PROJECT / 04 FEATURE ARTICLE

Samantha AIAssistant

TYPE Embedded AISTATUS prototype

A Samantha-inspired voice assistant for the FoloToy AI Passport, with an OS1-style interface and real-time voice interaction.

01 Voice interaction

A voice-first Samantha-inspired interface

  1. 01Idle
  2. 02Listening
  3. 03Thinking
  4. 04Speaking
  5. 05Listening / Idle

02 Project overview

An embedded assistant for the AI Passport

Samantha AI Assistant is a voice-first, Samantha-inspired interface for the FoloToy AI Passport. Its OS1-style visual layer presents the assistant response as subtitles while keeping the screen focused on the live interaction.

User speech transcripts are intentionally hidden. The project builds on the device's existing XiaoZhi voice pipeline and cloud/device MCP foundation; physical Passport runtime validation is still pending.

03 Target hardware

A constrained embedded display and audio system

  • FoloToy AI Passport
  • ESP32-C3
  • 8 MB flash
  • No PSRAM
  • ST7789 / ST7789P3
  • 240×320 display
  • ES8311 audio codec
  • CW2017 fuel gauge
  • Pre-rendered firmware animations
  • 128×128
  • 20 FPS
  • Assistant response subtitles
  • User speech transcripts hidden

04 Implementation

A project-specific interaction and visual layer

The repository derives from XiaoZhi ESP32 v2.5.0. XiaoZhi provides the underlying voice, network, protocol, application, and MCP foundation, including the existing voice pipeline and device/cloud MCP support.

The Samantha interaction and visual layer is the project-specific implementation. Firmware animations are pre-rendered at 128×128 pixels and 20 FPS; assistant response subtitles are shown, while user speech transcripts are intentionally hidden.

The canonical software build has been validated with ESP-IDF 6.1. Runtime validation on the physical Passport is still pending, so this page makes no claim of successful hardware validation.

05 Interaction states

The assistant returns to listening

  1. 01

    Idle

    The embedded interface waits for the next voice interaction.

  2. 02

    Listening

    The existing XiaoZhi voice pipeline handles the incoming voice interaction; user speech transcripts remain hidden.

  3. 03

    Thinking

    The device and cloud voice/MCP foundation carries the assistant request and response.

  4. 04

    Speaking

    A pre-rendered firmware animation accompanies the response subtitle.

  5. 05

    Return

    After speaking, the interface returns to Listening or Idle.