is told out loud, for your followers, and each of them listens in their own time.

  5. 04 / 05

    Get closer through mood

    Each voice message and its photos are analysed: their keywords and emotions show in the preview, and everyone’s mood on their profile. You see what the other feels, and the app brings together those whose moods match.

  6. 05 / 05

    Lend your voice to research

    In the research app, volunteers read about ten texts, each in an emotion, or improvise: angry, joyful, sad, surprised, thoughtful, in love… eleven emotions, each with its colour. Their consent is one button, and the recordings, never shared nor sold, teach AIs to recognise the emotion a voice carries.

Specifications

Collection
Personal products
Date
2021
Context
An entrepreneurial project, with Tanel Petelot, when we were students at CentraleSupélec.
With
Tanel Petelot
Platforms
iOS